LANA

lana

Authors: (1) Renato P. dos Santos, CIAGE – Centre for Generative Artificial Intelligence in Cognition and Education. Table of Links Abstract and Introduction Materials And Methods Results and Analyses Prompts and generated texts Conceptualizing chemical reactions Deepening on understanding of chemical reactions Question about combustion Question about a graph of gases turning into water over time Question about the difference between atoms, molecules, and moles Deepening on the concept of mole Question about changing of state Question about an animated representation of water molecules undergoing phase changes Question about plasma, a state of matter Question about chemical bondings Question about illustration of chemical bonds Question about the essence of the type of chemical bonding Further analysis Conclusions Limitations of the study and possible future studies Author Contributions, Conflicts of interest, Acknowledgements, and References Limitations of the study and possible future studies Despite its inherent limitations, including its single case design and the potential for bias, the exploratory depth of the study uncovered hidden potential within these systems, even amid serious concerns about generalizability. Future research could include: • Refining the crafting of prompts. • Exploring new features of these and other GenAIbots being introduced with increasing frequency. • Establishing standardised benchmarks to evaluate and compare chatbots and AI systems' performance, accuracy, and reliability. • Conducting long-term studies to observe the evolution of chatbots' capabilities and their impact on user interactions over time. • Conducting research with real students in classroom settings and beyond to assess these AI systems' practical educational applications and challenges. • Investigating chatbots' learning and adaptation capabilities to individual user needs and preferences over time. • Researching the integration of multimodal inputs (e.g., text, voice, image) to enhance chatbot capabilities and user interaction experiences. When implementing GenAIbots in Chemistry learning, it is crucial to evaluate benefits and drawbacks judiciously, ensuring accurate information delivery and considering the implications of reduced human interaction. These concerns may be alleviated by integrating GenAIbots with other educational tools or activities promoting collaborative dialogue among learners. Author Contributions The author confirms sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation. Conflicts of interest There are no conflicts to declare. Acknowledgements The authors warmly acknowledge Melanie Swan for her invaluable suggestion, which led to the transition from using the term 'objects-to-think-with' to 'agents-to-thinkwith'. References Adiguzel, T., Kaya, M. H., & Cansu, F. K. (2023). Revolutionising education with AI: Exploring the transformative potential of ChatGPT. Contemporary Educational Technology, 15(3), ep429. https://doi.org/10.30935/cedtech/13152 Baidoo-Anu, D., & Owusu Ansah, L. (2023). Education in the Era of Generative Artificial Intelligence (AI): Understanding the Potential Benefits of ChatGPT in Promoting Teaching and Learning. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4337484 Bardin, L. (1977). L’analyse de contenu. PUF - Presses Universitaires de France. Bitzenbauer, P. (2023). ChatGPT in physics education: A pilot study on easy-to-implement activities. Contemporary Educational Technology, 15(3), ep430. https://doi.org/10.30935/cedtech/13176 Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., … Amodei, D. (2020). Language Models are Few-Shot Learners. ArXiv, 2005.14165. http://arxiv.org/abs/2005.14165 Castro Nascimento, C. M., & Pimentel, A. S. (2023). Do Large Language Models Understand Chemistry? A Conversation with ChatGPT. Journal of Chemical Information and Modeling, 63(6), 1649–1655. https://doi.org/10.1021/acs.jcim.3c00285 David, E. (2023, September 21). Microsoft to add DALL-E 3 to Bing Chat. The Verge; Vox Media. https://www.theverge.com/2023/9/21/23873690/microsoft-new-ai-features-bingsearch-shopping-dall-e-3 Dewi, C. A., Pahriah, P., & Purmadi, A. (2021). The Urgency of Digital Literacy for Generation Z Students in Chemistry Learning. International Journal of Emerging Technologies in Learning, 16(11), 88–103. https://doi.org/10.3991/ijet.v16i11.19871 Dunlop, L., Hodgson, A., & Stubbs, J. E. (2020). Building capabilities in chemistry education: happiness and discomfort through philosophical dialogue in chemistry. Chemistry Education Research and Practice, 21(1), 438–451. https://doi.org/10.1039/C9RP00141G Flavell, J. H. (1976). Metacognitive aspects of problem-solving. In L. B. Resnick (Ed.), The Nature of Intelligence (pp. 231–236). Lawrence Erlbaum. Floridi, L., & Chiriatti, M. (2020). GPT-3: Its Nature, Scope, Limits, and Consequences. Minds and Machines, 30(4), 681–694. https://doi.org/10.1007/s11023-020-09548-1 Franciscu, S. (2023). ChatGPT: A Natural Language Generation Model for Chatbots. https://doi.org/10.13140/RG.2.2.24777.83044 Gregorcic, B., & Pendrill, A.-M. (2023). ChatGPT and the frustrated Socrates. Physics Education, 58(3), 035021. https://doi.org/10.1088/1361-6552/acc299 Latour, B. (1991). Nous n’avons jamais été modernes: Essai d’anthropologie symétrique. La Découverte. Leon, A. J., & Vidhani, D. (2023). ChatGPT Needs a Chemistry Tutor, Too. Journal of Chemical Education. https://doi.org/10.1021/acs.jchemed.3c00288 Liu, F., Budiu, R., Zhang, A., & Cionca, E. (2023, October 1). ChatGPT, Bard, or Bing Chat? Differences Among 3 Generative-AI Bots. https://www.nngroup.com/articles/ai-botcomparison/ Metz, C., & Hsu, T. (2023, September 20). ChatGPT Can Now Generate Images, Too. New York Times - Technology. https://www.nytimes.com/2023/09/20/technology/chatgpt-dalle3-images-openai.html Mishra, A., Soni, U., Arunkumar, A., Huang, J., Kwon, B. C., & Bryan, C. (2023). PromptAid: Prompt Exploration, Perturbation, Testing and Iteration using Visual Analytics for Large Language Models. ArXiv, 2304.01964. https://doi.org/10.48550/arXiv.2304.01964 Mollick, E. R. (2023, April 26). A guide to prompting AI (for what it is worth): A little bit of magic,but mostly just practice. One Useful Thing Blog. https://www.oneusefulthing.org/p/a-guidetoprompting-ai-for-what Okonkwo, C. W., & Ade-Ibijola, A. (2021). Chatbots applications in education: A systematic review. Computers and Education: Artificial Intelligence, 2, 100033. https://doi.org/10.1016/j.caeai.2021.100033 OpenAI. (2023). GPT-4 Technical Report. OpenAI. https://cdn.openai.com/papers/gpt-4.pdf Papert, S. A. (1980). Mindstorms - Children, Computers and Powerful Ideas. Basic Books. http://www.arvindguptatoys.com/arvindgupta/mindstorms.pdf Permatasari, M. B., Rahayu, S., & Dasna, W. (2022). Chemistry Learning Using Multiple Representations: A Systematic Literature Review. J.Sci.Learn.2022, 5(2), 334–341. https://doi.org/10.17509/jsl.v5i2.42656 Pimentel, A., Wagener, A., Silveira, E. F. da, Picciani, P., Salles, B., Follmer, C., & Jr., O. N. O. (2023). Challenging ChatGPT with Chemistry-Related Subjects. ChemRxiv. https://doi.org/10.26434/chemrxiv-2023-xl6w3 Schlosser, M. (2019). Agency. In The Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/archives/win2019/entries/agency/ Swan, M. (2015, January 21). We Should Consider the Future World as one of Multi-Species Intelligence. Response to The Edge Question 2015: What Do You Think About Machines That Think?; Edge.org. http://edge.org/response-detail/26070 Taylor, C. A., Hogarth, H., Hacking, E. B., & Bastos, E. (2022). Posthuman Object Pedagogies: Thinking with Things to Think with Theory for Innovative Educational Research. Cultural and Pedagogical Inquiry, 14(1), 206–221. https://doi.org/10.18733/cpi29662 Timilsena, N. P., Maharjan, K. B., & Devkota, K. M. (2022). Teachers’ And Students’ Experiences In Chemistry Learning Difficulties. Journal of Positive School Psychology, 6(10), 2856–2867. https://www.journalppw.com/index.php/jpsp/article/view/13764 Tümay, H. (2016). Reconsidering learning difficulties and misconceptions in chemistry: emergence in chemistry and its implications for chemical education. Chemistry Education Research and Practice, 17(2), 229–245. https://doi.org/10.1039/C6RP00008H Turkle, S. (1984). The Second Self: Computers and the Human Spirit. Simon and Schuster. Yin, R. K. (2011). Applications of Case Study Research. SAGE Publications, Inc. This paper is under CC BY-SA 4.0 DEED license. available on arxiv

Overcoming Limitations in AI Chatbot Research for Future Educational Impact

ভবিষ্যতের শিক্ষাগত প্রভাবের জন্য এআই চ্যাটবট গবেষণায় সীমাবদ্ধতা অতিক্রম করা

Überwindung von Einschränkungen in der KI-Chatbot-Forschung für zukünftige Bildungseffekte

Superar las limitaciones en la investigación de chatbots de IA para lograr un impacto educativo futuro

Surmonter les limites de la recherche sur les chatbots IA pour un impact éducatif futur

भविष्य के शैक्षिक प्रभाव के लिए AI चैटबॉट अनुसंधान में सीमाओं पर काबू पाना

将来の教育への影響のために AI チャットボット研究の限界を克服する

미래 교육적 영향을 위한 AI 챗봇 연구의 한계 극복

Superando as limitações na pesquisa do AI Chatbot para um impacto educacional futuro

Преодоление ограничений в исследованиях чат-ботов с использованием искусственного интеллекта для будущего воздействия на образование

Gelecekteki Eğitim Etkisi için Yapay Zeka Chatbot Araştırmalarında Sınırlamaların Aşılması

Khắc phục những hạn chế trong nghiên cứu Chatbot AI để tạo ra tác động giáo dục trong tương lai

克服人工智能聊天机器人研究的局限性，为未来教育带来影响

This paper is available on arxiv under CC BY-SA 4.0 DEED license.
Authors:
(1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de;
(2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany;
(6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ;
(7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction
Materials and Methods
Data
Results
Conclusions, Open Research Section and Conflict of Interest/Competing Interests
Author Contributions, Supplementary Information, Acknowledgments and References
Additional Insight from the Conceptual Model 
AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
Investigation of Resilience Loss in AMSR2’s X-Band
Results Proving Robustness
Potential Driving Forces and Sources of Bias INVESTIGATION OF RESILIENCE LOSS IN AMSR2’S X-BAND Text S7. The influence of forest loss for the years 2021 to 2022 The X-band of AMSR2 results from the observation of a shorter wavelength compared to the two C-bands. This leads to less sensitivity to aboveground biomass, which is why the results are less likely to indicate whether the ARF is loosing resilience or not. For completeness, Figures S7 and S8 show the results for the X-band. While the spatial correlation is more often increasing than not, which is comparable to the results in the C-bands, this is not the case for variance and AC1 (Figure S7). While the percentage of cells in Figure S8a) exhibiting a positive trend in all three indicators is still above what would be expected by random (12.5 %), it is far less than detected for AMSR2’s C-bands. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction Materials and Methods Data Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Potential Driving Forces and Sources of Bias Abstract and Introduction Abstract and Introduction Materials and Methods Materials and Methods Data Data Results Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Results Proving Robustness Potential Driving Forces and Sources of Bias Potential Driving Forces and Sources of Bias INVESTIGATION OF RESILIENCE LOSS IN AMSR2’S X-BAND Text S7. The influence of forest loss for the years 2021 to 2022 The X-band of AMSR2 results from the observation of a shorter wavelength compared to the two C-bands. This leads to less sensitivity to aboveground biomass, which is why the results are less likely to indicate whether the ARF is loosing resilience or not. For completeness, Figures S7 and S8 show the results for the X-band. While the spatial correlation is more often increasing than not, which is comparable to the results in the C-bands, this is not the case for variance and AC1 (Figure S7). While the percentage of cells in Figure S8 a) exhibiting a positive trend in all three indicators is still above what would be expected by random (12.5 %), it is far less than detected for AMSR2’s C-bands. a)

Loss of Amazon Rainforest Resilience: Investigation of Resilience Loss in AMSR2’s X-Band

This paper is available on arxiv under CC BY-SA 4.0 DEED license.
Authors:
(1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de;
(2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany;
(6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ;
(7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction
Materials and Methods
Data
Results
Conclusions, Open Research Section and Conflict of Interest/Competing Interests
Author Contributions, Supplementary Information, Acknowledgments and References
Additional Insight from the Conceptual Model 
AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
Investigation of Resilience Loss in AMSR2’s X-Band
Results Proving Robustness
Potential Driving Forces and Sources of Bias RESULTS PROVING ROBUSTNESS Text S8. Robustness with respect to the parameterization of the STL algorithm The results in Figures 5 and 6 are based on the residual found by the STL algorithm with default parameterization. Yet, Figure S9 and S10 prove that the analysis is not influenced by this choice as the results are almost identical when the length of the seasonal smoother equals 13 (instead of 7). Text S9. Robustness with respect to the size of the sliding windows The choice of the sliding window size is difficult for short time series as that of AMSR-E’s and AMSR2’s VOD. Yet, the results are robust w.r.t it, as Figure S11 based on sliding windows of 3 years is almost identical to Figure 5 with windows of 5 years. This also holds for the summarizing Figure S12 when compared to Figure 6. It is to note that the spatially averaged time series of the three indicators are more noisy for the choice of smaller windows, which is due to the fact the the calculation of the indicators becomes more unstable when calculated on only 36 data points, corresponding to the 3 years. The right choice of the window size in an approach to resilience analysis like the one shown here is challenging, as the increase can only be determined reliably based on enough indicator time steps either. The fact that the results are robust with respect to the window size gives confidence in the choice of 5 years as default. Text S10. Robustness with respect to the measure of change In the main text, any change in indicators is measured in terms of linear trends. Yet, another option would be the Kendall’s τ as a measure of steadiness of increase. Figure S13 is the analogue of Figure 5 but with increase in time series quantified in terms of Kendall’s τ . In Figure 6, increase in the indicators is quantified in terms of linear trends. The results are almost identical when measuring the indicators’ increase by the Kendall τ rank coefficient. Text S11. Robustness with respect to the distance defining neighboring cells The definition of the spatial correlation depends on the distance taken into account when taking the average correlation with neighboring cells. Figure S15 shows the trends of the spatial correlation for the different VIs comparing different radii defining the set of neighboring cells. As expected, a greater number of neighbors spatially smooths the results, but does not influence the general spatial pattern of the trends of spatial correlation. The turquoise circle on the bottom of the map shows an exemplary circular band of neighboring cells. In the main text, all results are based on a radius of 100 km. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction Materials and Methods Data Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Potential Driving Forces and Sources of Bias Abstract and Introduction Abstract and Introduction Materials and Methods Materials and Methods Data Data Results Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Results Proving Robustness Potential Driving Forces and Sources of Bias Potential Driving Forces and Sources of Bias RESULTS PROVING ROBUSTNESS Text S8. Robustness with respect to the parameterization of the STL algorithm The results in Figures 5 and 6 are based on the residual found by the STL algorithm with default parameterization. Yet, Figure S9 and S10 prove that the analysis is not influenced by this choice as the results are almost identical when the length of the seasonal smoother equals 13 (instead of 7). Text S9. Robustness with respect to the size of the sliding windows The choice of the sliding window size is difficult for short time series as that of AMSR-E’s and AMSR2’s VOD. Yet, the results are robust w.r.t it, as Figure S11 based on sliding windows of 3 years is almost identical to Figure 5 with windows of 5 years. This also holds for the summarizing Figure S12 when compared to Figure 6. It is to note that the spatially averaged time series of the three indicators are more noisy for the choice of smaller windows, which is due to the fact the the calculation of the indicators becomes more unstable when calculated on only 36 data points, corresponding to the 3 years. The right choice of the window size in an approach to resilience analysis like the one shown here is challenging, as the increase can only be determined reliably based on enough indicator time steps either. The fact that the results are robust with respect to the window size gives confidence in the choice of 5 years as default. Text S10. Robustness with respect to the measure of change In the main text, any change in indicators is measured in terms of linear trends. Yet, another option would be the Kendall’s τ as a measure of steadiness of increase. Figure S13 is the analogue of Figure 5 but with increase in time series quantified in terms of Kendall’s τ . In Figure 6, increase in the indicators is quantified in terms of linear trends. The results are almost identical when measuring the indicators’ increase by the Kendall τ rank coefficient. Text S11. Robustness with respect to the distance defining neighboring cells The definition of the spatial correlation depends on the distance taken into account when taking the average correlation with neighboring cells. Figure S15 shows the trends of the spatial correlation for the different VIs comparing different radii defining the set of neighboring cells. As expected, a greater number of neighbors spatially smooths the results, but does not influence the general spatial pattern of the trends of spatial correlation. The turquoise circle on the bottom of the map shows an exemplary circular band of neighboring cells. In the main text, all results are based on a radius of 100 km.

Loss of Amazon Rainforest Resilience: Results Proving Robustness

This paper is available on arxiv under CC BY-SA 4.0 DEED license.
Authors:
(1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de;
(2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany;
(6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ;
(7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction
Materials and Methods
Data
Results
Conclusions, Open Research Section and Conflict of Interest/Competing Interests
Author Contributions, Supplementary Information, Acknowledgments and References
Additional Insight from the Conceptual Model 
AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
Investigation of Resilience Loss in AMSR2’s X-Band
Results Proving Robustness
Potential Driving Forces and Sources of Bias POTENTIAL DRIVING FORCES AND SOURCES OF BIAS Text S12. The influence of the South American Monsoon system as a potentially bistable system In the Amazon basin, the South American monsoon system and the atmospheric moisture recycling form a complex and potentially bi-stable system. Hence the monthly precipitation itself could exhibit signs of CSD, which would then drive vegetation changes that are wrongly interpreted as the destabilization of the vegetation. In Figure S16, the signs of the trend of variance, AC1 and spatial are compared between VOD and precipitation (Funk et al., 2014). We argue that due to the high amount of cells where the values differ in sign for the VIs and precipitation (sum of diagonal cells in boxes expected to be 50 % for random assignment), the observed changes in the indicators are likely a measure of critical slowing down in the vegetation and not only a direct effect of variability in precipitation. Text S13. The influence of changes in precipitation as a potential driving force If precipitation is the main driver of ARF resilience loss, one would expect that in regions with stronger decreases in precipitation (negative trend), more critical slowing down can be observed. This comparison in displayed in Figure S17. In parts, decreased precipitation could potentially explain the destabilization detected in the western Amazon basin by both bands of AMSR-E as well as in the western and northeastern regions detected by AMSR2’s bands. However it is important to note here that the number of cells with an increasing amount of precipitation and decreasing indicators and those with decreasing amount of precipitation and increasing indicators add up to only about 50 %, which is what is to be expected for a random assignment of positive and negative trends in both. Text S14. The influence of changes in temperature Instead of decreasing precipitation (see Figure S17, an increase in temperature can also lead to vegetation destabilization through water stress by higher evapo-transpirative demand. If this was the driver of the CSD, increasing temperatures must coincide with positive trends in the indicators of CSD in Figure S18. Yet changes in temperature cannot explain the detected destabilization in the years 2002 to 2011. For the time period of AMSR2, destabilization detected in the northeast (band C1) and the east (band C2) might at least partially be accounted for rising temperatures (the percentage of cells with rising/declining temperatures and in-/decreasing indicators falls between 48.6 % and up to 59.3 %). Text S15. The potential bias induced by saturation In case of optical sensors, their saturation over high biomass regions often renders the VIs insensitive to the subtle changes due to perturbations that a resilience analysis based on CSD relies on. The argument is that with decreasing biomass, the VIs become more sensitive, which then translates into a signal of CSD as a direct artifact of the higher perception of perturbations. Even though VOD results from observations of microwaves, we want to address the question of saturation in VOD. As Figure S19 shows, we cannot exclude this as an artifact for AMSRE’s band C as well as AMSR2’s two bands, as they hardly observed any cells with increasing biomass. Interestingly, AMSR-E’s band X detected an increase in biomass as well as a loss of resilience in most of the cells (77.3 % to 78.2 %). Thus, its signal of destabilization of almost the entire ARF during the years 2002 to 2011 is not an artifact of saturation of the band. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction Materials and Methods Data Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Potential Driving Forces and Sources of Bias Abstract and Introduction Abstract and Introduction Materials and Methods Materials and Methods Data Data Results Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Results Proving Robustness Potential Driving Forces and Sources of Bias Potential Driving Forces and Sources of Bias POTENTIAL DRIVING FORCES AND SOURCES OF BIAS Text S12. The influence of the South American Monsoon system as a potentially bistable system In the Amazon basin, the South American monsoon system and the atmospheric moisture recycling form a complex and potentially bi-stable system. Hence the monthly precipitation itself could exhibit signs of CSD, which would then drive vegetation changes that are wrongly interpreted as the destabilization of the vegetation. In Figure S16, the signs of the trend of variance, AC1 and spatial are compared between VOD and precipitation (Funk et al., 2014). We argue that due to the high amount of cells where the values differ in sign for the VIs and precipitation (sum of diagonal cells in boxes expected to be 50 % for random assignment), the observed changes in the indicators are likely a measure of critical slowing down in the vegetation and not only a direct effect of variability in precipitation. Text S13. The influence of changes in precipitation as a potential driving force If precipitation is the main driver of ARF resilience loss, one would expect that in regions with stronger decreases in precipitation (negative trend), more critical slowing down can be observed. This comparison in displayed in Figure S17. In parts, decreased precipitation could potentially explain the destabilization detected in the western Amazon basin by both bands of AMSR-E as well as in the western and northeastern regions detected by AMSR2’s bands. However it is important to note here that the number of cells with an increasing amount of precipitation and decreasing indicators and those with decreasing amount of precipitation and increasing indicators add up to only about 50 %, which is what is to be expected for a random assignment of positive and negative trends in both. Text S14. The influence of changes in temperature Instead of decreasing precipitation (see Figure S17, an increase in temperature can also lead to vegetation destabilization through water stress by higher evapo-transpirative demand. If this was the driver of the CSD, increasing temperatures must coincide with positive trends in the indicators of CSD in Figure S18. Yet changes in temperature cannot explain the detected destabilization in the years 2002 to 2011. For the time period of AMSR2, destabilization detected in the northeast (band C1) and the east (band C2) might at least partially be accounted for rising temperatures (the percentage of cells with rising/declining temperatures and in-/decreasing indicators falls between 48.6 % and up to 59.3 %). Text S15. The potential bias induced by saturation In case of optical sensors, their saturation over high biomass regions often renders the VIs insensitive to the subtle changes due to perturbations that a resilience analysis based on CSD relies on. The argument is that with decreasing biomass, the VIs become more sensitive, which then translates into a signal of CSD as a direct artifact of the higher perception of perturbations. Even though VOD results from observations of microwaves, we want to address the question of saturation in VOD. As Figure S19 shows, we cannot exclude this as an artifact for AMSRE’s band C as well as AMSR2’s two bands, as they hardly observed any cells with increasing biomass. Interestingly, AMSR-E’s band X detected an increase in biomass as well as a loss of resilience in most of the cells (77.3 % to 78.2 %). Thus, its signal of destabilization of almost the entire ARF during the years 2002 to 2011 is not an artifact of saturation of the band.

Loss of Amazon Rainforest Resilience: Potential Driving Forces and Sources of Bias

This paper is available on arxiv under CC BY-SA 4.0 DEED license.
Authors:
(1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de;
(2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany;
(6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ;
(7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction
Materials and Methods
Data
Results
Conclusions, Open Research Section and Conflict of Interest/Competing Interests
Author Contributions, Supplementary Information, Acknowledgments and References
Additional Insight from the Conceptual Model 
AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
Investigation of Resilience Loss in AMSR2’s X-Band
Results Proving Robustness
Potential Driving Forces and Sources of Bias ADDITIONAL INSIGHT FROM THE CONCEPTUAL MODEL Text S1. Stability in the conceptual model The interaction of vegetation and precipitation is plotted in Figure S1. In the conceputal model, vegetation is a function of absolute incoming precipitation only and is not dependent on the cell’s location. Hence, it is the same for each cell and plotted as the single black solid line. Precipitation on the other hand is a function of the vegetation of the cell, but also of the vegetation of the cell’s eastern neighbors. Hence, it differs for all cells, which are encoded by the colors as in Figure 2. For the calculation of one cell’s precipitation as a function of vegetation, its eastern neighbors are considered to be in equilibrium. Furthermore, precipitation depends on the amount of moisture coming in from the Atlantic ocean, or in other words on the control parameter B. The solid lines represent the state in the beginning of the experiment with B = 1010, and the dashed lines in the end with B = 990. Where the black and coloured lines cross at some (V∗, P∗), the corresponding value of vegetation supports the value of precipitation P∗, but also is V∗ supported by P∗. Hence, this point is considered a fixed point, or equilibrium. As the moisture inflow is reduced in the simulations from 1010 (solid lines) to 990 (dashed lines), the fixed points with high vegetation values vanish, explaining the tipping in the Figures S4 and S5 . Text S2. Precipitation decrease in the conceptual model The results of simulations with decreasing moisture inflow from the Atlantic ocean (decreasing B(t)) were presented in the main text in Figure 4. Figure S2 zooms in into cell-wise total precipitation and highlights its stochasticity as well as the almost linear decrease up to the Tipping Point of vegetation. Text S3. Deforestation scenario leads to Tipping in the conceptual model The results of simulations with decreasing moisture inflow from the Atlantic ocean (decreasingB(t)) were presented in the main text in Figure 4. Figure S3 is identical but presents the results from simulations with unchanging inflow from the Atlantic, i.e. B(t) = 1000 for all t. Instead, the amount vegetation in the east-most cell is considered the bifurcation parameter and constantly decreased from 1 to 0, thereby mimicking deforestation. Text S4. Distribution of the trends for the three resilience indicators In the conceptual model, all indicators show clear trends when approaching the Tipping Point of vegetation, as illustrated in Figure S4. There, clearly all three indicators show a strong shift towards positive values with only very few negative ones for AC1. In a spatially extended system like the ARF, one would thus expect all three indicators to increase when the corresponding grid cell loses resilience. Text S5. Distribution of the trends for the three resilience indicators In Figure 4 the time of emergence (ToE) was defined as the first time when a trend becomes significant. Yet, as more conservative approach one can define the ToE as the time when a trend becomes permanently significant, i.e. when all trends later in time are significant as well. The results with this definition are displayed in Figure S5. The uncertainty in the timing of this type of ToE differs much less for AC1 compared to the definition in the main text. This indicates that AC1 might be more prone to false alarms. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction Materials and Methods Data Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Potential Driving Forces and Sources of Bias Abstract and Introduction Abstract and Introduction Materials and Methods Materials and Methods Data Data Results Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Results Proving Robustness Potential Driving Forces and Sources of Bias Potential Driving Forces and Sources of Bias ADDITIONAL INSIGHT FROM THE CONCEPTUAL MODEL Text S1. Stability in the conceptual model The interaction of vegetation and precipitation is plotted in Figure S1. In the conceputal model, vegetation is a function of absolute incoming precipitation only and is not dependent on the cell’s location. Hence, it is the same for each cell and plotted as the single black solid line. Precipitation on the other hand is a function of the vegetation of the cell, but also of the vegetation of the cell’s eastern neighbors. Hence, it differs for all cells, which are encoded by the colors as in Figure 2. For the calculation of one cell’s precipitation as a function of vegetation, its eastern neighbors are considered to be in equilibrium. Furthermore, precipitation depends on the amount of moisture coming in from the Atlantic ocean, or in other words on the control parameter B. The solid lines represent the state in the beginning of the experiment with B = 1010, and the dashed lines in the end with B = 990. Where the black and coloured lines cross at some (V∗, P∗), the corresponding value of vegetation supports the value of precipitation P∗, but also is V∗ supported by P∗. Hence, this point is considered a fixed point, or equilibrium. As the moisture inflow is reduced in the simulations from 1010 (solid lines) to 990 (dashed lines), the fixed points with high vegetation values vanish, explaining the tipping in the Figures S4 and S5 . Text S2. Precipitation decrease in the conceptual model The results of simulations with decreasing moisture inflow from the Atlantic ocean (decreasing B(t)) were presented in the main text in Figure 4. Figure S2 zooms in into cell-wise total precipitation and highlights its stochasticity as well as the almost linear decrease up to the Tipping Point of vegetation. Text S3. Deforestation scenario leads to Tipping in the conceptual model The results of simulations with decreasing moisture inflow from the Atlantic ocean (decreasingB(t)) were presented in the main text in Figure 4. Figure S3 is identical but presents the results from simulations with unchanging inflow from the Atlantic, i.e. B(t) = 1000 for all t. Instead, the amount vegetation in the east-most cell is considered the bifurcation parameter and constantly decreased from 1 to 0, thereby mimicking deforestation. Text S4. Distribution of the trends for the three resilience indicators In the conceptual model, all indicators show clear trends when approaching the Tipping Point of vegetation, as illustrated in Figure S4. There, clearly all three indicators show a strong shift towards positive values with only very few negative ones for AC1. In a spatially extended system like the ARF, one would thus expect all three indicators to increase when the corresponding grid cell loses resilience. Text S5. Distribution of the trends for the three resilience indicators In Figure 4 the time of emergence (ToE) was defined as the first time when a trend becomes significant. Yet, as more conservative approach one can define the ToE as the time when a trend becomes permanently significant, i.e. when all trends later in time are significant as well. The results with this definition are displayed in Figure S5. The uncertainty in the timing of this type of ToE differs much less for AC1 compared to the definition in the main text. This indicates that AC1 might be more prone to false alarms.

Loss of Amazon Rainforest Resilience: Additional Insight from the Conceptual Model

Loss of Amazon Rainforest Resilience: Author Contributions

This paper is available on arxiv under CC BY-SA 4.0 DEED license.
Authors:
(1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de;
(2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany;
(5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany;
(6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ;
(7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction
Materials and Methods
Data
Results
Conclusions, Open Research Section and Conflict of Interest/Competing Interests
Author Contributions, Supplementary Information, Acknowledgments and References
Additional Insight from the Conceptual Model 
AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
Investigation of Resilience Loss in AMSR2’s X-Band
Results Proving Robustness
Potential Driving Forces and Sources of Bias AMSR2 UNTIL 2022 EXCLUDED DUE TO MISSING HUMAN LAND USE DATA Text S6. The influence of forest loss for the years 2021 to 2022 Since the data sets for excluding HLU are only available until 2020, the VOD data by AMSR2 was only analyzed until 2020 even though it is available until present times. The maps in Figure S6a) and c) show the difference between the number of indicators with positive trends if analyzed until 2020 (as in the main text, Figure 6) and until 2022 for AMSR2’s band C and X, respectively. Differences mainly occur where sudden changes in VOD after 2020 can have a strong influence on the detrending method STL. Since STL acts on sliding windows over the whole data set as is (to detrend) and over the separate months (for de-seasonalizing), the influence of strong changes after 2020 increase over time. This becomes apparent in the the spatially averaged time series in b) and d), where all indicators start at the same level independent for both cases, where the analysis is performed until 2020 (solid lines) as well as until 2022 (dotted lines), but diverge afterwards. Such increasingly strong discrepancies in the period until 2020, where the data is equivalent, must stem from STL, which over time becomes more sensitive to strong changes in later time steps due to its sliding window approach. Yet, this is only the case for strong changes in the VOD after 2020, as it could result from forest loss and Human interference. Hence, we argue that the exclusion of HLU based on the years until 2020 is not sufficient to analyze data thereafter. This explanation is supported by high rates of forest loss since the end of 2020, with more than 18.000 km2 deforestation and 27.000 km2 degradation registered by the Instituto Nacional de Pesquisas Espaciais’s monitoring programs (INPE, 2023). Hence, the results for AMSR2 analyzed until 2022 cannot be relied on (see Figure S6). This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. This paper is available on arxiv under CC BY-SA 4.0 DEED license. Authors: Authors: (1) Lana L. Blaschke, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany and lana.blaschke@pik-potsdam.de; (2) Da Nian, Potsdam Institute for Climate Impact Research, Potsdam, Germany; (3) Sebastian Bathiany,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (4)  Maya Ben-Yami,  Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, and Potsdam Institute for Climate Impact Research, Potsdam, Germany; (5) Taylor Smith, Institute of Geosciences, University of Potsdam, 14476 Potsdam, Germany; (6)  Chris A. Boulton, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK ; (7) Niklas Boers, Earth System Modelling, School of Engineering and Design, Technical University of Munich, Munich, Germany, Potsdam Institute for Climate Impact Research, Potsdam, Germany, Global Systems Institute, University of Exeter, EX4 4QE Exeter, UK, and Department of Mathematics, University of Exeter, EX4 4QF Exeter, UK. Table of Links Abstract and Introduction Materials and Methods Data Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Potential Driving Forces and Sources of Bias Abstract and Introduction Abstract and Introduction Materials and Methods Materials and Methods Data Data Results Results Conclusions, Open Research Section and Conflict of Interest/Competing Interests Conclusions, Open Research Section and Conflict of Interest/Competing Interests Author Contributions, Supplementary Information, Acknowledgments and References Author Contributions, Supplementary Information, Acknowledgments and References Additional Insight from the Conceptual Model Additional Insight from the Conceptual Model AMSR2 until 2022 Excluded Due to Missing Human Land Use Data AMSR2 until 2022 Excluded Due to Missing Human Land Use Data Investigation of Resilience Loss in AMSR2’s X-Band Investigation of Resilience Loss in AMSR2’s X-Band Results Proving Robustness Results Proving Robustness Potential Driving Forces and Sources of Bias Potential Driving Forces and Sources of Bias AMSR2 UNTIL 2022 EXCLUDED DUE TO MISSING HUMAN LAND USE DATA Text S6. The influence of forest loss for the years 2021 to 2022 Since the data sets for excluding HLU are only available until 2020, the VOD data by AMSR2 was only analyzed until 2020 even though it is available until present times. The maps in Figure S6 a) and c) show the difference between the number of indicators with positive trends if analyzed until 2020 (as in the main text, Figure 6) and until 2022 for AMSR2’s band C and X, respectively. Differences mainly occur where sudden changes in VOD after 2020 can have a strong influence on the detrending method STL. Since STL acts on sliding windows over the whole data set as is (to detrend) and over the separate months (for de-seasonalizing), the influence of strong changes after 2020 increase over time. This becomes apparent in the the spatially averaged time series in b) and d) , where all indicators start at the same level independent for both cases, where the analysis is performed until 2020 (solid lines) as well as until 2022 (dotted lines), but diverge afterwards. Such increasingly strong discrepancies in the period until 2020, where the data is equivalent, must stem from STL, which over time becomes more sensitive to strong changes in later time steps due to its sliding window approach. Yet, this is only the case for strong changes in the VOD after 2020, as it could result from forest loss and Human interference. Hence, we argue that the exclusion of HLU based on the years until 2020 is not sufficient to analyze data thereafter. This explanation is supported by high rates of forest loss since the end of 2020, with more than 18.000 km2 deforestation and 27.000 km2 degradation registered by the Instituto Nacional de Pesquisas Espaciais’s monitoring programs (INPE, 2023). Hence, the results for AMSR2 analyzed until 2022 cannot be relied on (see Figure S6). a) c) b) d)

Loss of Amazon Rainforest Resilience: AMSR2 until 2022 Excluded Due to Missing Human Land Use Data

Loss of Amazon Rainforest Resilience: Conclusions

Loss of Amazon Rainforest Resilience: Abstract and Introduction

Loss of Amazon Rainforest Resilience: Results

Loss of Amazon Rainforest Resilience: Materials and Methods

Virtual reality allows us to rebuild a more human society from the ground up - from financial systems to governance and the very concept of “state". "System" says the metaverse is the wrong way for humanity. It resists its development and tries to stop its popularity. In my opinion, rebuilding the existing world and its "system" is impossible. But it is possible in virtual reality. Consider this, we can create a world without bureaucracy, institutions, and an authoritarian legacy. With this article, I opt out of a current "system" built in our real world and encourage people not to fear metaverses as they are our future. I believe that virtual worlds should never be considered as a way to escape from reality - they are just as real and can be lived in. "The Matrix" and science fiction classics: can we consider these works suitable metaverse examples? Every writer is a kind of philosopher because every science fiction work is a mental experiment. The author makes some assumptions like the existence of intelligent machines or completely virtual reality, and on this basis, models the possible development of events. Philosophers do approximately the same thing. science fiction Stanisław Herman Lem is outstanding in this respect. Many of his ideas of the middle of the last century then passed into philosophy. Lana Wachowski and Lilly Wachowski in "The Matrix" also worked out some ideas and problems quite well, but I disagree with them on one of the main points. The main characters of "The Matrix" talk about the feeling of unreality of the world created by machines. They say this feeling appears because of some trifles and starts to drive them crazy. I am sure a human cannot distinguish a quality virtual reality from the real world. And in principle, there is no need for that. So there's no way we can tell we're living in a virtual world instead of the real one? Perhaps we can if the simulation turns out to be imperfect. For example, objects sometimes fall through surfaces. Or if the creators start communicating with us and showing us all sorts of unrealistic miracles, such as turning the Statue of Liberty upside down or moving the waters of the sea. However, I doubt even that would convince all people of the world's unreality. The hardest case is a perfect simulation completely indistinguishable from reality, in which the creators do not interfere. Then, it is not clear how we could test its artificiality. Any experiments would proceed exactly as they would in reality. It is difficult even to imagine what could be the proof of virtuality of such a world. Science is powerless here, leaving pure philosophy with no way to test hypotheses. We know that simulations can be created with computers. In 50 or 100 years, we will surely be able to create a virtual reality that is completely indistinguishable from the real world. The question then arises: how can we be sure that we ourselves are living in the real world and not in one of the simulations? Especially if we ourselves do not notice any difference. If there is no difference, it probably doesn't matter, does it? Let's imagine that tomorrow, we somehow reliably learn that our world is a computer simulation like the Matrix. I think it would shock people at first. Some people will think their whole life has been an illusion and a lie. They will feel terrible. But I think, little by little, people will get used to it and move on because the feeling of the reality of the world is not connected with what the world is made of. Subjective sensations will not become less real to us, nor will our experiences become less valuable if we learn that we live in a simulation. The Earth will continue to revolve around the Sun, friends will remain friends, and we will still need to do something to live. It will turn out that trees, buildings, and our own bodies are made up of bits rather than atoms - but what will that change for a particular person? In a sense, we already know that we live in a simulation. The world people think is real is merely the result of our bodies. We do not perceive physical reality directly; we only receive a model created by our eyes, ears, skin, and brain. Directly, we can perceive only our subjective experience: sensations, emotions, thoughts, and so on. And any external world is infinitely distant from us. That is why I tend to draw the line not between the real and the virtual but between the corporeal and the virtual. All worlds outside the human body (or, rather, outside the human mind) are virtual to a certain extent, including "real" physical reality. But no one seems to be confused by the fact that our human vision of reality has little to do with quantum mechanics, which shows the true structure of reality. We are not really interested in what the world around us is made of. The main thing is that it works according to certain rules. That is, it's all about a person's willingness to recognize the world as real and to behave as if everything around them has some significance. None of us think, "Everything around us is made of strange, incomprehensible atoms, so it's all an illusion". We're just living. Then why should we think this way about zeros and ones? Then what does it mean to be real? What is a person ready to recognize as real? First of all, the world must be objective; it must exist beyond our mind and beyond our will so that objects are stable and the spoon does not bend even when we think it is not there. Cause and effect relationships must work in this world so that people's actions lead to some changes, and those changes lead to the next change. This has to do with the way our brains work. It has long been thought that it receives undistorted information from the senses. In fact, the brain filters it to suit its expectations, which are influenced by previous experiences. Everything that enters the brain mixes with episodes of its memory - as if anticipating a picture of what is happening. This mechanism works already at the level of vision. Apparently, it is impossible to get rid of this peculiarity of the brain in principle because it was evolutionarily developed as a "prediction machine", which very much dislikes being wrong in expectations. Therefore, objectivity, causality, and predictability should be at the core of any virtual reality that wants to be human-friendly. Also, the world should not restrict people from interacting with each other. It is the contacts between people that weave together a shared reality, so a living virtual world is hardly possible without people's freedom to do both wonderful and terrible things together. Aren't the technical aspects important? The resolution and color of the picture, the presence of touch... It's all important, but in my opinion, it's secondary. Simply because we already have examples of virtual realities that are technically bad, but people are still willing to spend a lot of time in them to live their whole lives. Look at Second Life. It has outdated graphics with no real three-dimensionality. Users still sit in front of their monitors and see a flat picture. But that doesn't really bother them. Second Life, for almost 20 years of history, has become a prototype of the meta-universe - concerts and interviews with famous people take place in this virtual world. On the other hand, there are worlds with full-fledged VR, like VRChat and Rec Room, and none of them have been as successful as Second Life. So, people don't care so much about technical specs. The generation of 50-year-olds tends to think of games and other virtual worlds as something second-rate to mainstream reality. I don't think it has to do with their technical scarcity. Rather, they consider games meaningless, empty pastimes that do not impact real life. But this perception is outdated. Digital worlds have their own economies, people use them to solve social and political problems, and virtual realities have become more than just self-sufficient. They are already influencing the main, "real" reality. Virtual worlds should not be seen as a means of escapism, an escape from reality. Virtual worlds extend the physical world, and building a meaningful life in them is also possible. However, the technical aspects are important in terms of mass appeal: the worse a virtual world covers the human senses, the more potential users it scares away. Few people want to play computer games without sound, but a whole virtual world without sounds would probably seem like a completely unacceptable option. When will virtual reality become indistinguishable from the physical world? Right now, virtual worlds in the form of games and meta-universe are facing "childish" problems like the inconvenience of VR headsets. Of course, living a fulfilling life in a virtual reality with only pictures, sounds, dizziness, and neck pain is hard. I think it will take us 20 or 30 years just to start covering vision and hearing perfectly. Maybe partially cover touch, too. Virtual reality will become truly mainstream when it can simulate the smells, tastes, and bodily sensations of eating, drinking, and sex. It's unlikely to do this without a full-fledged neural interface that connects directly to the central nerves, as in "The Matrix". And it seems to me that such an interface will appear only by the end of the century. With the development of technical means, even the external difference between real and virtual worlds will become increasingly blurred until simulation becomes completely indistinguishable from physical reality. The virtual world does not have to copy the real world completely. The simulation can be completely fantastic, even fairy-tale-like - the main thing is that it remains fundamentally understandable and familiar to humans.

Self-Sovereignty in Virtual Reality: Humans Won’t Know They’re in the Matrix, and They Don’t Need To

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Dihia LANASRI, OMDENA, New York, USA; (2) Juan OLANO, OMDENA, New York, USA; (3) Sifal KLIOUI, OMDENA, New York, USA; (4) Sin Liang Lee, OMDENA, New York, USA; (5) Lamia SEKKAI, OMDENA, New York, USA. Table of Links Abstract & Introduction Background Related Work Our Methodology Experiments and Results Conclusion, Acknowledgments & References Abstract With the proliferation of hate speech on social networks under different formats, such as abusive language, cyberbullying, and violence, etc., people have experienced a significant increase in violence, putting them in uncomfortable situations and threats. Plenty of efforts have been dedicated in the last few years to overcome this phenomenon to detect hate speech in different structured languages like English, French, Arabic, and others. However, a reduced number of works deal with Arabic dialects like Tunisian, Egyptian, and Gulf, mainly the Algerian ones. To fill in the gap, we propose in this work a complete approach for detecting hate speech on online Algerian messages. Many deep learning architectures have been evaluated on the corpus we created from some Algerian social networks (Facebook, YouTube, and Twitter). This corpus contains more than 13.5K documents in Algerian dialect written in Arabic, labeled as hateful or non-hateful. Promising results are obtained, which show the efficiency of our approach. Hate Speech · Algerian dialect · Deep Learning · DziriBERT · FastText Keywords 1 Introduction Hate speech detection, or detection of offensive messages in social networks, communication forums, and websites, is an exciting and hot research topic. Many hate crimes and attacks in our current life started from social network posts and comments MacAvaney et al. [2019]. Studying this phenomenon is imperative for online communities to keep a safe environment for their users. It also has a significant benefit for security authorities and states to ensure the safety of citizens and prevent crimes and attacks. A universally accepted definition of hate speech is currently unavailable Bogdani et al. [2021] because of the variation of cultures, societies, and local languages. Other difficulties include the diversity of national laws, the variety of online communities, and forms of online hate speech. Various definitions are proposed. According to the Encyclopedia of the American Constitution: "Hate speech is speech that attacks a person or group based on attributes such as race, religion, ethnic origin, national origin, sex, disability, sexual orientation, or gender identity." Nockleby [2000]. Today, many authors largely used this definition Guellil et al. [2022]. Facebook considers hate speech as "a direct attack on people based on protected characteristics—race, ethnicity, national origin, religious affiliation, sexual orientation, caste, sex, gender, gender identity, and serious disease or disability. We also provide some protections for immigration status." [1] . Davidson et al., who defines hate speech as "language that is used to express hatred towards a targeted group or is intended to be derogatory, to humiliate, or to insult the members of the group" propose one of the most accepted definitions Davidson et al. [2017]. Alternatively, the one proposed by Fortuna et al., "Hate speech is a language that attacks or diminishes, that incites violence or hate against groups, based on specific characteristics such as physical appearance, religion, descent, national or ethnic origin, sexual orientation, gender identity or other, and it can occur with different linguistic styles, even in subtle forms or when humor is used." Fortuna and Nunes [2018]. The literature review shows that the term Hate speech (which is the most commonly used) has various synonym terms such as abusive speech, offensive language, cyberbullying, or sexism detection Schmidt and Wiegand [2017]. Many works have been published in the context of hate speech detection for different standard and structured languages, like French Battistelli et al. [2020], English Alkomah and Ma [2022], Spanish Plaza-del Arco et al. [2021], and Arabic Albadi et al. [2018]. These languages are known for their standardization with well-known grammar and structure, which make the language processing well mastered. However, detecting hate speech in dialects, mainly Arabic ones such as Libyan, Egyptian, and Iraqi, etc. is still challenging and complex work Mulki et al. [2019]. Even if they are derived from the literal Arabic language, each country’s specific vocabulary and semantics are added or defined. In this work, we are interested in detecting hate speech in the Algerian dialect. This latter is one of the complex dialects Mezzoudj et al. [2019] characterized by the variety of its sub-dialects according to each region within the country. Algeria is a country with 58 regions; each one has a specificity in its spoken language with different words and meanings. The same word may have various meanings for each region; for example, ’Flouka’ in the east means ’earrings.’ In the north, it means ’small boat’. Moreover, new ’odd’ words are continually added to the Algerian vocabulary. The Algerian dialect is known for its morphological and orthographic richness. Facing this situation, treating and understanding the Algerian dialect for hate speech detection is a complex work. The importance of this project for the Algerian context encourages us to work on this problem. To the best of our knowledge, only few works have been proposed for hate speech detection in the Algerian dialect Boucherit and Abainia [2022], Menifi et al. [2022]. Some other related topics are treated like sentiment analysis Abdelli et al. [2019], sexism detection Guellil et al. [2021] which may be exploited to analyze the hate speech. In this paper, we proposed a complete end-to-end natural language processing (NLP) approach for hate speech detection in the Algerian dialect. Our approach covers the main steps of an NLP project, including data collection, data annotation, feature extraction, and then model development based on machine and deep learning, model evaluation, and inference. In this paper, we proposed a complete end-to-end natural language processing (NLP) approach for hate speech detection in the Algerian dialect. Our approach covers the main steps of an NLP project, including data collection, data annotation, feature extraction, and then model development based on machine and deep learning, model evaluation, and inference. This paper is structured as follows: Section 2 presents a necessary background, Section 3 reviews the most important related works, Section 4 details our proposed approach and evaluated models, Section 5 discusses the obtained results, and Section 6 concludes the paper. [1] Community Standards; Available on:https://www.facebook.com/communitystandards/objectionable_content

Hate Speech Detection in Algerian Dialect Using Deep Learning: Abstract & Intro

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Dihia LANASRI, OMDENA, New York, USA; (2) Juan OLANO, OMDENA, New York, USA; (3) Sifal KLIOUI, OMDENA, New York, USA; (4) Sin Liang Lee, OMDENA, New York, USA; (5) Lamia SEKKAI, OMDENA, New York, USA. Table of Links Abstract & Introduction Background Related Work Our Methodology Experiments and Results Conclusion, Acknowledgments & References 2 Background Hate speech is commonly defined as any communication that disparages a target group of people based on some characteristic such as race, color, ethnicity, gender, sexual orientation, nationality, religion, or other characteristic De Gibert et al. [2018]. 2.1 Hate speech According to Al-Hassan and Al-Dossari [2019] hate speech is categorized into five categories: (1) gendered hate speech, including any form of misogyny and sexism; (2) religious hate speech including any religious discrimination, such as Islamic sects, anti-Christian, etc.; (3) racist hate speech including any racial offense or tribalism, and xenophobia; (4) disability including any sort of offense to an individual suffering from health problems; and (5) political hate speech can refer to any abuse and offense against politicians Guellil et al. [2022]. 2.2 Algerian Dialect and Arabic Languages Arabic is the official language of 25 countries[2] . More than 400 million people around the world speak this language. Arabic is also recognized as the 4th most-used language on the Internet Boudad et al. [2018]. Arabic is classified into three categories Habash [2022]: (1) Classical Arabic (CA), which is the form of the Arabic language used in literary texts. The Quran is considered the highest form of CA text Sharaf and Atwell [2012]. (2) Modern Standard Arabic (MSA) is used for writing and formal conversations. (3) Dialectal Arabic is used in daily life communication, informal exchanges, etc. Boudad et al. [2018] like the Algerian dialect, Tunisian dialect, etc. [2] https://worldpopulationreview.com/country-rankings/arabic-speaking-countries

Hate Speech Detection in Algerian Dialect Using Deep Learning: Background

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Dihia LANASRI, OMDENA, New York, USA; (2) Juan OLANO, OMDENA, New York, USA; (3) Sifal KLIOUI, OMDENA, New York, USA; (4) Sin Liang Lee, OMDENA, New York, USA; (5) Lamia SEKKAI, OMDENA, New York, USA. Table of Links Abstract & Introduction Background Related Work Our Methodology Experiments and Results Conclusion, Acknowledgments & References 3 Related Work A restricted number of works have been published in the context of hate speech detection dealing with Arabic dialects. This section will analyze the most important NLP approaches dedicated to (1) the Algerian dialect and (2) Other Arabic dialects like Iraqi, Egyptian, Syrian, and Tunisian. This analysis helps us to identify the used approaches, models, and corpora. 3.1 Hate speech detection in Algerian dialect Guellil et al. [2022, 2021] Developed the first approach and corpus in Algerian dialect for hate speech detection against women in Arabic community on social media. This corpus contains more than 373K YouTube comments. Two different algorithms for feature extraction were used: Word2vec with machine learning models (GaussianNB, LogisticRegression, RandomForset, SGDClassifier, and LinearSVC) and FastText with Deep learning models (deep Convolutional Neural Network (CNN), long short-term memory (LSTM) network and Bi-directional LSTM (BiLSTM) network). Simulation results demonstrated the best performance of the CNN model with FastText. Boucherit and Abainia [2022] addressed the problem of detecting offensive and abusive content in Facebook comments. The corpus contains 8.7K comments in Algerian dialect written in Arabic and Latin characters, manually annotated as usual, abusive, and offensive. They used BiLSTM, CNN, FastText, SVM, and Multinomial Naive Bayes (NB) as classifiers. The experimental results showed that SVM and Multinomial NB classifiers outperformed all the other classifiers. Abainia et al. [2022] addressed the offensive language detection in the Amazigh language, which is one of the underresourced languages. They were interested in the Kabyle dialect. A new corpus of offensive Amazigh language is proposed containing 6.2K documents collected from Facebook and manually annotated as usual or offensive. A new lexicon of offensive and abusive Amazigh words with 12.6k entries is also developed. Many models have been evaluated, like SVM and Multinomial Naive Bayes classifiers tested with tf-idf. FastText was tested with deep learning models CNN and BiLSTM. The naive statistical classifier based on lexicon checking was the winner classifier. Mazari and Kheddar [2023] introduced a new dataset for Algerian dialect toxic text detection. An annotated multi-label dataset is built, containing around 14K comments extracted from Facebook, YouTube, and Twitter and labeled as hate speech, offensive language, and cyberbullying. Several tests have been conducted using many classification models of traditional machine learning: Random Forest, Naıve Bayes, Linear Support Vector (LSV), Stochastic Gradient Descent (SGD), and Logistic Regression. Furthermore, several assessments have been conducted using Deep Learning models such as CNN, LSTM, Gated Recurrent Unit (GRU), BiLSTM and Bidirectional-GRU (Bi-GRU). Results demonstrated the best performance of LSV, BiLSTM, and MLP when associated with the SGD model. Guellil et al. [2020] proposed a system for detecting hateful speech in Arabic political debates. The approach was evaluated against a hateful corpus concerning Algerian political debates. It contains 5K YouTube comments in MSA and Algerian dialects, written in both Arabic and Latin characters. Both classical algorithms of classification (Gaussian NB, Logistic Regression, Random Forest, SGD Classifier, and Linear SVC(LSVC)) and deep learning algorithms (CNN, multilayer perceptron (MLP), LSTM, and BiLSTM) are tested. For extracting features, the authors use Word2vec and FastText with their two implementations, namely, Skip Gram and CBOW. Simulation results demonstrate the best performance of LSVC, BiLSTM and MLP. Mohdeb et al. [2022] proposed an approach for analysis and the detection of dialectal Arabic hate speech that targeted African refugees and illegal migrants on the YouTube Algerian space. The corpus contains more than 4K comments annotated as Incitement, Hate, Refusing with non-hateful words, Sympathetic, and Comment. The transfer learning approach has been exploited for classification. The experiments show that the AraBERT monolingual transformer outperforms the mono-dialectal transformer DziriBERT and the cross-lingual transformers mBERT and XLM-R. 3.2 Hate speech detection in other Arabic dialects Various datasets or corpora were published in different dialects, which can be used for different purposes like hate speech, racism, violence, etc. detection. ALBayari and Abdallah [2022] is the first work to propose a corpus built from Instagram comments. This corpus contains 198K comments, written in MSA and three different dialects: Egyptian, Gulf, and Levantine. The comments were annotated as neutral, toxic, and Bullying. Al-Ajlan and Ykhlef [2018] and Haidar et al. [2019] datasets are collected from Twitter containing respectively 20K and 34K multi-dialectal Arabic tweets annotated as bullying and non-bullying labels. These tweets were from various dialects (Lebanon, Egypt, and the Gulf area). Moreover, two other datasets were proposed by Mubarak et al. [2017]. The first one with 1.1K tweets in different dialects and the second dataset contains 32K inappropriate comments collected from a famous Arabic news site and annotated as obscene, offensive, or clean. Albadi et al. [2018] proposed the religious hate speech detection where a multi-dialectal dataset of 6.6K tweets was introduced. It included an identification of the religious groups targeted by hate speech. Alakrot et al. [2018] also provided a dataset of 16K Egyptian, Iraqi, and Libyan comments collected from YouTube. The comments were annotated as either offensive, inoffensive, or neutral. T-HSAB Haddad et al. [2019] and L-HSAB Mulki et al. [2019] are two publicly available corpora for abusive hate speech detection. The first one is in the Tunisian dialect, combining 6K comments. The second one is in Levantine dialect (Syrian, Lebanese, Palestinian, and Jordanian dialects) containing around 6K tweets. These documents are labeled as Abusive, Hate, or Normal. Mubarak et al. [2020] looked at MSA and four major dialects (Egyptian, Levantine, Maghrebi, and Gulf). It presented a systematic method for building an Arabic offensive language tweet dataset that does not favor specific dialects, topics, or genres with 10K tweets. For tweet labeling, they used the count of positive and negative terms based on a polarity lexicon. FastText and Skip-Gram (AraVec skip-gram, Mazajak skip-gram); and deep contextual embeddings, namely BERTbase-multilingual and AraBERT are used. They evaluated different models: SVM, AdaBoost, and Logistic regression. Mulki and Ghanem [2021] introduced the first Arabic Levantine Twitter dataset for Misogynistic language (LeT-Mi) to be a benchmark dataset for automatic detection of online misogyny written in the Arabic and Levantine dialect. The proposed dataset consists of 6.5K tweets annotated either as neutral (misogynistic-free) or as one of seven misogyny categories: discredit, dominance, cursing/damning, sexual harassment, stereotyping and objectification, derailing, and threat of violence. They used BOW + TF-IDF, SOTA, LSTM, BERT, and Majority class as classifiers. Duwairi et al. [2021] investigated the ability of CNN, CNN-LSTM, and BiLSTM-CNN deep learning networks to classify or discover hateful content posted on social media. These deep networks were trained and tested using the ArHS dataset, which consists of around 10K tweets that were annotated to suit hateful speech detection in Arabic. Three types of experiments are reported: first, the binary classification of tweets into Hate or Normal. Ternary classification of tweets into (Hate, Abusive, or Normal), and multi-class classification of tweets into (Misogyny, Racism, Religious Discrimination, Abusive, and Normal). Aldjanabi et al. [2021] have built an offensive and hate speech detection system using a multi-task learning (MTL) model built on top of a pre-trained Arabic language model. The Arabic MTL model was experimented with two different language models to cover MSA and dialect Arabic. They evaluated a new pre-trained model ’MarBERT’ to classify both dialect and MSA tweets. They propose a model to explore multi-corpus-based learning using Arabic LMs and MTL to improve the classification performance. Haidar et al. [2017] presented a solution for the issue of cyberbullying in both Arabic and English languages. The proposed solution is based on machine learning algorithms using a dataset from Lebanon, Syria, the Gulf Area, and Egypt. That dataset contained 35K Arabic texts. In this research, Naïve Bayes and SVM models were chosen to classify the text. The SVM model achieved greater precision. Abdelali et al. [2016] The authors built a large dataset that consists of offensive Arabic words from different dialects and topics. The tweets were labeled into one of these categories: offensive, vulgar, hate speech, or clean. Since the offensive tweets involve implicit insults, the hate speech category was the tweets that contain racism, religious, and ethnic words. Different classifiers were employed in this study; the SVM model with a radial function kernel was mainly used with lexical features and pre-trained static embedding, while Adaptive Boosting and Logistic regression classifiers were employed when using Mazajak embedding. SVM gave the best precision. According to this literature analysis, we detect that the topic of hate speech detection in the Algerian dialect is not widely considered, and only few works deal with this problem. Furthermore, a lack of Algerian datasets prepared for hate speech is found. All these findings motivate our proposal.

Hate Speech Detection in Algerian Dialect Using Deep Learning: Related Work

Hate Speech Detection in Algerian Dialect Using Deep Learning: Our Methodology

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Dihia LANASRI, OMDENA, New York, USA; (2) Juan OLANO, OMDENA, New York, USA; (3) Sifal KLIOUI, OMDENA, New York, USA; (4) Sin Liang Lee, OMDENA, New York, USA; (5) Lamia SEKKAI, OMDENA, New York, USA. Table of Links Abstract & Introduction Background Related Work Our Methodology Experiments and Results Conclusion, Acknowledgments & References 5 Experiments and Results To train and evaluate our models, we used TensorFlow or Pytorch deep learning frameworks. We used Google Colab and Kaggle GPUs to accelerate the experiments. In table ??, we will provide the detailed results that we obtained. The LinearSVC model offered a competitive accuracy but struggled with the recall for the hate speech class. The precision and recall trade-off indicates possible challenges in differentiating between the subtle nuances of hate and non-hate speech in the dialect. The model exhibited high precision and recall for class 0 but showed room for improvement for class 1, particularly in terms of recall. This suggests that while the model is quite good at identifying class 0, it could be improved for identifying class 1. Linear Support Vector Classifier (LinearSVC): One of the worst models in terms of capabilities, although it is diverging from the baseline it is unclear whether these results will hold in out of distribution cases, especially when we know that there is no underlying process in the model that captures semantic representations of the documents. gzip + KNN: the model exhibits a noteworthy precision score, signifying its accuracy in correctly classifying instances as hate speech or not. However, the relatively lower recall score suggests that it missed identifying some hate speech instances. This discrepancy might be attributed to the model’s lack of specialized handling for the nuances of the Algerian dialect, potentially causing it to overlook certain hate speech patterns unique to that context. Dziribert-FT-HEAD: Despite this, the model’s overall accuracy remains commendably high, indicating its robust performance in making accurate predictions. Additionally, the balanced precision and recall values underline its ability to strike a reasonable trade-off between minimizing false positives and false negatives, a crucial aspect in hate speech detection. The F1 Score, being the harmonic mean of precision and recall, further validates the model’s capacity to effectively identify positive samples while avoiding misclassification of negative ones. The model consistently demonstrates strong performance across multiple evaluation metrics, especially in terms of accuracy and F1 score. These results reaffirm the practicality and effectiveness of employing deep learning techniques for the challenging task of hate speech detection. Unfortunately, the results of this model are among the worst ones. The literature shows the strength of LSTM and BiLSTM in this kind of NLP project, but this is not the case for this project. The low precision is due to the incapability of the model to classify correctly the hate class. FastText is a good word embedding model that captures the context and semantics of a document. However, in this case, it does not perform well because of the fine-tuning done where we took an Arabic FastText and fine-tune it on Algerian dataset written in Arabic characters. LSTM and BiLSTM with FastText-DZ: We utilize both PEFT and LoRA to fine-tune DZiriBERT, a model specifically adapted to the Algerian dialect. By employing these techniques, we were able to create a highly effective and efficient model for hate speech detection in the Algerian dialect while keeping computational costs at a minimum. DZiriBert with Peft+LoRA: The outcomes obtained from these models are noteworthy; nonetheless, their performances pale when compared with the parameter-efficient fine-tuning on the DZiriBERT model. Multilingual-E5-base Fine Tuned and sbert-distill-multilingual Fine Tuned: The results returned by this model are satisfying considering the relatively low quantity of data it was finetuned on, this exhibits further that the pretraining plays the major role on downstream takes such as classification in our case, especially that the base model is an encoder only architecture which captures contextual information from the input data, making it useful for a wide range of text classification tasks. DzaraShield: The results are slightly inferior to Dzarashield. One possible explanation is the increased complexity of the architecture when compared to the Dzarabert base model. Consequently, fine-tuning becomes a more intricate task due to the larger hyperparameter search space and the limited resources in terms of computing power and data availability. As a result, it is reasonable to expect that these models would perform similarly in real-world scenarios. AraT5v2-HateDetect: 5.1 Results Discussion The DzaraShield model has demonstrated remarkable capability in detecting hate speech in the Algerian dialect. Its outstanding precision score highlights its reliability in accurately identifying instances of hate speech. Additionally, it maintains a balanced precision and recall, indicating that it does not excessively sacrifice precision to achieve its higher recall. Such a balanced model holds considerable advantages, particularly when both false positives and false negatives carry significant consequences. For the other models, mainly LSTM or BiLSTM with Dziri FastText, more fine-tuning should be performed to enhance the results. Moreover, future work may include hyperparameter tuning, class balancing techniques, or the integration of more complex models to improve performance across both classes. The disparity between precision and recall in certain models warrants further investigation. Delving deeper into this issue could yield valuable insights into specific aspects of the dialect that might be contributing to this imbalance. Future experiments should prioritize understanding and addressing these discrepancies, with the goal of enhancing recall without compromising precision. The results from various experimental models underscore the intricacies involved in hate speech detection in the Algerian dialect. While traditional machine learning and deep learning approaches provided some valuable insights, they fell short in capturing the dialect’s nuanced characteristics. In contrast, the DzaraShield model emerged as the most successful approach, emphasizing the pivotal role of Encoder-only models in the realm of projects of this nature. These findings offer valuable insights for future work in this area and underscore the potential of leveraging domainspecific knowledge, advanced fine-tuning techniques, and sophisticated architectures for the effective detection of hate speech in under-studied and complex dialects such as Algerian.

Hate Speech Detection in Algerian Dialect Using Deep Learning: Experiments and Results

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Dihia LANASRI, OMDENA, New York, USA; (2) Juan OLANO, OMDENA, New York, USA; (3) Sifal KLIOUI, OMDENA, New York, USA; (4) Sin Liang Lee, OMDENA, New York, USA; (5) Lamia SEKKAI, OMDENA, New York, USA. Table of Links Abstract & Introduction Background Related Work Our Methodology Experiments and Results Conclusion, Acknowledgments & References 6 Conclusion The importance of hate speech detection on social networks has encouraged many researchers to build solutions (corpora and classifiers) to detect suspect messages. The literature review shows that most works are interested in text in structured languages like English, French, Arabic, etc. However, few works deal with dialects, mainly the Algerian one, which is known for its complexity and variety. To fill in the gap, we propose in this paper a complete NLP approach to detect hate speech in the Algerian dialect. We built an annotated corpus of more than 13,5K documents, which is used to evaluate various deep learning architectures. The obtained results are very promising, where the most accurate was the DzaraShield . Looking ahead, there is significant potential to enhance inference speed, particularly for the Dziribert-based and multilingual models. While this project primarily focused on Arabic characters, our next step will be to address the dialect when written in Latin characters. Embracing both Arabic and Latin characters will more accurately capture the nuances of the written Algerian dialect. Finally, we plan to expand our corpus size and explore alternative deep-learning architectures. 7 Acknowledgments We would like to thank every person who has contributed to this project: Micha Freidin, Viktor Ivanenko, Piyush Aaryan, Yassine Elboustani, Tasneem Elyamany, Cephars Bonacci, Nolan Wang and Lydia Khelifa Chibout. We would also like to thank Omdena organization for giving us this valuable opportunity References Sean MacAvaney, Hao-Ren Yao, Eugene Yang, Katina Russell, Nazli Goharian, and Ophir Frieder. Hate speech detection: Challenges and solutions. PloS one, 14(8):e0221152, 2019. Mirela P Bogdani, Federico Faloppa, and Xheni Karaj. Beyond definitions. a call for action against hate speech in albania. a comprehensive study november 2021. 2021. JT Nockleby. hate speech in encyclopedia of the american constitution. electronic journal of academic and special librarianship. 2000. Imane Guellil, Ahsan Adeel, Faical Azouaou, Mohamed Boubred, Yousra Houichi, and Akram Abdelhaq Moumna. Ara-women-hate: An annotated corpus dedicated to hate speech detection against women in the arabic community. In Proceedings of the Workshop on Dataset Creation for Lower-Resourced Languages within the 13th Language Resources and Evaluation Conference, pages 68–75, 2022. Thomas Davidson, Dana Warmsley, Michael Macy, and Ingmar Weber. Automated hate speech detection and the problem of offensive language. In Proceedings of the international AAAI conference on web and social media, volume 11, pages 512–515, 2017. Paula Fortuna and Sérgio Nunes. A survey on automatic detection of hate speech in text. ACM Computing Surveys (CSUR), 51(4):1–30, 2018. Anna Schmidt and Michael Wiegand. A survey on hate speech detection using natural language processing. In Proceedings of the fifth international workshop on natural language processing for social media, pages 1–10, 2017. Delphine Battistelli, Cyril Bruneau, and Valentina Dragos. Building a formal model for hate detection in french corpora. Procedia Computer Science, 176:2358–2365, 2020. Fatimah Alkomah and Xiaogang Ma. A literature review of textual hate speech detection methods and datasets. Information, 13(6):273, 2022. Flor Miriam Plaza-del Arco, M Dolores Molina-González, L Alfonso Urena-López, and M Teresa Martín-Valdivia. Comparing pre-trained language models for spanish hate speech detection. Expert Systems with Applications, 166: 114120, 2021. Nuha Albadi, Maram Kurdi, and Shivakant Mishra. Are they our brothers? analysis and detection of religious hate speech in the arabic twittersphere. In 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), pages 69–76. IEEE, 2018. Hala Mulki, Hatem Haddad, Chedi Bechikh Ali, and Halima Alshabani. L-hsab: A levantine twitter dataset for hate speech and abusive language. In Proceedings of the third workshop on abusive language online, pages 111–118, 2019. Fréha Mezzoudj, Mourad Loukam, and Fatma Zohra Belkredim. Arabic algerian oranee dialectal language modelling oriented topic. International Journal of Informatics and Applied Mathematics, 2(2):1–14, 2019. Oussama Boucherit and Kheireddine Abainia. Offensive language detection in under-resourced algerian dialectal arabic language. arXiv preprint arXiv:2203.10024, 2022. Djamila Menifi, Wiam Moussa, and Ahmed Cherif Mazari. Transfer Learning and Deep Learning for Multilingual Algerian Dialect Hate Speech Detection. PhD thesis, 2022. Adel Abdelli, Fayçal Guerrouf, Okba Tibermacine, and Belkacem Abdelli. Sentiment analysis of arabic algerian dialect using a supervised method. In 2019 International Conference on Intelligent Systems and Advanced Computing Sciences (ISACS), pages 1–6. IEEE, 2019. Imane Guellil, Ahsan Adeel, Faical Azouaou, Mohamed Boubred, Yousra Houichi, and Akram Abdelhaq Moumna. Sexism detection: The first corpus in algerian dialect with a code-switching in arabic/french and english. arXiv preprint arXiv:2104.01443, 2021. Ona De Gibert, Naiara Perez, Aitor García-Pablos, and Montse Cuadros. Hate speech dataset from a white supremacy forum. arXiv preprint arXiv:1809.04444, 2018. Areej Al-Hassan and Hmood Al-Dossari. Detection of hate speech in social networks: a survey on multilingual corpus. In 6th international conference on computer science and information technology, volume 10, pages 10–5121, 2019. Naaima Boudad, Rdouan Faizi, Rachid Oulad Haj Thami, and Raddouane Chiheb. Sentiment analysis in arabic: A review of the literature. Ain Shams Engineering Journal, 9(4):2479–2490, 2018. Nizar Y Habash. Introduction to Arabic natural language processing. Springer Nature, 2022. Abdul-Baquee M Sharaf and Eric Atwell. Qurana: Corpus of the quran annotated with pronominal anaphora. In Lrec, pages 130–137, 2012. Kheireddine Abainia, Kenza Kara, and Tassadit Hamouni. A new corpus and lexicon for offensive tamazight language detection. In Proceedings of the 7th International Workshop on Social Media World Sensors, pages 1–6, 2022. Ahmed Cherif Mazari and Hamza Kheddar. Deep learning-based analysis of algerian dialect dataset targeted hate speech, offensive language and cyberbullying. International Journal of Computing and Digital Systems, 2023. Imane Guellil, Ahsan Adeel, Faical Azouaou, Sara Chennoufi, Hanene Maafi, and Thinhinane Hamitouche. Detecting hate speech against politicians in arabic community on social media. International Journal of Web Information Systems, 16(3):295–313, 2020. Djamila Mohdeb, Meriem Laifa, Fayssal Zerargui, and Omar Benzaoui. Evaluating transfer learning approach for detecting arabic anti-refugee/migrant speech on social media. Aslib Journal of Information Management, 74(6): 1070–1088, 2022. Reem ALBayari and Sherief Abdallah. Instagram-based benchmark dataset for cyberbullying detection in arabic text. Data, 7(7):83, 2022. Monirah A Al-Ajlan and Mourad Ykhlef. Optimized twitter cyberbullying detection based on deep learning. In 2018 21st Saudi Computer Society National Computer Conference (NCC), pages 1–5. IEEE, 2018. Batoul Haidar, Maroun Chamoun, and Ahmed Serhrouchni. Arabic cyberbullying detection: Enhancing performance by using ensemble machine learning. In 2019 international conference on internet of things (ithings) and ieee green computing and communications (greencom) and ieee cyber, physical and social computing (cpscom) and ieee smart data (smartdata), pages 323–327. IEEE, 2019. Hamdy Mubarak, Kareem Darwish, and Walid Magdy. Abusive language detection on arabic social media. In Proceedings of the first workshop on abusive language online, pages 52–56, 2017. Azalden Alakrot, Liam Murray, and Nikola S Nikolov. Dataset construction for the detection of anti-social behaviour in online communication in arabic. Procedia Computer Science, 142:174–181, 2018. Hatem Haddad, Hala Mulki, and Asma Oueslati. T-hsab: A tunisian hate speech and abusive dataset. In International conference on Arabic language processing, pages 251–263. Springer, 2019. Hamdy Mubarak, Ammar Rashed, Kareem Darwish, Younes Samih, and Ahmed Abdelali. Arabic offensive language on twitter: Analysis and experiments. arXiv preprint arXiv:2004.02192, 2020. Hala Mulki and Bilal Ghanem. Let-mi: an arabic levantine twitter dataset for misogynistic language. arXiv preprint arXiv:2103.10195, 2021. Rehab Duwairi, Amena Hayajneh, and Muhannad Quwaider. A deep learning framework for automatic detection of hate speech embedded in arabic tweets. Arabian Journal for Science and Engineering, 46:4001–4014, 2021. Wassen Aldjanabi, Abdelghani Dahou, Mohammed AA Al-qaness, Mohamed Abd Elaziz, Ahmed Mohamed Helmi, and Robertas Damaševicius. Arabic offensive and hate speech detection using a cross-corpora multi-task learning ˇ model. In Informatics, volume 8, page 69. MDPI, 2021. Batoul Haidar, Maroun Chamoun, and Ahmed Serhrouchni. A multilingual system for cyberbullying detection: Arabic content detection using machine learning. Advances in Science, Technology and Engineering Systems Journal, 2(6): 275–284, 2017. Ahmed Abdelali, Kareem Darwish, Nadir Durrani, and Hamdy Mubarak. Farasa: A fast and furious segmenter for arabic. In Proceedings of the 2016 conference of the North American chapter of the association for computational linguistics: Demonstrations, pages 11–16, 2016. Amine Abdaoui, Mohamed Berrimi, Mourad Oussalah, and Abdelouahab Moussaoui. Dziribert: a pre-trained language model for the algerian dialect. arXiv preprint arXiv:2109.12346, 2021. Sourab Mangrulkar, Sylvain Gugger, Lysandre Debut, Younes Belkada, and Sayak Paul. Peft: State-of-the-art parameterefficient fine-tuning methods. https://github.com/huggingface/peft, 2022. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021. Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, and Furu Wei. Text embeddings by weakly-supervised contrastive pre-training. arXiv preprint arXiv:2212.03533, 2022. E Moatez Billah Nagoudi, A Elmadany, and M Abdul-Mageed. Arat5: Text-to-text transformers for arabic language understanding and generation. arXiv preprint arXiv:2109.12068, 2021.

Stories, News & Mentions about LANA

Lana’s Stories on HackerNoon