Comedian, podcast host, actor and author Bert Kreischer is bringing his Tops Off World Tour to BOK Center for a Friday, May 17 performance. Tickets go on sale 10 a.m. Friday, Oct. 13 and will be ...

Events announced this week: Bert Kreischer coming to BOK, Ron White returning to River Spirit

COLUMBIA − After graduating only months ago from the University of Missouri, Corrine Baum went to Israel to teach English to disadvantaged children. Just a week ago, day-to-day life was normal ...

MU alum escapes Israel as war unfolds

DECATUR, Ill. (WICS) — Millikin University (MU) will once again be hosting its annual event, Boo at MU. Visitors can enjoy a Trunk or Treat in the University Commons parking lot as well as activities, music, candy, prizes, and more in a family-friendly ...

Millikin University will host its annual Boo at MU

DECATUR, Ill. (WICS) — Millikin University (MU) will once again be hosting its annual event, Boo at MU. Visitors can enjoy a Trunk or Treat in the University Commons parking lot as well as ...

The MU Faculty Council approved a statement of solidarity with those affected by new state laws and MU Health Care’s halt on gender-affirming care for minors, two weeks after it failed to hold a ...

MU Faculty Council expresses solidarity with those affected by gender-affirming care policy

After selling out four shows in Tempe and the TD Bank Garden in Boston, enormously popular comedian Bert Kreischer is understandably adding dates to his seemingly unending tour. He'll add 34 ...

Bert Kreischer adds Austin, San Antonio dates to Tops Off World Tour

Bert Kreischer attends the premiere of Whitney presents The OnlyFans Roast of Bert Kreischer on OFTV at Sunset at the Edition on March 29, 2023 in West Hollywood, California. (Photo by Phillip ...

Comedian Bert Kreischer is coming to San Antonio for stand-up tour

EL PASO, Texas (KFOX14/CBS4) — Comedian Bert Kreischer, aka "The Machine" is coming to El Paso. Renowned comedian, podcast host, actor and author Bert Kreischer is extending his Tops Off World ...

Bert Kreischer, 'The Machine,' extends Tops Off World Tour to El Paso in January 2024

LSE continues to straddle the gap between the LS and Onyx, but now with four-wheel-drive. Compared to the fisty-cuffs battle between the Toyota Hilux and Ford Ranger for the title of country’s ...

Newly added mid-spec 4WD Isuzu MU-X shows top isn’t always best

Micron Technology MU announced on Thursday that it has officially started the construction of a new memory chip factory in Boise, Idaho. This will mark the United States’ first memory chip ...

Micron (MU) Begins New Fab Construction to Expand US Chip Output

Mubert

mubert

10 Best AI Music Generators for Streamlined Music Generation Process

Təkmilləşdirilmiş Musiqi Yaratma Prosesi üçün 10 Ən Yaxşı AI Musiqi Generatorları

10 nejlepších hudebních generátorů AI pro zjednodušený proces generování hudby

Οι 10 καλύτερες γεννήτριες μουσικής AI για βελτιωμένη διαδικασία δημιουργίας μουσικής

Los 10 mejores generadores de música con IA para un proceso de generación de música optimizado

Os 10 mellores xeradores de música AI para un proceso de xeración de música simplificado

10 pi bon jeneratè mizik AI pou pwosesis jenerasyon mizik rasyonalize

A 10 legjobb AI zenegenerátor az egyszerűsített zenegenerálási folyamathoz

合理化された音楽生成プロセスのための 10 のベスト AI 音楽ジェネレーター

Mpamorona Mozika AI 10 tsara indrindra ho an'ny fizotry ny famoronana mozika mirindra

Imashini 10 nziza yumuziki wa AI kubikorwa byogutunganya umuziki

10 Gjeneruesit më të mirë të muzikës me AI për procesin e gjenerimit të muzikës më të thjeshtë

10 Tijeneretara ta Vuyimbeleri bya AI letinene ta Endlelo ro Endla Vuyimbeleri lebyi Olovisiweke

88 Stories To Learn About Music

Meet the Winners of Startups in North and Central America

The role of music for business and user experience as well as ways to make the best of it, discussed by me, Alex Kochetkov, founder and CEO of Mubert , the first generative music streaming service. The impact of music in the life of human beings cannot be overrated. People would rather do without sports, newspapers, movies, internet and smartphones. But music is important. When we listen to masterpieces of Debussy or Wagner it is truly astounding how such beauty can be created. The essence of talent is a mystery. We all have been to live shows at least once, and this feeling evoked by music is unexplainable. According to numerous scientific studies, music makes people think of themselves, distract from the outside world and forget about reality, calm down and acquire additional meaning to life. It is all about subconsciousness. Music is magic and music has power. Prominent businesses leverage this knowledge and it bears fruit. “Music gives a soul to the universe, wings to the mind, flight to the imagination and life to everything.” ― Plato. I think the great Athenian philosopher will not mind if I add “music gives profit to business owners”. In a world where companies fight for unparalleled user experience, sound appears to be a decisive competitive differentiator. The secret is in the ‘right’ music being played. Again, the relevance matters a lot when we speak of personalized UX. Customers will leave public places in the case of the music not suiting their brand identity. On the contrary, they are more likely to stay longer, come back and buy more if they like what they hear. The trick is to find the balance and satisfy both brand profile and customer tastes. There is more to it. Customers are concerned about the moral aspects of music accompaniment in public places. They want things to be right, and musicians should be appreciated and paid for their music being used. And, if people find out businesses act unethically and illegally, their attitude towards a business will change to negative. A watershed between companies and their clientele implies a conflict of interests. While customers want justice, every business owner cares about cutting costs. Paying royalties is a hit to the wallet. Streaming services are popular, however, they cannot provide a solution to cover all needs at once. The development of music recommender systems or MRS is the subproduct of streaming services which have mushroomed in recent years. Thanks to services such as Apple Music, Spotify, and Pandora,  people acquire access to immense catalogs. In the old days it took quite an effort to find the music you want. Today, millions of tracks and artists can be found in a matter of seconds. People hate having multiple choices. In this respect, MRS are used to filter and narrow down offered music based on the specific preferences of the user. Basically, highly relevant recommendations are the exact feature listeners want to see in those online streaming services. Though, they have a number of serious flaws. Music recommender systems: far from perfect MRS vulnerability boils down to one thing: they fail to take into account a vast number of factors that influence decision-making in music. Current recommendation systems waltz around the user-item interaction approach and descriptors based on content. In reality, music choices at a specific point in time are driven by internal and external factors as well as context. Emotional state, plus weather conditions, plus social surrounding form this particular narrow  halfway we want to walk through here and now. It is does not mean we will necessarily go back there again. Nowadays, MRS are unable to provide relevant recommendations for newcomers; the ‘cold start’ issue as we know it — not enough data. Also, when we manage our playlists, it’s not the order of tracks that matters: we pay attention to the entire ensemble and the transition from one track to another. Adding one more track to the existing playlist, so that it fits perfectly, is another challenge. Recommender systems, in fact, represent one side of the problem with streaming services. Streaming services: the big picture The debate on the sustainability of streaming services revolve around money and ethics. To what extent are these platforms transparent, fair in terms of artists recognition? Record labels have been holding a grudge for a long time. In the case of Spotify, the platform did not hesitate to make deals with artists directly cutting out the record labels — this is one of the practices that questions morality. Armies of paying subscribers across the globe do not seem to have saved the day. Despite its growing customer base Spotify, for instance, used to be unprofitable for years since the day of its inception. The price of subscription is one of the major motivators for the illegal consumption of music. Streaming services — the largest of them — may be different in design; they provide the same catalog with approximately 300 million tracks. Here we cannot talk about differentiation as compared to video streaming platforms that have original content. Embarrassing news on data security appeared at the beginning of the year. There has been a lot of speculation on how Spotify : there is a lot happening in terms of online marketing and data sharing you would never think of. The giants of the online streaming market are the ones that are targeted by constant GDPR complaints. They refer to user rights structural violations: streaming platforms have automated response systems for access requests. Users do not get sufficient data, for instance, who their data is shared with. In addition, little protection is provided to user privacy. sells user data You have to face it: the future is here. Artificial intelligence changes companies, businesses, industries; it has changed the way we interact with the world. At first, AI approached the area of figures and computation. When neural networks entered the world of art painters, musicians, and artists all over the world started getting nervous. In the context of art and AI there are a lot of debates on what makes music human, whether there is anything an algorithm cannot learn. The big question is — will AI replace or help? Let’s put the alarmist paradigm aside and look in the direction of the opportunities. How music met AI From a business perspective, sound is an important element of the UX. By providing personalized experiences through music, a company creates its own unique atmosphere and differentiates itself. At the same time, businesses in horeca domain are interested in functional music: it’s not meant to distract by vivid emotional coloring. French restaurant clients, for instance, would hardly enjoy their glass of wine listening to Bjork or Puccini’s Tosca. So, functional music must provide a light background, a supplement to an environment. Driven by this idea, we elaborated the framework to tackle all the existing challenges and provide an on-demand solution. Mubert AI-based generative music streaming services platform, kills two, or even three birds with one stone. The service broadcasts non-stop music via API; is capable of giving businesses exactly what they want —personalization, cost-cutting, differentiation, unique UX for their customers. Algorithms write music in real-time mode: they compile never-ending streams of music that do not need to be managed. AI accesses a large database of patterns written by human composers. Neuronets fulfil the task of conductor: it takes into account the current circumstances and manages tiny musical pieces accordingly. Mubert’s innovative approach comes with ease of use. The platform’s technology does not have all stereotypical functions that other streaming services have — the appropriate streams are in one-click proximity. Moreover, the platform is adaptive to external commands or device sensors which makes it a suitable solution for smart homes and voice assistants. Ultimately, the service provides businesses with copyright protection. Mubert addresses legal issues via copyright-free music — no contracts, no royalties. What separates a business from personalized music accompaniment? Only the purchase of subscription. That’s it. The flexibility of generative music has been successfully tested in a number of industries, and gaming is one of them. Historically, gaming streamers were banned from services for  illegally using music in the background. The problem with copyright holders is not new, and , with its 1 million monthly users, reaped the benefits of generative music. The partnership resulted in around 66 thousand unique streamers creating approximately 780 hours of unique content within the first month. Omlet Arcade Personalized user experience literally means getting personal with customers: their lives are much more than content, they go far beyond that. The audience is sophisticated. Everyone needs to feel privileged, special, and have a sense of being considered important. Music appeals both emotionally and intangibly, which further pays off with rather tangible sums. We are all customers and when it comes down to what we like — nobody can deceive us. We know when a business cares and tries its best to please us. Leveraging cutting-edge tech for unparalleled UX is a win-win option every business owner should consider.

Music and tech: the perfect mixture for business purposes

Authors:
(1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel;
(2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Related Work Methodology Experimental Settings Results Conclusion and Limitations Bibliographical References 5. Results The results are summarized in Table 1. The table rows are organized into three distinct groups. The first group features combinations of various baseline models. The second group showcases combinations incorporating our proprietary HeArBERT model, while the third group highlights combinations involving the contrasting HeArBERTNT model. We train multiple machine-translation combinations, in both directions Arabic-to-Hebrew and Hebrewto-Arabic, based on the same encoder-decoder architecture, initialized with different language model combinations. We assign various combinations of language models to the encoder and decoder components, ensuring that the selected models align with the respective source and target languages. In other words, we make sure that the language model which we use to initialize the encoder, is compatible with the system’s source language. Since mBERT and CAMeLBERTET , as well as our two models HeArBERT and HeArBERTNT , can potentially handle both languages, we experiment with combinations where each of them is assigned to both, the encoder and decoder components at the same time. The results demonstrate that the second group, which utilizes our HeArBERT to initialize a Hebrew decoder, surpasses the performance of all other combinations. Notably, the pairing of GigaBERT with HeArBERT is the standout performer across both directions. It surpasses the performance of using HeArBERT for initializing the Arabic encoder by only a few minor points. The performance of HeArBERTNT , as reported in the third group is consistently lower than HeArBERT in both directions. The difference seems to be more significant in the Arabic-to-Hebrew direction. Overall, the results show that transliterating the Arabic training texts into the Hebrew script as a pre-processing step is beneficial for an Arabic-to-Hebrew machine translation system. The impact of the transliteration proves to be less pronounced in the reverse translation direction. Using the extended (ET) version of CAMeLBERT has a reasonable performance. However, it performs much worse than the best model in both directions, indicating that extending the vocabulary with transliterated Arabic tokens does contribute to better capturing the meaning of Hebrew tokens in context. This implies that joint pre-training on both languages is essential for achieving a more robust language representation. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Authors: Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Abstract and Introduction Related Work Related Work Methodology Methodology Experimental Settings Experimental Settings Results Results Conclusion and Limitations Conclusion and Limitations Bibliographical References Bibliographical References 5. Results The results are summarized in Table 1. The table rows are organized into three distinct groups. The first group features combinations of various baseline models. The second group showcases combinations incorporating our proprietary HeArBERT model, while the third group highlights combinations involving the contrasting HeArBERTNT model. We train multiple machine-translation combinations, in both directions Arabic-to-Hebrew and Hebrewto-Arabic, based on the same encoder-decoder architecture, initialized with different language model combinations. We assign various combinations of language models to the encoder and decoder components, ensuring that the selected models align with the respective source and target languages. In other words, we make sure that the language model which we use to initialize the encoder, is compatible with the system’s source language. Since mBERT and CAMeLBERTET , as well as our two models HeArBERT and HeArBERTNT , can potentially handle both languages, we experiment with combinations where each of them is assigned to both, the encoder and decoder components at the same time. The results demonstrate that the second group, which utilizes our HeArBERT to initialize a Hebrew decoder, surpasses the performance of all other combinations. Notably, the pairing of GigaBERT with HeArBERT is the standout performer across both directions. It surpasses the performance of using HeArBERT for initializing the Arabic encoder by only a few minor points. The performance of HeArBERTNT , as reported in the third group is consistently lower than HeArBERT in both directions. The difference seems to be more significant in the Arabic-to-Hebrew direction. Overall, the results show that transliterating the Arabic training texts into the Hebrew script as a pre-processing step is beneficial for an Arabic-to-Hebrew machine translation system. The impact of the transliteration proves to be less pronounced in the reverse translation direction. Using the extended (ET) version of CAMeLBERT has a reasonable performance. However, it performs much worse than the best model in both directions, indicating that extending the vocabulary with transliterated Arabic tokens does contribute to better capturing the meaning of Hebrew tokens in context. This implies that joint pre-training on both languages is essential for achieving a more robust language representation. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Training a Bilingual Language Model by Mapping Tokens onto a Shared Character Space: Results

Authors:
(1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel;
(2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Related Work Methodology Experimental Settings Results Conclusion and Limitations Bibliographical References 4. Experimental Settings Machine Translation. Our machine-translation architecture is based on a simple encoder-decoder framework, which we initialize using weights of the BERT model in focus.[5] For example, if we want our model to translate from Hebrew to Arabic, we might initialize the encoder with the model weights of mBERT and the decoder with the weights of CAMelBERT. To fine-tune the model on machine translation, we use the new “Kol Zchut” (in English, “All Rights”) Hebrew-Arabic parallel corpus[6] which contains over 4,000 parallel articles in the civil-legal domain, corresponding to 140,000 sentence-pairs in Arabic and Hebrew containing 2.13M and 1.8M words respectively. To the best of our knowledge, ours is the first work to report machine translation results using this resource; therefore, no established baseline or benchmark exists for comparison. As the corpus is provided without an official train/test split, we apply a random split with 80% of the data being allocated for training and the remaining 20% for testing, using the train_test_split function of scikit-Learn with a random seed of 42. We evaluate our HeArBERT-based translation against an equivalent system initialized using other models. The standard BLEU metric (Papineni et al., 2002) is employed to contrast the system’s generated translation with the sole reference translation present in the corpus. Each system is fine-tuned for the duration of ten epochs, and we report the best performance observed across all epochs. Baseline Language Models. To contrast HeArBERT with an equivalent model trained on texts in both Arabic and Hebrew scripts, we pre-train another model and tokenizer following the same procedure, but without the transliteration preprocessing for the Arabic data. This model is denoted with a subscript "NT" (no transliteration): HeArBERTNT. We compare our model with a number of existing models. The initial model, mBERT, was originally pre-trained on a range of languages, including both Arabic and Hebrew. We also chose a couple of monolingual Arabic language models, with specific versions from Hugging Face provided in a footnote. Specifically, we use CAMeLBERT[7] and GigaBERT[8]. In some experiments, we adopt a technique inspired by Rom and Bar (2021). This involves expanding the vocabulary of an existing Arabic language-model’s tokenizer by appending a Hebrew-transliterated version of each Arabic token and associating it with the original token identifier. We denote such extended models by adding “ET” (extended tokenizer) to the model name. All these models share the same architecture size as our proposed model. This paper is available on arxiv under CC BY 4.0 DEED license. [5] We use HuggingFace’s EncoderDecoderModel. [6] https://releases.iahlt.org/ [7] CAMeL-Lab/bert-base-arabic-camelbert-mix [8] lanwuwei/GigaBERT-v4-Arabic-andEnglish Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Authors: Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Abstract and Introduction Related Work Related Work Methodology Methodology Experimental Settings Experimental Settings Results Results Conclusion and Limitations Conclusion and Limitations Bibliographical References Bibliographical References 4. Experimental Settings Machine Translation. Our machine-translation architecture is based on a simple encoder-decoder framework, which we initialize using weights of the BERT model in focus.[5] For example, if we want our model to translate from Hebrew to Arabic, we might initialize the encoder with the model weights of mBERT and the decoder with the weights of CAMelBERT. To fine-tune the model on machine translation, we use the new “Kol Zchut” (in English, “All Rights”) Hebrew-Arabic parallel corpus[6] which contains over 4,000 parallel articles in the civil-legal domain, corresponding to 140,000 sentence-pairs in Arabic and Hebrew containing 2.13M and 1.8M words respectively. To the best of our knowledge, ours is the first work to report machine translation results using this resource; therefore, no established baseline or benchmark exists for comparison. As the corpus is provided without an official train/test split, we apply a random split with 80% of the data being allocated for training and the remaining 20% for testing, using the train_test_split function of scikit-Learn with a random seed of 42. We evaluate our HeArBERT-based translation against an equivalent system initialized using other models. The standard BLEU metric (Papineni et al., 2002) is employed to contrast the system’s generated translation with the sole reference translation present in the corpus. Each system is fine-tuned for the duration of ten epochs, and we report the best performance observed across all epochs. Baseline Language Models. To contrast HeArBERT with an equivalent model trained on texts in both Arabic and Hebrew scripts, we pre-train another model and tokenizer following the same procedure, but without the transliteration preprocessing for the Arabic data. This model is denoted with a subscript "NT" (no transliteration): HeArBERTNT. Baseline Language Models. We compare our model with a number of existing models. The initial model, mBERT, was originally pre-trained on a range of languages, including both Arabic and Hebrew. We also chose a couple of monolingual Arabic language models, with specific versions from Hugging Face provided in a footnote. Specifically, we use CAMeLBERT[7] and GigaBERT[8]. In some experiments, we adopt a technique inspired by Rom and Bar (2021). This involves expanding the vocabulary of an existing Arabic language-model’s tokenizer by appending a Hebrew-transliterated version of each Arabic token and associating it with the original token identifier. We denote such extended models by adding “ET” (extended tokenizer) to the model name. All these models share the same architecture size as our proposed model. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [5] We use HuggingFace’s EncoderDecoderModel. [6] https://releases.iahlt.org/ [7] CAMeL-Lab/bert-base-arabic-camelbert-mix [8] lanwuwei/GigaBERT-v4-Arabic-andEnglish

Benchmarking HeArBERT: Comparative Analysis with Traditional and Extended Language Models

Authors:
(1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel;
(2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Related Work Methodology Experimental Settings Results Conclusion and Limitations Bibliographical References 2. Related Work K et al. (2020) have suggested that structural similarity of languages is essential for language model’s multilingual generalization capabilities. Their suggestion was further discussed by Dufter and Schütze (2020), who highlighted the essential components for a model to possess “multilinguality”, and show that the order of the words in the sentence is key to the model’s cross-lingual generalization capabilities. mBERT, as introduced by Devlin et al. (2019), was a pioneering language model that encompassed multiple languages, including Arabic and Hebrew. However, both Arabic and Hebrew are significantly under-represented in the pre-training data, resulting in inferior performance compared to the equivalent monolingual models on various downstream tasks (Antoun et al., 2020; Lan et al., 2020; Chriqui and Yahav, 2022; Seker et al., 2022). GigaBERT (Lan et al., 2020) is another multilingual model, trained for English and Arabic. However, the best results for most of the known NLP tasks are typically achieved by one of the large monolingual models in both Arabic and Hebrew. CAMeLBERT (Inoue et al., 2021), is one of those models. It is trained on texts written in Modern Standard Arabic (MSA), Classical Arabic, as well as dialectal variants. In the realm of Hebrew language models, AlephBERT (Seker et al., 2022) stands out as one of the leading performers, alongside others like HeBERT (Chriqui and Yahav, 2022). Among other datasets, the monolingual models mentioned above use the relevant parts of the OSCAR dataset (Ortiz Suárez et al., 2020) for training. Our model relies solely on the OSCAR dataset for both Hebrew and Arabic, resulting in a considerably smaller total number of words for each language in comparison to the existing monolingual language models. The effect of transliteration on cross-lingual generalization were discussed previously in (Dhamecha et al., 2021; Chau and Smith, 2021) and more recently in (Moosa et al., 2023; Purkayastha et al., 2023). None of these works study the languages of our focus: Arabic and Hebrew. Dhamecha et al. (2021) focused on languages from the Indo-Aryan family, which has been studied before for cross-lingual generalization and also has several publicly available multilingual models. To the best of our knowledge, our work is first to study generalization between Arabic and Hebrew and no multilingual masked language models that include both languages have been published apart from mBERT. Chau and Smith (2021) address the generalization from high- to low-resourced languages. However, both Arabic and Hebrew are currently considered medium- to high-resourced languages. Furthermore, their evaluation focuses solely on tokenlevel classification tasks, such as dependency parsing and part-of-speech tagging, whereas our evaluation targets machine translation, a sequence-to-sequence bilingual task. Purkayastha et al. (2023) employ Romanization for transliteration, whereas we transliterate Arabic into the Hebrew script. Analogous to Chau and Smith (2021), their evaluation centers on token-level classification tasks, which are not addressed in our work. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Authors: Authors: (1) Aviad Rom, The Data Science Institute, Reichman University, Herzliya, Israel; (2) Kfir Bar, The Data Science Institute, Reichman University, Herzliya, Israel. Table of Links Abstract and Introduction Abstract and Introduction Related Work Related Work Methodology Methodology Experimental Settings Experimental Settings Results Results Conclusion and Limitations Conclusion and Limitations Bibliographical References Bibliographical References 2. Related Work K et al. (2020) have suggested that structural similarity of languages is essential for language model’s multilingual generalization capabilities. Their suggestion was further discussed by Dufter and Schütze (2020), who highlighted the essential components for a model to possess “multilinguality”, and show that the order of the words in the sentence is key to the model’s cross-lingual generalization capabilities. mBERT, as introduced by Devlin et al. (2019), was a pioneering language model that encompassed multiple languages, including Arabic and Hebrew. However, both Arabic and Hebrew are significantly under-represented in the pre-training data, resulting in inferior performance compared to the equivalent monolingual models on various downstream tasks (Antoun et al., 2020; Lan et al., 2020; Chriqui and Yahav, 2022; Seker et al., 2022). GigaBERT (Lan et al., 2020) is another multilingual model, trained for English and Arabic. However, the best results for most of the known NLP tasks are typically achieved by one of the large monolingual models in both Arabic and Hebrew. CAMeLBERT (Inoue et al., 2021), is one of those models. It is trained on texts written in Modern Standard Arabic (MSA), Classical Arabic, as well as dialectal variants. In the realm of Hebrew language models, AlephBERT (Seker et al., 2022) stands out as one of the leading performers, alongside others like HeBERT (Chriqui and Yahav, 2022). Among other datasets, the monolingual models mentioned above use the relevant parts of the OSCAR dataset (Ortiz Suárez et al., 2020) for training. Our model relies solely on the OSCAR dataset for both Hebrew and Arabic, resulting in a considerably smaller total number of words for each language in comparison to the existing monolingual language models. The effect of transliteration on cross-lingual generalization were discussed previously in (Dhamecha et al., 2021; Chau and Smith, 2021) and more recently in (Moosa et al., 2023; Purkayastha et al., 2023). None of these works study the languages of our focus: Arabic and Hebrew. Dhamecha et al. (2021) focused on languages from the Indo-Aryan family, which has been studied before for cross-lingual generalization and also has several publicly available multilingual models. To the best of our knowledge, our work is first to study generalization between Arabic and Hebrew and no multilingual masked language models that include both languages have been published apart from mBERT. Chau and Smith (2021) address the generalization from high- to low-resourced languages. However, both Arabic and Hebrew are currently considered medium- to high-resourced languages. Furthermore, their evaluation focuses solely on tokenlevel classification tasks, such as dependency parsing and part-of-speech tagging, whereas our evaluation targets machine translation, a sequence-to-sequence bilingual task. Purkayastha et al. (2023) employ Romanization for transliteration, whereas we transliterate Arabic into the Hebrew script. Analogous to Chau and Smith (2021), their evaluation centers on token-level classification tasks, which are not addressed in our work. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

HeArBERT: A Bilingual Model for Arabic-Hebrew Translation Using Transliteration

Learning Semantic Knowledge from Wikipedia:  Learning Concept Hierarchies from Document Categories

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

It’s been 15 years since the Marvel Cinematic Universe (MCU) began. Over 20 films have come out, but there are still plenty more stories to tell. In 2023, fans will get treated to three new films featuring established characters like Ant-Man and Captain Marvel. Outside of the MCU, Sony is still churning out their Spider-Man adjacent films. They have two films in the pipeline that are set to release this year. All in all, 2023 is shaping up to be a big year for Marvel fans. Here are all of the Marvel movies coming out in 2023. Marvel Movies Coming Out in 2023 Ant-Man and the Wasp: Quantumania Guardians of the Galaxy Vol. 3 Spider-Man: Across the Spider-Verse The Marvels Kraven the Hunter 1. Ant-Man and the Wasp: Quantumania (02-17-2023) In their third movie, Ant-Man and the Wasp go on their biggest (or smallest) adventure yet. They’ve come across and defeated their fair share of villains, but nothing can prepare them for the one that comes next. That’s because the big bad of Quantumania is none other than Kang the Conqueror (or at least, some version of him). Kang is being set as the big baddie of the MCU, and it’s exciting to see what he has in store for Ant-Man and the Wasp. But another exciting aspect of the movie is that Ant-Man’s daughter, who has appeared in the last two films, is finally following in her father’s footsteps and is becoming a full-fledged superhero. Quantumania is shaping up to be the best Ant-Man film yet, and it can’t come soon enough. 2. Guardians of the Galaxy Vol. 3 (05-05-23) Out of all the Marvel movies coming out in 2023, this might be the one that fans are the most excited to see. Vol. 3 is set to be the closing chapter for the Guardians of the Galaxy. Dave Bautista, who plays Drax, has about how this is his last performance as the character. publicly talked In , Guardians of the Galaxy Director James Gunn said, “This is the end for us, the last time people will see this team of Guardians.” Deadline’s Hero Nation Podcast That doesn’t mean that all of these characters will never be seen again, but it means that things will never be the same again. If this truly is the end, the Guardians had a hell of a ride with two successful movies and even a recently released Christmas special. 3. Spider-Man: Across the Spider-Verse (06-02-23) The only animated movie on this list, Spider-Man: Across the Spider-Verse has big shoes to fill. The first Spider-Verse movie was a huge hit; it even won the . But Across the Spider-Verse has all the potential in the world to be just as big as its predecessor. Oscar for Best Animated Feature In the first film, we got a taste of the Spider-Verse; we saw different Spider-People such as Spider-Man Noir, Spider-Ham, and Spider-Gwen. But it seems that this movie is taking viewers deeper into the multiverse. Although we don’t know the full synopsis of the movie, it looks like the Spider-People are in some kind of conflict with each other, with Spider-Man 2099 potentially being the movie’s main bad guy. 4. The Marvels (07-28-23) The last MCU movie of the year, The Marvels brings three superheroes together, Captain Marvel, Ms. Marvel, and Monica Rambeau. We don’t know much about this movie, but it has been brewing for a while. Captain Marvel made her debut in her self-titled movie, and she also had a prominent role in Avengers: Endgame. Monica also had a small appearance in Captain Marvel, but she had a more significant storyline in the Disney+ show, WandaVision. Then, there’s Ms. Marvel, who had her own Disney+ show. At the end of her show, we see Ms. Marvel switching places with Captain Marvel. Why this is happening is yet to be revealed, but all of our questions will be answered when the movie comes out. 5. Kraven the Hunter (10-06-23) The only movie on this list that isn’t a sequel, Kraven the Hunter is a film that takes place outside of the MCU. It’s being produced by Sony, and it’s a Spider-Man adjacent film that will focus on Kraven, a Spidey villain. This isn’t the first time that Sony has done this type of movie. They recently made Venom, Venom: Let There Be Carnage, and Morbius. Also, they have a and an in the works. Hypno-Hustler movie El Muerto movie Kraven is one of the few prolific Spider-Man villains who have yet to appear on the big screen, and it’s interesting to see if he can be the star of a movie. Only time will tell, but this movie does sound intriguing if not anything. More in Entertainment: Every Doctor Strange Appearance in Order All of the Ant-Man Movies in Order The Dark Knight Movies in Order Feature image source

Stories, News & Mentions about MUBERT

Mubert’s Stories on HackerNoon

Mubert's News & Mentions