Published first on TheFly – the ultimate source for real-time, market-moving breaking financial news.

Cigna revives merger talks with Humana, Bloomberg reports

Shares of Humana, which has a market capitalization of about $32 billion, were up about 6% in after-hours trading on Friday, while those of Cigna were down about 5%. Cigna was valued at about $94 billion, according to data compiled by LSEG. Cigna and Humana declined to comment.

Cigna resumes mega merger talks with rival health insurer Humana, Bloomberg News reports

Shares of Humana were up about 5% in after-hours trading on Friday, while those of Cigna were down about 2%. Last year, Reuters reported that Cigna ended its attempt to negotiate an acquisition of Humana after the pair failed to agree on a price. Cigna did not immediately respond to a Reuters ...

Cigna resumes merger discussions with Humana, Bloomberg News reports

Executives at Fortune 500 companies from JPMorgan Chase to Cigna Healthcare are reassuring investors that they remain committed to diversity, equity, and inclusion principles, even as mounting attacks on DEI erode certain programs.

Jamie Dimon Among Major CEOs Defying DEI Backlash for Businesses - Bloomberg

GuideWell, a mutual insurance holding company serving more than 38 million people, announced the appointment of Thurman Justice as President and Chief Executive Officer (CEO) of Triple-S Management Corporation,

GuideWell Announces Appointment of Thurman Justice as Triple-S Chief Executive Officer

It’s time to pick health coverage for next year on the Affordable Care Act’s insurance marketplaces, and experts caution against focusing your choice on price alone. In order

It's time to buy health insurance through the marketplace. Experts suggest doing your research first

According to the study “... most expressed flexible time off or work arrangements, followed by a private health insurance plan and mental health support.”

Amid cost of living crisis, SG workers want more health support from employers—study

Exchange insurers are charging more, expanding geographic footprints and creating disease-specific policies to separate themselves from the competition.

Marketplace open enrollment has arrived. Here's what to know

The ACA marketplace’s open enrollment season is also open season for scammers, who look to prey on people overwhelmed by the health insurance marketplace.

No, you won’t get a $6,400 spending allowance from a free health insurance plan

Medicare beneficiaries should watch for shopping surprises when choosing a Medigap plan to avoid paying too much for coverage.

Watch for These 3 Medigap Shopping Surprises to Avoid Overpaying

Cigna

cigna

Inside the Degen Mind: Understanding Crypto Communities and Why We Are the Way We Are

በዴጄን አእምሮ ውስጥ-የ Crypto ማህበረሰቦችን መረዳት እና ለምን እነሱ እንደነበሩ መረዳት

Unutar uma Degena: Razumijevanje kripto zajednica i zašto su takve kakve jesu

Dentro de la mente de Degen: comprensión de las comunidades criptográficas y por qué son como son

डेगन माइंड के अंदर: क्रिप्टो समुदायों को समझना और वे जिस तरह से हैं, क्यों हैं

Inside the Degen Mind: הבנת קהילות קריפטו ומדוע הן כפי שהן

Degen の心の中: 暗号通貨コミュニティを理解し、なぜそれが今のような形になっているのか

Na kati ya makanisi ya Degen: Kososola ba Communautés Crypto mpe Pourquoi Bazali Ndenge Bazali

Дегений оюун санааны дотор: Крипто нийгэмлэгүүд ба тэд яагаад ийм байдгийг ойлгох

डिजेन माइन्ड भित्र: क्रिप्टो समुदायहरू बुझ्दै र तिनीहरू किन छन् जसरी तिनीहरू छन्

Wewnątrz umysłu Degena: zrozumienie społeczności kryptograficznych i dlaczego są takie, jakie są

Degen aklynyň içinde: Kripto jemgyýetlerine we näme üçin olaryň beýle bolandygyna düşünmek

Bên trong tâm trí Degen: Hiểu về cộng đồng tiền điện tử và lý do tại sao họ lại như vậy

Thanks to big data, today an organization can quickly obtain the necessary information from an unordered data set and deploy it effectively. The growing popularity of big data analytics has led to a significant increase in the number of companies providing big data solutions and related services. Today, there are in the global market, and this number may be very confusing. Big data skills, experience, and related technical expertise can significantly vary among vendors, which may puzzle anyone looking for a big data service provider. almost 1000 big data analytics companies How to identify the best big data consulting firms I set myself the goal to identify the most reliable big data consulting companies on today’s IT market. To do this, I investigated hundreds of data providers and studied their characteristics. Each company was analyzed according to the set of specifically designed criteria. Among the criteria are big data expertise, major clients, offering strategy, track record, and online reviews. Below you can see the results of the conducted market research. 1. Itransition Founded in 1998, Itransition is a custom software development company that provides big data consulting and platform engineering services. The company also offers business intelligence (BI) portals development and boosts them with machine learning (ML) and other technologies. The company offers data store design, data migration, performance optimization, and support services. The company’s client base includes IBM, Expedia, Toyota, eBay, Xerox, PayPal, and Adidas. With more than 150 completed projects and 810 clients worldwide, IAOP recognizes the company as one of The Global Outsourcing 100. Denver, CO Headquarters: Expedia, Xerox, Toyota, eBay, IBM, PayPal, Adidas Core clients: $25 — $49 Pay per hour: 1000–9999 Number of employees: Website: https://www.itransition.com/ 2. Saviant Consulting Saviant is a consulting company with more than 120 experts in its team and more than 100 clients in various countries, including UK, Canada, and the US. The company specifically focuses on Energy, Logistics, and Retail. Among the company’s big data services are big data analytics and management, full-cycle big data solution development, and more. Pune, India Headquarters: JJ Food Service, BuilderMT, Transparent Technologies, Inc Core clients: Not Revealed Pay per hour: 50–249 Number of employees: Website: https://www.saviantconsulting.com/ 3. Squadex Headquartered in California, USA, Squadex is a technology consulting and engineering company that provides big data services. It offers big data analytics, consulting, and big data platform engineering. Squadex is an AWS certified consulting partner that works with companies of different size, from small startups to large enterprises. Palo Alto, CA Headquarters: Model N, NinthDecimal, Nitrio Core clients: Not Revealed Pay per hour: 250–999 Number of employees: Website: https://squadex.com/ 4. CBIG Consulting Founded in 2002, CBIG Consulting is an American company offering business intelligence and big data analytics solutions. Besides analytics and consulting, the company provides big data solutions architecture. For this purpose, CBIG Consulting employs Hadoop distributions, Hive, Sqoop, etc. The company’s industrial focus is on Finance, Healthcare, and Retail. Chicago, IL Headquarters: Abbott, Astellas, Baxter, Cigna, DePaul University Core clients: $150 — $199 Pay per hour: 50–249 Number of employees: Website: https://www.cbigconsulting.com/ 5. Clavax Technologies Founded in 2011, Clavax is a software development company headquartered in San Jose, California. It delivers big data consulting and implementation. By employing big data technologies, the company provides services for data warehousing, data processing, and data integration. Clavax’ client list includes PWC, Football.com, Bookfari, and Honeywell. San Jose, CA Headquarters: PWC, Football.com, Bookfari, Honeywell, Jabil Core clients: $25 — $49 Pay per hour: 50–249 Number of employees: Website: https://www.clavax.com/ 6. ThirdEye Data ThirdEye Data is a big data services company that delivers big data consulting, analytics and engineering while using artificial intelligence, deep learning, and machine learning. ThirdEye Data has served such companies as Nokia, Ford, Amgen, GettyImages, Actian, and others. The company focuses on Finance, Government, IT, Retail, and Manufacturing. Santa Clara, CA Headquarters: Microsoft, Nokia, Ford, Amgen, GettyImages, Actian, Google Core clients: $50 — $99 Pay per hour: 50–249 Number of employees: Website: https://thirdeyedata.io/ 7. Signity Solutions Established in 2009, Signity Solutions is a company that delivers custom software development and big data solutions. The company helps with big data analysis and brings data solutions empowered by machine learning. Signity works for the Commerce, Education, Media, and other industries. Rajiv Gandhi IT Park, India Headquarters: Samsung, Grupio, The Project Factory Core clients: $25 — $49 Pay per hour: 50–249 Number of employees: Website: https://www.signitysolutions.com 8. Caserta Founded in 2001, Caserta is a NY-based IT company that delivers a wide range of data-related services. The company offers big data consulting and solution architecture/development. Caserta also provides data visualization, data warehousing services, and more. The company has worked with such clients as New York Times, Capital One, Equinox, and HBO. New York, NY Headquarters: New York Times, Capital One, Equinox, HBO Core clients: Not Revealed Pay per hour: 10–49 Number of employees: Website: https://caserta.com/ 9. RapidOps, Inc RapidOps is a custom software development company based in the USA. The company’s expertise includes big data analytics, consulting, and visualization. The company’s team of 60+ professional developers delivers its services to the Manufacturing, Retail, and IT industries. Charlotte, NC Headquarters: YoDish, Zurn Industries, DassaultSystemes Core clients: $25 — $49 Pay per hour: 50–249 Number of employees: Website: https://www.rapidops.com/ 10. Aegis Soft Tech Since 2003, Aegis Soft Tech has been providing software development to Healthcare, Finance, Retail, Manufacturing, and other industries. Aegis provides big data consultation, management, and analysis. Among that, Aegis helps with data migration processes. The company’s pool of developers expertizes in Hadoop, Cloudera, and Hortonworks platforms. Rajkot, India Headquarters: Sunzinet, Zydus, PointRF, Efacec Core clients: < $25 Pay per hour: 50–249 Number of employees: Website: https://www.aegissofttech.com/ 11. GetInData GetInData is a Poland-based company that unites more than 30 professionals. The company has carried out more than 30 projects. GetInData delivers the following big data services: big data infrastructure development, analytics, consulting, engineering, and maintenance. GetInData delivered its services to such companies as Spotify, Allegro, Agora, and Freshmail. Warsaw, Poland Headquarters: Spotify, Allegro, Agora, Freshmail, Truecaller, GoEuro Core clients: $50 — $99 Pay per hour: 10–49 Number of employees: Website: https://getindata.com/ 12. Kavi Global Established in 2008, Kavi Global is an analytical company that delivers a whole spectrum of big data services. They include consulting, big data solution architecture, and platform implementation. The company also offers data mining and data management services. Kavi Global’s industrial expertise includes Transportation, Manufacturing, Education, and Healthcare. Barrington, IL Headquarters: General Electric, TEAMHealth, Thomson Reuters, Ryder Core clients: $150 — $199 Pay per hour: 10–49 Number of employees: Website: https://www.kaviglobal.com/ 13. Indium Software Since 1999, Indium Software has been providing technological solutions for 19 various industries. The company has worked with 750 clients from all over the world. Employing such technologies as Hadoop, Spark, and Apache Kafka, Indium Software delivers big data architecture and consulting services. Princeton, NJ Headquarters: Not Revealed Core clients: < $25 Pay per hour: 250–999 Number of employees: Website: https://www.indiumsoftware.com/ 14. Fayrix Israel-based Fayrix is a custom software development company and big data services provider with 1530 IT experts on board. The company deals with various projects, from initial data analysis and consulting to big data infrastructure design, big data software development, and its implementation into a client’s existing technological landscape. Pitu’akh, Israel Headquarters: Biosence Webster, SCR, S7 Airlines Core clients: < $25 Pay per hour: 1000–9999 Number of employees: Website: https://fayrix.com/ 15. Denologix Established in 2002, Denologix is an analytics and information management company providing a wide range of big data services, including data analysis, consulting, and big data solution implementation and support. The company deploys a large number of technological tools for big data, including Hortonworks Data Platform, Cloudera СВР Hadoop Distribution, and more. Among the company’s clients are Adidas, Allstream, Aviva, Canada Post, and others. Toronto, Canada Headquarters: Adidas, Allstream, Aviva, Canada Post, Bell, Brookfield Core clients: $150 — $199 Pay per hour: 50–249 Number of employees: Website: http://www.denologix.com/ 16. Data Reply Established in 2015, Data Reply is a big data and analytics consultancy based in London, UK. Besides big data consulting and training, the company offers data engineering and data science services. The company focuses on such industries as Retail, Media, Telecommunications, and Commerce. London, United Kingdom Headquarters: Not revealed Core clients: $100 — $149 Pay per hour: 10–49 Number of employees: Website: https://www.datareply.co.uk/ 17. Dexlock Established in 2012, Dexlock is an Indian company that delivers big data services. Among them are large data store development and deep learning/machine learning implementation. The company’s technological expertise includes Hadoop, Apache Spark, Apache Kafka, and more. Dexlock works with the Healthcare, Retail, Insurance, Energy and other industries. Kochi, India Headquarters: FindCharisma, Splash of Colors, StatFuel, MyPeepz Core clients: < $25 Pay per hour: 50–249 Number of employees: Website: http://www.dexlock.com/ 18. 47Billion 47Billion is a bespoke software development company that provides mobile and web development and UI/UX design. Besides these services, the company delivers end-to-end big data analysis, consulting, and engineering. The company’s industrial focus includes IT, Gaming, and Consumer Services. 47Billion’s clients include Cisco, Nazara, and Apsalar. Bangalore, India Headquarters: Cisco, Nazara, Apsalar Core clients: : $25 — $49 Pay per hour 50–249 Number of employees: Website: https://47billion.com/ 19. 51zero 51zero is a UK-based company that specializes in big data services providing. The company delivers big data consulting and solutions development and focuses on the financial service domain. The company’s pool of technological experts deploys such big data technologies as Spark, Apache Kafka, Docker, and others. London, United Kingdom Headquarters: HSBC, The Royal Bank of Scotland, Barclays Core clients: $200 — $300 Pay per hour: 2–9 Number of employees: Website: https://www.51zero.com/ 20. SCI-BI SCI-BI is a company based in India. With more than seven years of experience, the company specializes in big data consulting and engineering. The company also helps with data integration strategy. SCI-BI’s big data technological expertise includes Cloudera and Hortonworks. The company focuses explicitly on Logistics and Retail. Chennai, India Headquarters: Executive Planning Services, USA SaasByDesign Inc, USA Caravel Logistics Core clients: < $25 Pay per hour: 10–49 Number of employees: Website: http://sci-bi.com/

The Top Big Data Consulting Firms

Healthcare Innovation in 2018

245 Tools & Resources for Travelers and Digital Nomads

What is Brand Management Software? Brand Management tools are helpful to businesses in order to manage their brand globally and locally and you can easily ensure yourself that your branding is in your control. Brand owners or managers can easily manage the complete brand messaging of their business with the help of brand management software. A business owner or brand manager can easily create an effective discipline for those who represent their brand; agencies, vendor partners and other marketing sources represent the brand in the market so the brand manager can easily create the effective roadmap for them and it really increases the brand reputation in the market. The user can easily get a centralized database where he/she can create, manage, store, share and distribute the brand assets and relevant files. This thing really increases brand consistency in all different stages. If you need to create new marketing techniques and strategies for your brand and you need the best process to regulate it, brand management software is the best solution for you where you can easily manage workflow process. You can also approve new marketing techniques and other activities to boost your brand power and awareness. Let’s take a look, here you can find the leading brand management software for your business. 1. Brand24 Discover what people are saying online about your brand — as it unfolds in real time. Brand24 gives you instant access to mentions about your brand across the web, from social networks to influential publishers. React and engage quickly to real-time comments made about your brand. Follow-up on positive comments from your brand ambassadors, or respond to a dissatisfied customer before the story gets ahead of you. 2. Widen Collective Widen Collective is a cloud-based digital asset management solution that allows users to find, share, upload, download and analyze visual content anywhere, on any device. The product is suitable for marketers, creative teams, IT managers and software architects of SMBs and enterprises. It is designed to cater to the needs of various industries that include apparel, construction, consumer goods, hospitality, healthcare, engineering, medical devices, travel and retail. 3. Grade.us If you’ve got your pulse on growing tends in digital marketing, you know the importance of online reviews. Not only are they an increasingly valuable ranking factor for search engines, they are becoming an inevitable step for your customers as they look to make a purchase decision themselves! In other words, they help you get more customers because of visibility and influence. 4. Justuno Justuno is the most innovative and robust conversion optimization suite for sales conversions, email list growth, and onsite promotions using real-time user data. We deliver powerful results for agencies, grow audiences for brands, and increase conversions for merchants. 5. Cision Cision covers all aspects of your communication needs, helping you reach, target and engage your audience. The Cision Communications Cloud® features a suite of metrics and automated analyses that translate your activities and earned media coverage into valuable insights and most importantly verifiable return-on-investment. 6. Reputology Reputology is the review monitoring and management solution designed for organizations that operate multiple physical locations. We help restaurants, gyms, hotels, auto dealers, hospitals, marketing agencies, retail stores, banks & more. 7. Percolate Percolate offers solutions to introduce visibility into the marketing process, improve coordination of work, and effectively build marketing campaigns and content. The world’s largest brands — including Mastercard, Cisco, Cigna, Electronic Arts, and Bosch — use Percolate to create a coordinated customer experience, reduce production costs, and understand marketing impact. 8. Bynder Bynder’s success is not only down to the quality of our work and outstanding customer care, but it’s driven by the attitude and approach of our global team. We believe focused and passionate teamwork should influence everything we do, and our desire to produce exceptional solutions runs deep. We are united with you in the drive to get the best from your project; we listen hard to understand your business and your needs, and we’re 100% committed to making all our features the best they can possibly be. 9. Prisync Prisync is a competitor price tracking & dynamic pricing software for all sizes of e-commerce companies from all around the world. It tracks online prices and stock availabilities and lets e-commerce companies define smart pricing rules depending on their unit product costs and competitor prices to apply dynamic pricing on their webshop to increase sales and profit margins. Prisync is already helping various sizes of e-commerce companies from more than 40 countries. 10. Asset Bank The first version of Asset Bank was established in 2005. We were inspired to enter the world of Digital Asset Management (DAM) after developing a bespoke image management system for our client GKN Driveline. As part of the development process we realised there were quite a few companies already offering DAM systems, but at an exorbitant cost. Our vision was to offer quality DAM software at a great price. 11. Frontify Frontify is the SaaS-based brand management platform for the digital age used by 2000 companies worldwide including Vodafone, Lufthansa, IBM, Nasdaq, and Bosch. With our cloud-based software, brand managers can create online Style Guides, manage their brand assets, create UI Pattern Libraries, Brand Portals and work collaboratively in creative projects. Strengthen your brand consistency over all channels using Frontify. Ideally suited for Enterprise level companies. 12. NiceJob NiceJob is a reputation marketing solution to aid businesses with gathering more and better quality reviews, marketing across social platforms, and improving SEO ranking. With the aim of helping businesses of any type or size to improve their online reputation, NiceJob provides tools such as review invitations, automated email and SMS follow-ups, review funnels, stories, SEO management, and more. 13. Third Light Third Light is a fast, feature-rich, digital asset management tool. Organise, tag, find, share, edit and download all of your digital files quickly, easily and securely. Online chat and share supports centralized, collaborative working with internal teams and external partners. Convert formats, trace copyright, protect and distribute files. Cloud-based or licensed solutions with free, seamless updates and API available. Great price, offers unlimited users and supports industry standards. 14. Outbrain Outbrain is a premium discovery platform that helps connect marketers to their target audience through personalized recommendations on the world’s leading publishers. We use proprietary interest and behavioral data to capture audience’s attention all the way to inspiring their next action. 15. CleanPix CleanPix is a web-based digital, brand, and media asset management solution which automatically processes files into multiple sizes and formats, enables users to organize assets into public and private galleries, and allows distribution of assets through public portals, access-controlled private galleries, and directly by email. The Core Elements of Brand Management Software Brand Management software is specially designed for all sizes business where a brand manager can easily organize and advertise their products and services to their most potential customers in the most effective manner. We can also understand this software as marketing category software. Here you can find the most essential elements of brand asset management software: Effective Marketing Campaigns: As I told you earlier that it manages to brand globally and locally however it also works to maintain brand loyalty internally and externally. Marketing is one and the only option to reach your audience so you can easily reach to your audience and create the effective brand awareness with the help of brand management tools. You can create a robust and error-less marketing campaign with these branding tools. Project Workflows: You can easily manage project workflows with brand management software; it usually takes care of your all work steps; you can also track the workflow with the same software. You can create new marketing strategies and regulate it to across all channels. You can create the effective work cycle at all channels of business and it will help you to create the best workflows. You can easily increase the team collaboration level with brand management. Design Assets: Design defines the business. It attracts the customer and increases brand value so never compromise with design assets. You can easily develop and market impactful designs which one shows the higher relevancy of your business. You can create designs in any format and share directly to social media with the help of brand management tools. You can even restore the asset and recreate the design with your team. Social Media Social media is the most powerful way for you if you really want to reach to your audience. You can easily create relevant and impressive designs with this type of tool and directly share to social media. You can easily integrate any of social media platforms into your brand management software. You can easily double your social media effort through brand management software. Common Functionality of Brand Management: Here, I am going to show you some most common functionality of brand management software. Social Media Integration: You can easily integrate your brand management software with your business social media handles like Facebook, Twitter, Google+, Linkedin, Tumblr, Reddit and much more. It really creates the easiest way to do social media for your business. Real-time analytics You can easily access real-time analytics across the complete brand management process. You can easily get an idea about critical insights through this software. User-defined permissions You can easily provide and denied access to the particular user. You can deliver and share your business assets to the right person with this feature. Search functionality It’s the best option to locate the right asset in right time. You can search and get needed assets anytime with this feature. Multi-format output You can easily create your content in multi-format like you can create content in PDF, HTML, and other formats. Fulfillment distribution You can easily distribute the relevant asset to the right users and groups in seconds even you can control your distribution among users. Asset management The User can easily create, store and manage all the marketing and design assets even he/she can share with others within certainly located online portal or database. Artwork automation It allows the user to streamline design processes and this thing really improves the brand accuracy and productivity level of the particular user. Approval Process Control The user can easily review the social media posts and marketing strategies and after they can manually or automatically approve it. Alert Feature The user can easily send alerts to a particular user or group regarding content reviews and approval stages. It will also help to generate custom email alerts to the selected groups. What things buyers should keep in mind? As I told you that it works for all size businesses as per their needs so before choosing the best software, you must have to understand your needs. If you are the small business owner, cloud-based brand management software is the best option for you because of its affordable and cost-effective options to manage your branding. 1. For larger and big enterprises, there is another category of corporate brand management software where you can find the on-premise solution for your brand. You can easily organize and control in-house branding activities with the help of this software. 2. You must also understand the third-party integration features of software because of it will make your way easier with proper integration. You can integrate your social media accounts, email accounts, online store, your website and much more thing to your brand management software. 3. Why You Need Brand Management Software For Your Business? You can do many things manually to increase your business but it only offers limited space. All the successful company works on big data and analytics and if you want to work with big data, brand management tools are the best option to make your system automated. You can easily discover more useful insights and understand your customer needs very well. You can even enter to the new market with the help of valuable and exact data. Mobile offers major traffic to brands than desktop, laptop, and other devices. Most returning users and new users (approx 51%) interact with mobile to the brand so the business turns their approach to mobile devices. You can easily integrate your brand software to your mobile app and create effective marketing and branding strategies for your business.

List of Top 10+ Brand Management Software With Ultimate Buyer Guide

Originally published on Scrappy & Strange, the Founder Collective Newsletter — Sign up here ! David Frankel relayed this quote from , which was delivered at annual gathering, as a reminder that capital is a constraint — though probably not the biggest constraint facing most startups. Hustle tends to be in much shorter supply. Here are a few inspiring stories of entrepreneurs illustrating the power of hustle over capital: Regie Gibson Cue Ball Capital’s Instead of investing in costly ad campaigns in the company’s cash-strapped early days, , with ultra-specific names like “RacksAndStands.com” to game Google’s search algorithm, paving the way for what has become a $4B business. Wayfair: Wayfair built 270 highly targeted websites a $2B stock photo service, was started by a hacker/amateur photographer who personally built the site and launched it with 100,000 images from his photo collection. ShutterStock , Hustle doesn’t stop once you have raised venture capital. Langley Steinert raised $5M to get CarGurus started, but instead of seeking VC to hire even more engineers, he used the revenue he was generating to hire — now the company is cruising towards an IPO. CarGurus: hundreds of sales reps to generate even more revenue Meet The Efficient Entrepreneur Who Replaced RadioShack Limor Fried, founder/CEO of AdaFruit. Photo: . Make: Magazine started out of her dorm room at MIT and has bootstrapped her ecommerce/electronics startup (now based in NYC) to $45M in revenue with 105 FTEs, and a product line which features 400+ hardware products and kits. Read this profile to learn more about the company, including how they’ve acquired 20M+ YouTube views, 100K Twitter followers, and 2M monthly website visits. Limor Fried AdaFruit Make: Magazine Unity Raised No VC for its First Four Years, Gets $400M in One Round , makers of the game engine that powers Pokemon Go and just raised . Even more interesting is the of how the founders hustled during the company’s first four years, building out , before raising . Unity 34% of top mobile games $400M from PE giant Silver Lake history mobile-friendly 3-D rendering tech forgoing salary, and even waiting tables to pay the bills venture capital Should Startups Care About Profitability? makes a compelling argument that profitability is overrated for companies that are well suited to venture capital. It’s a nuanced argument, but one that almost any entrepreneur would benefit from reading. Here’s an excerpt: Mark Suster If the market they are targeting is very large and fast growing, then the venture-backed businesses often make it harder for the non-venture-backed businesses to compete in the long-term. If the markets aren’t large then the company who managed its costs may be able to get a modest exit at a fair price and make the team wealthy precisely because they didn’t take on venture capital. The VC-backed businesses sometimes “blow up.” As is often the case — there are no obvious or right answers. responded with a post about the beneficial effects enjoyed by his portfolio companies that have prioritized profitability. shared his experience from the dot-com era, “I learned that the markets reward growth until they don’t. Then they reward profitability.” Also, be sure to read to the discussion. Fred Wilson Brad Feld Roger Ehrenberg’s thoughtful contribution Weirdness from Around the Web Filmmaker replaces missing eyeball with a video camera. [ ] Motherboard “During a recent type-in at Albuquerque soul food restaurant Nexus Brewery, around three dozen fans took turns clicking the keys of an Italian-made 1964 Olivetti Lettera 32 Typewriter.” [ ] MySanAntonio British researchers scan 6,000 faces to create database for plastic surgeons to use for reconstructive surgery [ ] 3Ders The first blue pigment created in over 200 years to be used in a crayon [ ] Colossal Founder Collective in the News Deconstructing the “Near” Perfect Deal — Our Investment in Julia Computing By D avid Frankel , Partner An excerpt: I’m fortunate to have been involved with some amazing entrepreneurs and startups over the years, but each of those deals had some flaw when I invested. These obviously weren’t disqualifying; it’s just the reality that high-risk seed-stage venture investing is predicated on … risk. That’s why I’m both excited and somewhat concerned that I’ve recently had the opportunity to invest in a deal that is as close to the platonic ideal that I’m ever likely to see. Want to know what makes a deal “near” perfect? Read More! What Startups & VCs Can Learn From Indiana Jones By Micah Rosenbloom , Partner An excerpt: It might be a useful exercise for founders to think of themselves as auteurs rather than entrepreneurs. Filmmakers and founders are both creative professionals, but directors have to contend with a world where capital is a fixed sum, not a seemingly endless stream. Directors must have a vision for their product, a rough understanding of what it would cost to realize that vision, and then execute within those constraints. Contrast that to startup founders who always have the possibility of raising another round of funding. Strict limitations for filmmakers have led to a slew of indie projects, made on shoe-string budgets by unknown creators, that go on to rival star-studded Hollywood blockbusters. For instance: These are career-making projects that put stars, and catchphrases, on the map. All were made — at most — with a series A level of funding. What makes these low-budget films so successful? There seem to be a few common themes: Want to know what Napoleon Dynamite can teach your startup? . Read the rest here There’s No Shame in a $100M Startup By E ric Paley , Partner & Joe Flaherty , C&C We’ve expanded our list of and found that billion dollar outcomes remain in short supply. entrepreneurs who have become VCs Portfolio Updates CEO Zach Bookman appeared on CNBC to discuss how tech companies can help streamline and strengthen government services OpenGov [CNBC] raised $50M from Cigna to fight diabetes Omada Health [Omada] announced a new package design and PharmacyOS PillPack [Wired] released data demonstrating that meal delivery apps don’t cannibalize restaurant sales. S ense360 [CNBC] breaks down the new changes to Apple’s podcast ecosystem Bumpers [Bumpers] is one of the Wall Street Journal’s “25 Tech Companies to Watch” Andela [WSJ] Job of the Week 3-D Printing market leader is looking to hire 100+ people in the next year, but first, they need to find an . Do you know how to seek out the best software engineers? The techniques to attract top notch talent in engineering, design, marketing, and sales? If so, Formlabs is building something amazing — be a part of it. Formlabs ace recruiter

“Hustling is the art of practicing capitalism without capital”

How to Use Tuples in TypeScript

From a young age, Vishwanadham Mandala had a keen interest in technology, setting the stage for a distinguished career in the IT sector. Growing up in India, he earned a BTech and MTech in Computer Science & Engineering, fueled by a passion to harness the power of data. His dedication to academia and a relentless quest for knowledge led him to Indiana University in the United States, where he completed an MS in Data Science. He continues to push the envelope by pursuing a PhD in AI & ML. Currently, Vishwanadham is a prominent leader in the field of data engineering. Known for his leadership mantra, "Consistent work, each day, every day," he blends strategic foresight with unwavering persistence. His expertise as an AI & ML leader in critical automotive applications has advanced automotive manufacturing processes and safety standards, enhancing public safety. As a mentor and speaker, Vishwanadham is committed to developing the next generation of technology professionals, ensuring that his legacy of excellence leaves a lasting impact on the industry. Additionally, he has been a member of the Forbes Technology Council since May 2024. Path to Big Data Fascination with the transformative power of data ignited Vishwanadham’s pursuit of a career in Big Data and data engineering. His journey through critical roles at IBM, Oracle, Sierra-Cedar, and Ciena allowed him to work globally, gaining diverse experiences and honing his skills. Reflecting on his career progression, Vishwanadham notes how these experiences shaped his expertise and paved the way to his current role at Cummins, where he spearheads significant projects and integrates AI/ML technologies to drive the company's future direction. Vishwanadham's academic background has been instrumental in his professional success. With degrees from prestigious institutions in India and the United States, he notes, "My diverse educational background gave me a solid foundation in technical knowledge and critical thinking skills." This exposure to different academic environments and perspectives has enriched his understanding of global technological trends and best practices, enabling him to excel in his career and make meaningful contributions to the IT industry. "The exposure to different academic environments and perspectives enriched my understanding of global technological trends and best practices," he explains. This comprehensive background not only allowed him to grow professionally but also equipped him with the skills necessary to drive innovation and efficiency at Cummins, solidifying his role as a leader in the integration of AI and ML technologies. Game-changing Projects Across Industries Vishwanadham led a groundbreaking project that significantly impacted Ciena’s strategic direction in the data space. His innovative approach and strategic leadership were instrumental in shaping Ciena's technology roadmap, highlighting his ability to drive impactful change and foster innovation within the organization. His work showcased his expertise in data engineering and his vision for future technological advancements. Vishwanadham's experience spans various industries, including manufacturing and healthcare, where he has successfully integrated AI and ML solutions to address unique challenges. He explains that his approach involves focusing on both fundamental and superficial facets of data. By adapting to each sector's unique challenges and nuances, he tailors his solutions to best match the environment. This method of integrating AI and ML solutions has enabled Vishwanadham to continuously learn and grow with each encounter. "This method has enriched my expertise, allowing me to tackle various challenges effectively and continuously learn and grow with each encounter," he notes. By leveraging his diverse experience and innovative approach, Vishwanadham has been able to make significant contributions to the fields of manufacturing and healthcare, demonstrating his versatility and proficiency in AI and data engineering. Innovative IoT Patents for Rural Healthcare One of Vishwanadham’s most notable patents involves a groundbreaking IoT solution for medical devices, specifically designed to enhance healthcare in rural areas. This innovative patent focuses on leveraging telemetry data processing to provide real-time health insights, which can significantly improve patient outcomes in remote locations. By utilizing AI and ML, the solution aims to deliver instantaneous and accurate health data, allowing for timely medical interventions. Vishwanadham explains that this technology "enables timely medical interventions and improves access to quality healthcare for rural populations." This IoT innovation has the potential to transform rural healthcare by bridging the gap between remote patients and advanced medical care. "By leveraging AI/ML, this innovation aims to offer instantaneous and accurate health data," says Vishwanadham, highlighting the critical role of advanced technology in improving healthcare accessibility and quality in underserved areas. This patent exemplifies his commitment to using technology to create impactful solutions that address real-world challenges. Leading and Mentoring in AI Vishwanadham emphasizes that leading successful AI and ML initiatives requires a strong foundation in data science, strategic vision, and a commitment to continuous learning. He believes that a disciplined work ethic, cooperative teamwork, and innovative thinking are essential qualities for driving progress in these fields. Additionally, "Effective communication and inspiring and guiding teams are also essential for driving innovation," Vishwanadham states, highlighting the multifaceted nature of leadership in AI and ML. In addition to his leadership skills, Vishwanadham is deeply committed to mentoring and teaching in the AI and ML fields. He actively shares his extensive knowledge and experience with aspiring professionals. "Mentoring helps individuals grow and contributes to the industry's evolution by fostering a culture of continuous learning and innovation," he explains. By guiding the next generation of AI and ML experts, Vishwanadham aims to create a robust talent pool capable of tackling future challenges and driving technological advancements. This dual focus on leadership and mentorship highlights Vishwanadham's holistic approach to advancing the AI and ML fields. He understands that the future of technology depends not only on current innovations but also on preparing the next generation to continue pushing the boundaries of what is possible. AI Innovations for Safety and Supply Chains Looking to the future, Vishwanadham aspires to leverage AI and ML to significantly advance citizen safety and improve supply chains by implementing cutting-edge technologies and innovative solutions. He is particularly excited about the potential of Edge AI/ML, which he describes as "a revolutionary approach to telemetry data processing that can provide instantaneous and accurate insights." One of his current projects focuses on integrating this technology into the automotive industry to enhance vehicle safety and protect citizens. "By staying ahead of the competition through constant learning and innovation, I aim to drive transformative change," he asserts. His commitment to using advanced AI and ML technologies to solve real-world problems highlights his vision for a safer and more efficient future, where technology plays a crucial role in protecting lives and optimizing operations. In a world becoming increasingly data-centric, Vishwanadham’s expertise serves as a beacon for those navigating the complexities of modern business landscapes. His journey reminds us that mastering data is not just about understanding technology but also about leveraging it to make remarkable contributions to society. As he continues to mentor the next generation of tech professionals, his influence extends beyond projects and patents, fostering a culture of continuous learning and innovation in the industry. From a young age, Vishwanadham Mandala had a keen interest in technology, setting the stage for a distinguished career in the IT sector. Growing up in India, he earned a BTech and MTech in Computer Science & Engineering, fueled by a passion to harness the power of data. His dedication to academia and a relentless quest for knowledge led him to Indiana University in the United States, where he completed an MS in Data Science. He continues to push the envelope by pursuing a PhD in AI & ML. Currently, Vishwanadham is a prominent leader in the field of data engineering. Known for his leadership mantra, "Consistent work, each day, every day," he blends strategic foresight with unwavering persistence. His expertise as an AI & ML leader in critical automotive applications has advanced automotive manufacturing processes and safety standards, enhancing public safety. As a mentor and speaker, Vishwanadham is committed to developing the next generation of technology professionals, ensuring that his legacy of excellence leaves a lasting impact on the industry. Additionally, he has been a member of the Forbes Technology Council since May 2024. Path to Big Data Fascination with the transformative power of data ignited Vishwanadham’s pursuit of a career in Big Data and data engineering. His journey through critical roles at IBM, Oracle, Sierra-Cedar, and Ciena allowed him to work globally, gaining diverse experiences and honing his skills. Reflecting on his career progression, Vishwanadham notes how these experiences shaped his expertise and paved the way to his current role at Cummins, where he spearheads significant projects and integrates AI/ML technologies to drive the company's future direction. Vishwanadham's academic background has been instrumental in his professional success. With degrees from prestigious institutions in India and the United States, he notes, "My diverse educational background gave me a solid foundation in technical knowledge and critical thinking skills." This exposure to different academic environments and perspectives has enriched his understanding of global technological trends and best practices, enabling him to excel in his career and make meaningful contributions to the IT industry. "The exposure to different academic environments and perspectives enriched my understanding of global technological trends and best practices," he explains. This comprehensive background not only allowed him to grow professionally but also equipped him with the skills necessary to drive innovation and efficiency at Cummins, solidifying his role as a leader in the integration of AI and ML technologies. Game-changing Projects Across Industries Vishwanadham led a groundbreaking project that significantly impacted Ciena’s strategic direction in the data space. His innovative approach and strategic leadership were instrumental in shaping Ciena's technology roadmap, highlighting his ability to drive impactful change and foster innovation within the organization. His work showcased his expertise in data engineering and his vision for future technological advancements. Vishwanadham's experience spans various industries, including manufacturing and healthcare, where he has successfully integrated AI and ML solutions to address unique challenges. He explains that his approach involves focusing on both fundamental and superficial facets of data. By adapting to each sector's unique challenges and nuances, he tailors his solutions to best match the environment. This method of integrating AI and ML solutions has enabled Vishwanadham to continuously learn and grow with each encounter. "This method has enriched my expertise, allowing me to tackle various challenges effectively and continuously learn and grow with each encounter," he notes. By leveraging his diverse experience and innovative approach, Vishwanadham has been able to make significant contributions to the fields of manufacturing and healthcare, demonstrating his versatility and proficiency in AI and data engineering. Innovative IoT Patents for Rural Healthcare One of Vishwanadham’s most notable patents involves a groundbreaking IoT solution for medical devices, specifically designed to enhance healthcare in rural areas. This innovative patent focuses on leveraging telemetry data processing to provide real-time health insights, which can significantly improve patient outcomes in remote locations. By utilizing AI and ML, the solution aims to deliver instantaneous and accurate health data, allowing for timely medical interventions. Vishwanadham explains that this technology "enables timely medical interventions and improves access to quality healthcare for rural populations." This IoT innovation has the potential to transform rural healthcare by bridging the gap between remote patients and advanced medical care. "By leveraging AI/ML, this innovation aims to offer instantaneous and accurate health data," says Vishwanadham, highlighting the critical role of advanced technology in improving healthcare accessibility and quality in underserved areas. This patent exemplifies his commitment to using technology to create impactful solutions that address real-world challenges. Leading and Mentoring in AI Vishwanadham emphasizes that leading successful AI and ML initiatives requires a strong foundation in data science, strategic vision, and a commitment to continuous learning. He believes that a disciplined work ethic, cooperative teamwork, and innovative thinking are essential qualities for driving progress in these fields. Additionally, "Effective communication and inspiring and guiding teams are also essential for driving innovation," Vishwanadham states, highlighting the multifaceted nature of leadership in AI and ML. In addition to his leadership skills, Vishwanadham is deeply committed to mentoring and teaching in the AI and ML fields. He actively shares his extensive knowledge and experience with aspiring professionals. "Mentoring helps individuals grow and contributes to the industry's evolution by fostering a culture of continuous learning and innovation," he explains. By guiding the next generation of AI and ML experts, Vishwanadham aims to create a robust talent pool capable of tackling future challenges and driving technological advancements. This dual focus on leadership and mentorship highlights Vishwanadham's holistic approach to advancing the AI and ML fields. He understands that the future of technology depends not only on current innovations but also on preparing the next generation to continue pushing the boundaries of what is possible. AI Innovations for Safety and Supply Chains Looking to the future, Vishwanadham aspires to leverage AI and ML to significantly advance citizen safety and improve supply chains by implementing cutting-edge technologies and innovative solutions. He is particularly excited about the potential of Edge AI/ML, which he describes as "a revolutionary approach to telemetry data processing that can provide instantaneous and accurate insights." One of his current projects focuses on integrating this technology into the automotive industry to enhance vehicle safety and protect citizens. "By staying ahead of the competition through constant learning and innovation, I aim to drive transformative change," he asserts. His commitment to using advanced AI and ML technologies to solve real-world problems highlights his vision for a safer and more efficient future, where technology plays a crucial role in protecting lives and optimizing operations. In a world becoming increasingly data-centric, Vishwanadham’s expertise serves as a beacon for those navigating the complexities of modern business landscapes. His journey reminds us that mastering data is not just about understanding technology but also about leveraging it to make remarkable contributions to society. As he continues to mentor the next generation of tech professionals, his influence extends beyond projects and patents, fostering a culture of continuous learning and innovation in the industry.

Mastering Data in the Modern Age with Vishwanadham Mandala

Авалоданне дадзенымі ў сучасную эпоху з Vishwanadham Mandala

বিশ্বধাম মন্ডালার সাথে আধুনিক যুগে ডেটা আয়ত্ত করা

Dominar les dades a l'Edat Moderna amb Vishwanadham Mandala

Mestring af data i den moderne tidsalder med Vishwanadham Mandala

Dominando los datos en la era moderna con Vishwanadham Mandala

שליטה בנתונים בעידן המודרני עם Vishwanadham Mandala

Padroneggiare i dati nell'era moderna con Vishwanadham Mandala

ヴィシュワナダム・マンダラによる現代のデータ活用

Vishwanadham Mandala로 현대의 데이터 마스터링

Menguasai Data di Zaman Moden dengan Vishwanadham Mandala

Opanowanie danych w erze nowoczesności z Vishwanadham Mandala

Освоение данных в современную эпоху с помощью Vishwanadham Mandala

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 6. Conclusion Arabic and Hebrew, both Semitic languages, display inherent structural resemblances and possess shared cognates. In an endeavor to allow a bilingual language model to recognize these cognates, we introduced a novel language model tailored for Arabic and Hebrew, wherein the Arabic text is transliterated into the Hebrew script prior to both pre-training and fine-tuning. We contrast our model by training another language model on the identical dataset but without the transliteration preprocessing step, in order to assess the impact of transliteration. We fine-tuned our model for the machine translation task, yielding promising outcomes. These results suggest that the transliteration step offers tangible benefits to the translation process. Comparing to the translation combination involving other language models, we see comparable results; this is encouraging given that the training data we utilized for pre-training the model is approximately 60% smaller than theirs. As a future avenue of research, we intend to train the model on an expanded dataset and explore scaling its architecture. In this study, our emphasis was on a transformer encoder. We are keen to investigate the effects of implementing transliteration within a decoder architecture, once such a model becomes available for Hebrew. Limitations The transliteration algorithm from Arabic to Hebrew is based a simple deterministic lookup table. However, sometimes the transliteration is not that straight forward, and this simple algorithm generates some odd rendering, which we would like to fix. For example, our algorithm does not place a final-form letter at the end of the Arabic word in Hebrew. Another challenge with transliteration into Hebrew is that for some words a Hebrew writer may choose to omit long vowel characters and the readers will still be able to understand the word. This phenomenon is referred to as “Ktiv Hasser” in Hebrew. Yet, there exists a preference for certain word representations over others. This inconsistency makes it more challenging for aligning the transliterated Arabic words to their cognates in the way they are written. Our transliteration algorithm always renders the Arabic word following the Arabic letters, which may be different than how this word is typically written in Hebrew. Another limitation is the relatively small size of the dataset which we used for pre-training the language model, comparing to other existing language models of the same architecture size. 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 6. Conclusion Arabic and Hebrew, both Semitic languages, display inherent structural resemblances and possess shared cognates. In an endeavor to allow a bilingual language model to recognize these cognates, we introduced a novel language model tailored for Arabic and Hebrew, wherein the Arabic text is transliterated into the Hebrew script prior to both pre-training and fine-tuning. We contrast our model by training another language model on the identical dataset but without the transliteration preprocessing step, in order to assess the impact of transliteration. We fine-tuned our model for the machine translation task, yielding promising outcomes. These results suggest that the transliteration step offers tangible benefits to the translation process. Comparing to the translation combination involving other language models, we see comparable results; this is encouraging given that the training data we utilized for pre-training the model is approximately 60% smaller than theirs. As a future avenue of research, we intend to train the model on an expanded dataset and explore scaling its architecture. In this study, our emphasis was on a transformer encoder. We are keen to investigate the effects of implementing transliteration within a decoder architecture, once such a model becomes available for Hebrew. Limitations The transliteration algorithm from Arabic to Hebrew is based a simple deterministic lookup table. However, sometimes the transliteration is not that straight forward, and this simple algorithm generates some odd rendering, which we would like to fix. For example, our algorithm does not place a final-form letter at the end of the Arabic word in Hebrew. Another challenge with transliteration into Hebrew is that for some words a Hebrew writer may choose to omit long vowel characters and the readers will still be able to understand the word. This phenomenon is referred to as “Ktiv Hasser” in Hebrew. Yet, there exists a preference for certain word representations over others. This inconsistency makes it more challenging for aligning the transliterated Arabic words to their cognates in the way they are written. Our transliteration algorithm always renders the Arabic word following the Arabic letters, which may be different than how this word is typically written in Hebrew. Another limitation is the relatively small size of the dataset which we used for pre-training the language model, comparing to other existing language models of the same architecture size. 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

Reflections on HeArBERT: Transliteration Benefits, Research Horizons, and Limitations

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 3. Methodology We begin by pre-training a new language model using texts written in both Arabic and Hebrew. This model, named HeArBERT, is subsequently finetuned to enhance performance in machine translation between Arabic and Hebrew. For pre-training, we utilize the de-duplicated Arabic (∼3B words) and Hebrew (∼1B words) versions of the OSCAR dataset (Ortiz Suárez et al., 2020). In this work, we aim to measure the impact of normalizing all texts to a shared script, so that cognates can be unified under the same token representation. Therefore, we transliterate the Arabic texts into the Hebrew script as a preprocessing step for both training and testing. Our transliteration procedure is designed following most of the guidelines published by The Academy of the Hebrew Language who has defined a Hebrew mapping for every Arabic letter[4], and the mapping provided in (Terner et al., 2020). Only Arabic letters are converted to their Hebrew equivalents, while non-Arabic characters remain unchanged. Our implementation is based on a simple lookup table, executed letter by letter, which is composed of the two mappings mentioned above, as shown in Appendix A. For evaluation, we independently train the model twice: once with the transliteration step and once without. We subsequently compare the performance of these two versions when fine-tuned on a downstream machine translation test set. Our model is based on the original BERT-base architecture. We train a WordPiece tokenizer with a vocabulary size of 30, 000, limiting its accepted alphabet size to 100. This approach encourages the learning of tokens common to both languages, allowing the tokenizer to focus on content rather than on special characters not inherent to either language. We choose to train only for the masked language model (MLM) task employed originally in BERT, ignoring the next-sentence-prediction component, as it has previously been proven less effective (Liu et al., 2019). Overall, we trained each model for the duration of 10 epochs, over the course of approximately 3 weeks, using 4 Nvidia RTX 3090 GPUs. Fine-tuning HeArBERT is done similar to finetuning the original BERT model, except for the addition of the transliteration step of Arabic letters that takes place prior to tokenization. In this preprocessing step, all non-Arabic letters remain intact, while Arabic letters are transliterated into their Hebrew equivalents, as described above. The preprocssing and pre-training process of HeArBERT is depicted in Figure 1 This paper is available on arxiv under CC BY 4.0 DEED license. [4] https://hebrew-academy. org.il/wp-content/uploads/ taatik-aravit-ivrit-1.pdf 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 3. Methodology We begin by pre-training a new language model using texts written in both Arabic and Hebrew. This model, named HeArBERT, is subsequently finetuned to enhance performance in machine translation between Arabic and Hebrew. For pre-training, we utilize the de-duplicated Arabic (∼3B words) and Hebrew (∼1B words) versions of the OSCAR dataset (Ortiz Suárez et al., 2020). In this work, we aim to measure the impact of normalizing all texts to a shared script, so that cognates can be unified under the same token representation. Therefore, we transliterate the Arabic texts into the Hebrew script as a preprocessing step for both training and testing. Our transliteration procedure is designed following most of the guidelines published by The Academy of the Hebrew Language who has defined a Hebrew mapping for every Arabic letter[4], and the mapping provided in (Terner et al., 2020). Only Arabic letters are converted to their Hebrew equivalents, while non-Arabic characters remain unchanged. Our implementation is based on a simple lookup table, executed letter by letter, which is composed of the two mappings mentioned above, as shown in Appendix A. For evaluation, we independently train the model twice: once with the transliteration step and once without. We subsequently compare the performance of these two versions when fine-tuned on a downstream machine translation test set. Our model is based on the original BERT-base architecture. We train a WordPiece tokenizer with a vocabulary size of 30, 000, limiting its accepted alphabet size to 100. This approach encourages the learning of tokens common to both languages, allowing the tokenizer to focus on content rather than on special characters not inherent to either language. We choose to train only for the masked language model (MLM) task employed originally in BERT, ignoring the next-sentence-prediction component, as it has previously been proven less effective (Liu et al., 2019). Overall, we trained each model for the duration of 10 epochs, over the course of approximately 3 weeks, using 4 Nvidia RTX 3090 GPUs. Fine-tuning HeArBERT is done similar to finetuning the original BERT model, except for the addition of the transliteration step of Arabic letters that takes place prior to tokenization. In this preprocessing step, all non-Arabic letters remain intact, while Arabic letters are transliterated into their Hebrew equivalents, as described above. The preprocssing and pre-training process of HeArBERT is depicted in Figure 1 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 [4] https://hebrew-academy. org.il/wp-content/uploads/ taatik-aravit-ivrit-1.pdf

Stories, News & Mentions about CIGNA

Cigna’s Stories on HackerNoon

Cigna's News & Mentions