FACT360’s Post

Wave goodbye to the lady and gentleman behind the recorded message! welcome AI interaction! Real-time, intelligent responses that speak to you in your language. The future of customer service is here! 🚀👋 #AIRevolution #GoodbyeRecording #HelloInnovation”

I've been getting more and more interested in how #generativeAI can be used for humanitarian analytics and situation awareness. Check out this paper I published in the recent ISCRAM 2024 conference where you can see outcomes of my Fulbright Poland work on using #LLMs to understand situation awareness related to #Ukrainian #refugees in #Poland. #chatGPT #datascience #spatialdatascience Title, abstract, and link to download the paper below. ------- Forced Displacement Situation Awareness with Large Language Models: Ukrainian Refugees in Poland Case Study We evaluated the utility of the large language model (LLM) ChatGPT to develop situation awareness related to the forced displacement of Ukrainian refugees into Poland. Utilizing text messages derived from the Help for Ukrainians in Poland Telegram message group, we used ChatGPT to translate messages in multiple languages into English and identify message topics and themes. Topics and themes from beginning of the war in Ukraine in 2022 were analyzed and visualized using K-means clustering and word clouds. The language identification and translation capabilities of the LLM were evaluated by two human evaluators and measured using Kappa (0.86) and BLUE scores (0.46) with the LLM performing effectively. We conclude that the ability of LLM using carefully developed language prompts for large data volume analysis with no need for manual human analysis shows promise for humanitarian analytics focused on rapidly identifying potential key trends, needs, and locations of displaced people. https://lnkd.in/dqbt24t2

Does the media play a role in shaping people's attitudes and votes towards migrants? How can we identify their role in the data? With my co-authors Mamadou SACKO, Ahmed Tritah, and Mathieu Picault we retrace the change in newspapers' media discourse on migrants after the occurrence of terrorist attacks in France, back in the years 2015-2016, using Natural Language Processing methods. We show how salience of immigration in the French media have increased for over 7 weeks after the attacks. Importantly, within the first three weeks, when linked to Muslims or (in)security related issues, we document that salience has been multiplied by 5. Besides, the tone of reporting decreased dramatically although this effect was shorter-lived, lasting up to two or three weeks.  In terms of electoral impacts, we find that municipalities with higher exposure to negative media coverage saw a corresponding rise in votes for about 3% on average, in favor of the main anti-migrant party in France during the Élections Régionales, held three weeks after the Bataclan attack. In some localities, the contribution of the media was dramatically high reaching figures as high as 20%. Laboratoire d'Économie d'Orléans CEPII Université de Tours

🇷🇺 🇩🇪 🇫🇷 🇪🇸 🇧🇬 🇮🇩 🇹🇷 🇮🇹 🇳🇱 🇨🇳 Big news. Searchlight's new AI -powered language translation tool is now live in DarkIQ and Cerberus, providing accurate #darkweb translations to cybersecurity professionals and law enforcement. With this release, you can instantly translate the top 10 languages on the #darkweb, including Russian, Chinese, and French. Our translation tool has been trained on more than 167 million dark web data points and uses Neural Machine Translation (#NMT) to translate with the same level of accuracy as a human – even when it comes to Russian dark web slang – so you can always decipher what criminals are really saying. #AIAdvancements #darkwebinvestigtion #darkwebmonitoring

New research shows how the dialect that is used to issue prompts to large language models can result in both biased and discriminatory outputs. The researchers found the LLMs would attribute less prestigious forms of employment for speakers of African American English versus Standard American English. Moreover, “…when LLMs are asked to pass judgment on defendants who committed murder, they choose the death penalty more often when the defendants speak African American English rather than Standardized American English, again without being overtly told that they are African American.” Finally, the authors found that “harm reduction measures like human feedback training not only don't address dialect prejudice but may make things worse by teaching LLMs to conceal their underlying racist training data with positive comments when queried directly on race.” #aigovernance #aisafety #aipolicy

Introducing Dialectal Arabic Speech-to-Text Models! ✨ VoiceAI now supports regional Arabic dialects including Gulf, Egyptian, and Levantine. Increase the accuracy of your call transcriptions and speech analytics for Arabic speakers. Read all about it in our latest article, link in the comments. ⬇️ #VoiceAI #STT #Transcription #AI 

Welcome to #futuretools In this episode of #future tools we look at the capability of artificial intelligence to commoditize knowledge and skills in Africa. Africa has so many gaps in knowledge and skills acquisition. With the development and adoption of AI platforms many skills and knowledge may be commoditize becoming available cheaply and efficiently to many in the continent. What this means is creation of more opportunities for those who acquire knowledge in AI systems. For example with a smart phone and AI language translator and with no language translation knowledge, a young man in remote areas of Africa is able to create documentaries on his/her routine activities and translate this to as many languages as he/she may wish Wee! the future is here for our continent.

Just to share a little slide from the first morning conferences at #CIOL2024 on gender bias in AI. From Hungarian, a language that doesn't encode gender, into English, a language which, for better or for worse, does.  👫 #ai #translation #translator #CIOL #traduction #languageservices #HAITrans

Big news. Searchlight's new AI -powered language translation tool is now live in DarkIQ and Cerberus, providing accurate #darkweb translations to cybersecurity professionals and law enforcement. With this release, you can instantly translate the top 10 languages on the #darkweb, including Russian, Chinese, and French. Our translation tool has been trained on more than 167 million dark web data points and uses Neural Machine Translation (#NMT) to translate with the same level of accuracy as a human – even when it comes to Russian dark web slang – so you can always decipher what criminals are really saying. #AIAdvancements #darkwebinvestigtion #darkwebmonitoring

If you are attending #ACL2024 in Bangkok next week and are interested in LLMs and natural language explanations join our poster session on August 14 @ 10:30, Poster session 6, Convention Center A1 I will be there to discuss our (Marco Valentino, Andre Freitas) latest work on "Inference to the Best Explanation in Large Language Models"