Hi all,

The last Research Showcase for 2024 will be live-streamed next Wednesday, December 11, at 9:00 AM PT / 17:00 UTC. Find your local time here. The theme for this showcase is AI for Wikipedia.

You are welcome to watch via the YouTube stream: https://youtube.com/live/_hk6KLD-0tg?feature=share. As usual, you can join the conversation in the YouTube chat as soon as the showcase goes live. 

This month's presentations: 

Language Models for Wikimedia: A Report from the Trenches and a Call to Arms

By Bob West, EPFL

I will showcase LLM-based solutions my lab has built for tackling problems on Wikipedia (e.g., fact extraction, generation of article descriptions and edit summaries), highlighting practical challenges linked to applying such solutions in vivo. I will conclude by sketching a vision for building a Wikimedia Foundation Foundation Model with the potential to assist humans in keeping Wikipedia great.

A Public AI Assistant for World Wide Knowledge

By Monica Lam, Stanford University

We are introducing an initiative to develop a public AI assistant to enhance access to worldwide knowledge. Our scalable, LLM-based framework adapts to any knowledge corpus, effectively handling both structured and unstructured data. Our pilot version supports users in drafting Wikipedia-style articles by researching the web, engages in fact-grounded conversations using information from Wikipedia in 25 languages, and writes SPARQL queries for Wikidata. Our assistant has been applied across diverse fields: aiding scientists with the Semantic Scholar corpus of academic papers, supporting historians in examining 18th-century newspapers, and assisting journalists in analyzing Federal Election Commission (FEC) campaign donation records. We are seeking collaboration to further expand its applications.

--

Kinneret Gordon Lead Research Community Officer Wikimedia Foundation