Spring is here, and we can’t get enough of the innovative NLP projects blooming all around us! Our March co:lab fridays meetup was hosted by Jay Alammar and Luis Serrano, who were joined by a fresh new lineup of outstanding and award winning community demos. Check out the recap below, or watch the full session.

This Month’s Hot Topic: NLP University

Luis shared some of the learning content that his team has been working on, including the latest module on transformer models. Currently a series of blog posts and videos, this content will eventually move to a dedicated place where people can easily browse topics of interest and enhance their understanding of large language models and NLP. See list of topics to date on our YouTube playlist or the Cohere blog.

March’s Coolest Demos: Duns n Dracs, DocuChat, Ask Quran, and AI Brand Intel

Our lineup this month featured four exciting demos that spanned very different topics, including gaming, productivity, religion, and social media monitoring. These makers also share their thinking around building products with NLP. Check them out!

Duns n Dracs by Erick Fonseca Perez — Eric presented his winning project from the Midudev Twitch hackathon in February, showcasing different capabilities of NLP. In the game, Cohere is the dungeon master and generates every storyline and entity using the `small` and `xlarge-nightly` models, giving the player unlimited possibilities. See the GitHub repo.

DocuChat by Amir Azizafshari — Amir demoed his semantic search tool for GitHub documentation that leverages Cohere’s embedding model and a vector database. It allows users to ask questions about a specific topic on GitHub, and the system will find the closest documentation that matches the query. See the GitHub repo.

Ask Quran by Team Menlo Park Lab — Misbah Syed presented their chatbot system that uses Cohere’s multilingual semantic search capabilities to search Quranic text for question answering. Their goal was to enable contextual search of information that was not possible with conventional keyword search. The project was a Cohere hackathon winner. See the GitHub repo.

AI Brand Intel, also by Team Menlo Park Lab — Misbah shared the team’s winning project from a recent multilingual semantic search hackathon event sponsored by Cohere and Qdrant. This app also uses Cohere’s multilingual embeddings model, as well as Cohere’s Embed, Summarize, and Classify endpoints, to enable businesses to monitor and analyze social media and news mentions, and respond to customers, in over 100 languages. See the GitHub repo.

Congrats to the Presenters

The presenting teams will all get a special badge on Cohere’s Discord, plus mention in our social channels and a one-on-one mentoring session with a Cohere co-founder. Congratulations all around!

Join Us Next Month

Co:lab friday community demos happen on the last Friday of every month. Be sure to register for our next event on May 26, 12 pm ET!

If you have a Cohere-powered demo you’d like to share at co:lab fridays, please fill out our application form.

P.S. Also, join the co:mmunity conversation on Discord.