In this video, we'll learn how to use rerankers, a new library by Jeremy Howard and his team at Answer AI. rerankers provides a uniform API over various re-ranking models, making it super easy to switch/change/test the different models. Re-ranking is a technique where you re-sort a set of documents against a provided query to make sure that the most relevant documents are at the top of the list. It's used to improve the retrieval quality in a Retrieval Augmented Generation pipelines.
➡️ Resources
🟡 Code - https://github.com/mneedham/LearnDataWithMark/blob/main/rerankers-playground/rerankers.ipynb
🟡 rerankers - https://github.com/answerdotai/rerankers
🟡 Blog - https://www.markhneedham.com/blog/2024/09/28/intro-to-rerankers/
➡️ Chapters
00:00 What is rerankers?
00:20 Why do we need reranking?
01:14 Initialising reranker
01:33 A hello world example
02:08 An Olympics opening ceremony example
02:35 Querying the data with vector search
03:48 Improving the results with rerankers