Pre-trained transformer models have significantly improved search results and yield much better results than traditional approach. With their good ability of text understanding, they can retrieve great results without relying on external ranking signals like PageRank or click-rates. This make these methods especially suitable for application where such information is missing.
In this talk, I will give an introduction into semantic search:
How to create the necessary representations
How to find relevant documents
How to optimize and adapt algorithms to your own use-case without the need of labeled data
Speaker: Nils Reimers
Slides: https://nils-reimers.de/talks/2022-08...