Add a ranker powered by text matching models

jina-ai / jina

☁️ Build multimodal AI applications with cloud-native stack

https://docs.jina.ai

Apache License 2.0

20.98k stars 2.22k forks source link

Closed nan-wang closed 4 years ago

nan-wang commented 4 years ago

Describe the feature

Add a ranker powered by text matching models

Your proposal

the ranker takes texts from query chunks and match chunks and call the model to calculate the matching scores

Environment

Screenshots

avhirupc commented 4 years ago

Do you mean something like Deep Rank , which takes in query and document and outputs a relevance score for ranking?

If not, Isn't this already supported in one of the examples?

https://github.com/jina-ai/examples/tree/master/southpark-search where matching between query and document based distilbert embedding is already implemented & ranked based on cosine similarity

jina-bot commented 4 years ago

This issue is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 4 days

celis commented 4 years ago

Hi @nan-wang, can you provide more details about what you call text matching models?, I see already implemented some tf-idf and bm25 rankers

hanxiao commented 4 years ago

i remove this feature from 0.4 as