Candidate finder new metrics, cleanup

Living-with-machines / DeezyMatch

A Flexible Deep Learning Approach to Fuzzy String Matching

https://living-with-machines.github.io/DeezyMatch/

Other

139 stars 34 forks source link

Candidate finder new metrics, cleanup #46

Closed kasra-hosseini closed 4 years ago

kasra-hosseini commented 4 years ago

[WIP]

@mcollardanuy @fedenanni Similar to before, I am opening a PR now, so you can see the changes as we go forward.

TODO:

[x] rename context with hstates to better represent the inner working of the code
[x] Add DeezyMatch predictions and cosine similarity to candidateFinder
[x] Cleanup the path assignments in inference
[x] Cleanup the path assignments in combineVecs
[x] Cleanup the path assignments in candidateFinder
[x] ~New way of combining vectors in rnn, consistent with candidateFinder~

kasra-hosseini commented 4 years ago

@mcollardanuy @fedenanni DeezyMatch predictions and cosine similarities are now added to the outputs of candidateFinder. Currently, the ranking is as before. See https://github.com/Living-with-machines/DeezyMatch/pull/46/commits/83aaa25aa37b0bae48583cd1bed53f88aca0b104

kasra-hosseini commented 4 years ago

In this commit https://github.com/Living-with-machines/DeezyMatch/pull/46/commits/b525b53d9dd7dd7403bd394b837d198513d9666a, we can choose the ranking metric (faiss distance, cosine similarity and prediction outputs). I have updated README. If something is not clear, please let me know.

kasra-hosseini commented 4 years ago

@mcollardanuy @fedenanni Sorry for the delay. I am done with this PR. Please let me know if something is not clear and/or any changes are needed.

kasra-hosseini commented 4 years ago

@mcollardanuy I should have said return predictions, sorry. You are right. We output the metrics and print them on screen, but the function returns predictions for further use in, for example, candidateFinder.

mcollardanuy commented 4 years ago

Hi @kasra-hosseini, all good! Thanks!