search

Ravoxsg / SummaReranker

Source code for SummaReranker (ACL 2022)
MIT License
24 stars 10 forks source link

Could you provide a detailed instruction #1

Closed StevenTang1998 closed 2 years ago

StevenTang1998 commented 2 years ago

Hello, congratulations on your work being accepted for ACL 2022!

I want to follow your work and reproduce the results. However, the script you provide didn't run successfully in my environment. There seems to be some local paths, and the location and format of the dataset are not particularly clear.

Could you offer some help and wish you a happy life!

Ravoxsg commented 2 years ago

Hi @StevenTang1998 , thanks!

So in order to run the re-ranker, you need to first generate a set of summary candidates This is done in src/candidate_generation/main_candidate_generation.py This will generate num_beams candidate for each source document in the dataset.

Then, you need to score these candidates to evaluate the re-ranking. This is done in src/candidate_generation/main_scores.py

Let me know if you need any help running these scripts.

Ravoxsg commented 2 years ago

@StevenTang1998 Please follow instructions from the other similar issue: https://github.com/Ravoxsg/SummaReranker/issues/2