search

Albert-Ma / PROP

WSDM'2021, PROP and SIGIR'2021,B-PROP
Apache License 2.0
110 stars 18 forks source link

Fail to download the pretrained model #1

Closed jingtaozhan closed 3 years ago

jingtaozhan commented 3 years ago

Hi, this is a great work! However, I failed to download the pretrained model. The given link points to an empty folder. Is the pretrained model available yet?

Albert-Ma commented 3 years ago

Hi, I released the prop_step400k base model which we got first place on the MSMARCO Document Ranking Leaderboard. Note that the prop_step400k base is different from the model we used in our WSDM 2021 paper and these models will be available soon. Too busy now.

jingtaozhan commented 3 years ago

Great! Congratulations! I assume you use PROP for second-stage ranking. Which model did you use for first-stage ranking? I'm currently working on first-stage retrieval and it performs very well on MS MARCO document ranking task. I think maybe we can work together and further improve the document retrieval performance.

Albert-Ma commented 3 years ago

Sorry for the delay. Yeah, PROP is designed for reranking. Currently, for the result we submitted to the leaderboard, we use BM25 to retrieve top100 doc2query augmented docs. We are digging into dense presentations too and I am glad to cooperate with you in a proper way. Maybe we can talk about this via email. Here is my email: maxinyu17g@ict.ac.cn