Neural Document Classification

castorini / castor

PyTorch deep learning models for text processing

http://castor.ai/

Apache License 2.0

178 stars 58 forks source link

Neural Document Classification #159

Closed achyudh closed 5 years ago

achyudh commented 5 years ago

This is a work in progress. Please don't merge until all the tasks below have been checked. @Ashutosh-Adhikari feel free to add relevant tasks to the list below.

[x] Checkpoint models and perform dev evaluation only at the end of an epoch
[ ] Replace RCV-1 dataset with the 103-class Lewis split
[x] Add support for Yelp, IMDB and DBPedia datasets
[x] Verify performance of LSTM with Regularization after model checkpointing

Ashutosh-Adhikari commented 5 years ago

Let us not replace RCV-1 mod-apte, as of now how about adding RCV-1 103 lewis split. As previously discussed.

achyudh commented 5 years ago

I agree. LSTM-Reg doesn't seem to be performing well on RCV-1 (103 classes) with splits from the SGM paper. Until we find a proper split for the dataset, we would have to stick with the current ModApte splits.

achyudh commented 5 years ago

Corresponding merge request in Castor-data: https://git.uwaterloo.ca/jimmylin/Castor-data/merge_requests/12