Castor

mirror of https://github.com/wassname/Castor.git synced 2026-08-28 12:42:36 +08:00

Files

T

History

Michael Tu 5bf33bf8ea Update README to use Castor-models and Instructions for Internal Users (#113 )

* Update instructions to use Castor-models
* Consolidate requirements.txt
* Refine README with convenience scripts
* Update internal instructions
* MP-CNN working dir minor edit

2018-05-25 18:14:10 -04:00

__init.py__.py

Update readme (#90 )

2017-12-05 21:58:55 -05:00

.gitignore

Update bridge (#87 ) (#89 )

2017-12-05 12:23:08 -05:00

args.py

Add flag to support onnx (#91 )

2017-12-06 09:55:02 -05:00

bridge.py

Migrate to from GitHub castorini/data to uWaterloo Castor-data (#103 )

2018-05-23 16:16:17 -04:00

create_dataset.sh

Migrate to from GitHub castorini/data to uWaterloo Castor-data (#103 )

2018-05-23 16:16:17 -04:00

external_features.py

Corpus idf (#23 )

2017-05-07 22:28:35 -04:00

main.py

Add flag to support onnx (#91 )

2017-12-06 09:55:02 -05:00

model.py

Use view instead of unsqueeze since ONNX v1.0 doesn't support it (#94 )

2017-12-08 15:04:29 -05:00

overlap_features.py

Migrate to from GitHub castorini/data to uWaterloo Castor-data (#103 )

2018-05-23 16:16:17 -04:00

README.md

Migrate to from GitHub castorini/data to uWaterloo Castor-data (#103 )

2018-05-23 16:16:17 -04:00

train.py

Add flag to support onnx (#91 )

2017-12-06 09:55:02 -05:00

trec_dataset.py

Update bridge (#87 ) (#89 )

2017-12-05 12:23:08 -05:00

wiki_dataset.py

Update bridge (#87 ) (#89 )

2017-12-05 12:23:08 -05:00

README.md

SM model

References:

Aliaksei _S_everyn and Alessandro _M_oschitti. 2015. Learning to Rank Short Text Pairs with Convolutional Deep Neural Networks. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '15). ACM, New York, NY, USA, 373-382. DOI: http://dx.doi.org/10.1145/2766462.2767738

Please ensure you have followed instructions in the main README doc before running any further commands in this doc.

Your repository root should be in your PYTHONPATH environment variable:

export PYTHONPATH=$(pwd)

To create the dataset:

cd Castor/sm_cnn/
./create_dataset.sh

We use trec_eval for evaluation:

cd ../utils/
./get_trec_eval.sh
cd ../sm_cnn

Training

Download the word2vec model from here and copy it to the data/ folder.

You can train the SM model for the 4 following configurations:

random - the word embedddings are initialized randomly and are tuned during training
static - the word embeddings are static (Severyn and Moschitti, SIGIR'15)
non-static - the word embeddings are tuned during training
multichannel - contains static and non-static channels for question and answer conv layers

To train on GPU 0 with static configuration:

python train.py --mode static --gpu 0

NB: pass --no_cuda to use CPU

The trained model will be save to:

saves/static_best_model.pt

Testing the model

python main.py --trained_model saves/TREC/multichannel_best_model.pt

Evaluation

The performance on TrecQA dataset:

TrecQA:

Best dev

Metric	rand	static	non-static	multichannel
MAP	0.8096	0.8162	0.8387	0.8274
MRR	0.8560	0.8918	0.9058	0.8818

Test

Metric	rand	static	non-static	multichannel
MAP	0.7441	0.7524	0.7688	0.7641
MRR	0.8172	0.8012	0.8144	0.8174

WikiQA:

Best dev

Metric	rand	static	non-static	multichannel
MAP	0.7109	0.7204	0.7049	0.7245
MRR	0.7169	0.7234	0.7075	0.7259

Test

Metric	rand	static	non-static	multichannel
MAP	0.6313	0.6378	0.6455	0.6476
MRR	0.6522	0.6542	0.6689	0.6646

NB: The results on WikiQA are based on the SM model hyperparameters.

To create your own word2vec.pt file

Download word2vec from here to the data/ folder

python $PYTHONPATH/utils/build_w2v.py --input ../../Castor-data/embeddings/word2vec/aquaint+wiki.txt.gz.ndim=50.bin

Note that $PYTHONPATH holds the location of the repository root.