mirror of
https://github.com/wassname/keras-language-modeling.git
synced 2026-10-04 12:40:33 +08:00
43 lines
773 B
Plaintext
43 lines
773 B
Plaintext
Best results achieved for each model:
|
|
|
|
Embedding + Max Pooling:
|
|
- Top 1 Precision:
|
|
- 0.492 on test 1
|
|
- 0.483 on test 2
|
|
- 0.495 on dev
|
|
- MRR:
|
|
- 0.624 on test 1
|
|
- 0.611 on test 2
|
|
- 0.624 on dev
|
|
|
|
Dense + CNN + Max Pooling:
|
|
- Top 1 precision:
|
|
- 0.507 on test 1
|
|
- 0.458 on test 2
|
|
- 0.515 on dev
|
|
- MRR:
|
|
- 0.635 on test 1
|
|
- 0.593 on test 2
|
|
- 0.642 on dev
|
|
|
|
Attentional LSTM + Max Pooling:
|
|
- Top 1 precision:
|
|
- 0.480 on test 1
|
|
- 0.465 on test 2
|
|
- 0.487 on dev
|
|
- MRR:
|
|
- 0.627 on test 1
|
|
- 0.613 on test 2
|
|
- 0.635 on dev
|
|
|
|
Unsupervised RNN language model + trained embeddings:
|
|
- Top 1 precision:
|
|
- 0.546 on test 1
|
|
- 0.527 on test 2
|
|
- 0.552 on dev
|
|
- Mrr:
|
|
- 0.670 on test 1
|
|
- 0.651 on test 2
|
|
- 0.671 on dev
|
|
|