mirror of
https://github.com/wassname/Castor.git
synced 2026-09-09 11:13:20 +08:00
+ sm model no external features baseline + sm model with IDF weights + sm model with IDF weights without removing punctuation --> barely better than df/idf (a la Pytorch). + sm model with stemming before computing IDF weights ^ all on the TrecQA dataset
1.9 KiB
1.9 KiB
| 1 | Baseline_method | idf | condition | dataset | train_MAP | dev_MAP | test_MAP | Notes/Comments |
|---|---|---|---|---|---|---|---|---|
| 2 | idf_sum_similarity | dataset | no_stopping + no_stemming | TrecQA | 0.7744 | 0.7668 | 0.7014 | IDF is computed over terms in the train, dev and test sets |
| 3 | idf_sum_similarity | dataset | with_stopping + with_stemming | TrecQA | 0.7078 | 0.7404 | 0.6816 | |
| 4 | idf_sum_similarity | dataset | with_stopping + with_stemming + stop(punct) | TrecQA | 0.7074 | 0.7337 | 0.6805 | also stopping punctuation |
| 5 | idf_sum_similarity | dataset | no_stopping + no_stemming | WikiQA | 0.2429 | 0.2489 | 0.2219 | There is less overlap between questions and answers in the WikiQA dataset |
| 6 | idf_sum_similarity | dataset | with_stopping + with_stemming | WikiQA | 0.2232 | 0.2472 | 0.2025 | |
| 7 | idf_sum_similarity | dataset | with_stopping + with_stemming + stop(punct) | WikiQA | 0.2236 | 0.2472 | 0.2025 | |
| 8 | ||||||||
| 9 | idf_sum_similarity | Corpus | with_stopping + with_stemming | TrecQA | 0.7032 | 0.7281 | 0.672 | IDF is computed over documents in disks1-5 and aquaint |
| 10 | idf_sum_similarity | Corpus | with_stopping + with_stemming | WikiQA | 0.2212 | 0.2461 | 0.2005 | IDF is computed over documents in Wikipedia |
| 11 | ||||||||
| 12 | sm_model | NA | no_idf_overlap_features | TrecQA | 0.597 | 0.6918 | 0.6445 | |
| 13 | sm_model | Dataset | with_corrected_idf_overlap_features | TrecQA | 0.919 | 0.7927 | 0.7384 | S&M model uses stopping only |
| 14 | sm_model | Dataset | with_corrected_idf_overlap_features – stop(punct) | TrecQA | 0.8325 | 0.7948 | 0.7562 | did not remove punctuation |
| 15 | sm_model | Dataset | with_stemmed_idf_overlap_features | TrecQA | 0.9414 | 0.8206 | 0.7363 | |
| 16 | sm_model | Corpus | with_corrected_idf_overlap_features | TrecQA | ||||
| 17 | sm_model | Dataset | with_fractional_idf_overlap_features | TrecQA | ||||
| 18 | sm_model | Corpus | with_fractional_idf_overlap_features | TrecQA | ||||
| 19 | sm_model | NA | no_idf_overlap_features | WikiQA | ||||
| 20 | sm_model | Dataset | with_corrected_idf_overlap_features | WikiQA | ||||
| 21 | sm_model | Corpus | with_corrected_idf_overlap_features | WikiQA | ||||
| 22 | sm_model | Dataset | with_fractional_idf_overlap_features | WikiQA | ||||
| 23 | sm_model | Corpus | with_fractional_idf_overlap_features | WikiQA |