Table 10 Distribution of error types for medical entity recognition (MER) and keyphrase extraction (KE). # columns indicate the occurrence number of corresponding error type, while % columns indicate the corresponding ratio
From: Sentences, entities, and keyphrases extraction from consumer health forums using multi-task learning
Model | MS | US | ISO | IT | ||||
|---|---|---|---|---|---|---|---|---|
# | % | # | % | # | % | # | % | |
Medical Entity Recognition (MER) | ||||||||
\(\text {IndoNLU}_{\text {LARGE}}\) (STL) | 55 | 11.55 | 83 | 17.44 | 323 | 67.86 | 15 | 3.15 |
\(\text {XLM-R}_{\text {LARGE}}\) (STL) | 61 | 13.23 | 60 | 13.02 | 328 | 71.15 | 12 | 2.60 |
Parallel MER - KE (\(\alpha = 0.7\)) | 83 | 17.22 | 58 | 12.03 | 326 | 67.63 | 15 | 3.11 |
Hierarchical Three-way (\(\alpha = 0.4; \beta = 0.4\)) | 69 | 14.94 | 69 | 14.94 | 309 | 66.88 | 15 | 3.25 |
Keyphrase Extraction (KE) | ||||||||
\(\text {IndoLEM}_{\text {BASE}}\) (STL) | 69 | 19.06 | 31 | 8.56 | 262 | 72.38 | 0 | 0.00 |
Parallel MER - KE (\(\alpha = 0.3\)) | 79 | 22.44 | 15 | 4.26 | 258 | 73.30 | 0 | 0.00 |
Hierarchical MER - KE (\(\alpha = 0.3\)) | 109 | 29.46 | 17 | 4.59 | 244 | 65.95 | 0 | 0.00 |
Parallel Three-way (\(\alpha = 0.4; \beta = 0.2\)) | 52 | 14.99 | 25 | 7.20 | 270 | 77.81 | 0 | 0.00 |