Skip to main content
. Author manuscript; available in PMC: 2025 Mar 5.
Published in final edited form as: Proc Int AAAI Conf Weblogs Soc Media. 2024 May 28;18:422–434. doi: 10.1609/icwsm.v18i1.31324

Table 3:

Comparison of the explanations obtained by LIME with three different types of textual cues annotated as Trigger, LoST indicators, and consequences using ROUGE scores and BLEU scores. ([Black, italicize, underlined] + bold) represents the highest values of [ATS: All Test Samples, TP: True Positives, TN: True Negatives], respectively. Higher the value of LoST indicators and lower (↓) the values with Trigger and consequences, better is the classifier.

Model Evaluation Trigger (↓) LoST Indicators (↑) Consequences (↓)
ATS TP TN ATS TP TN ATS TP TN

BERT ROUGE 0.0756 0.0705 0.0868 0.2773 0.3128 0.1985 0.0421 0.0442 0.0375
BLEU 0.0519 0.0446 0.0681 0.1656 0.1760 0.1426 0.0315 0.0349 0.0241
ALBERT ROUGE 0.0987 0.0973 0.1014 0.2201 0.2424 0.1777 0.0359 0.0427 0.0229
BLEU 0.0657 0.0595 0.0774 0.1363 0.1367 0.1356 0.0271 0.0318 0.0179
DistilBERT ROUGE 0.0826 0.0837 0.0809 0.2471 0.2738 0.2051 0.0351 0.0273 0.0474
BLEU 0.0569 0.0564 0.0579 0.1451 0.1412 0.1513 0.0260 0.0219 0.0324
DeBERTa ROUGE 0.0687 0.0645 0.0790 0.2461 0.2570 0.2193 0.0468 0.0465 0.0476
BLEU 0.0459 0.0389 0.0634 0.1586 0.1513 0.1724 0.0346 0.0341 0.0359

ClinicalBERT ROUGE 0.0728 0.0729 0.0726 0.2045 0.2304 0.1639 0.0369 0.0359 0.0387
BLEU 0.0463 0.0451 0.0485 0.1260 0.1272 0.1241 0.0286 0.0288 0.0284
PsychBERT ROUGE 0.0946 0.0729 0.1272 0.2310 0.2304 0.1949 0.0307 0.0359 0.0415
BLEU 0.0648 0.0451 0.0860 0.1336 0.1272 0.1414 0.0226 0.0288 0.0325
MentalBERT ROUGE 0.0820 0.0733 0.1047 0.2614 0.2821 0.2076 0.0160 0.0126 0.0248
BLEU 0.0464 0.0348 0.0768 0.1544 0.1521 0.1604 0.0107 0.0091 0.0147