Skip to main content
. 2026 Sep 14;28:e96771. doi: 10.2196/96771

Table 2.

Distribution of sentiment classes across models and perspectives.

Sentiment column Positive, n (%) Neutral, n (%) Negative, n (%)
DeepSeek-7B general 12 (0.07) 16330 (99.29) 105 (0.64)
DeepSeek-7B patient 7101 (43.18) 7597 (46.19) 1749 (10.63)
DeepSeek-7B physician 739 (4.49) 15339 (93.26) 369 (2.24)
Mistral-7B general 3709 (22.55) 10730 (65.24) 2008 (12.21)
Mistral-7B patient 1608 (9.78) 11668 (70.94) 3171 (19.28)
Mistral-7B physician 1926 (11.71) 13726 (83.46) 795 (4.83)
TextBlob 6384 (38.82) 9859 (59.94) 204 (1.24)
VADERa 8810 (53.57) 158 (0.96) 7479 (45.47)
ClinSent-lexicon 9784 (59.49) 2948 (17.92) 3715 (22.59)

aVADER: Valence Aware Dictionary and Sentiment Reasoner.