Sentiment Analysis

[JBarsotti]

Would it be possible to use Trove to perform sentiment analysis on clinical notes? I have access to all the MIMIC-IV discharge notes but am struggling to find a good tool to use for sentiment analysis. It seems like most models are not designed for such use.

[jason-fries]

Hi @JBarsotti, In the most general sense, Trove enables weakly supervised sequence labeling tasks (i.e., per-token classification). We focus mostly on tools for information extraction-style tasks on clinical text (NER, relation extraction), but you could use Trove for any task that can be formulated as sequence labeling or sentence classification. So depending on how you formulate your sentiment analysis task, I'd say it's worth giving Trove a shot.

[JBarsotti]

Thank you so much for the reply! I think I will definitely give it a shot. Will report back with my success or failure! 🙂

---

From: Jason Alan Fries @.> Sent: Sunday, May 21, 2023 10:01 PM To: som-shahlab/trove @.> Cc: Barsotti, Ercole @.>; Mention @.> Subject: [External] Re: [som-shahlab/trove] Sentiment Analysis (Issue #9)

Hi @JBarsottihttps://github.com/JBarsotti, In the most general sense, Trove enables weakly supervised sequence labeling tasks (i.e., per-token classification). We focus mostly on tools for information extraction-style tasks on clinical text (NER, relation extraction), but you could use Trove for any task that can be formulated as sequence labeling or sentence classification. So depending on how you formulate your sentiment analysis task, I'd say it's worth giving Trove a shot.

— Reply to this email directly, view it on GitHubhttps://github.com/som-shahlab/trove/issues/9#issuecomment-1556448761, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AR5U2T6XVYEGB3XVMVG7PQDXHLJIJANCNFSM6AAAAAAYJQJPEM. You are receiving this because you were mentioned.Message ID: @.***>

[jason-fries]

Awesome! Let me know if you have any questions :)