← All notebooks for this

Cricket & talk-show sentiment

Build a Bangla sentiment dataset from commentary or talk-show transcripts and chart the mood minute by minute.

How did the mood change during a match or a talk show? This notebook turns Bangla commentary or talk-show transcripts into a labelled sentiment dataset, then charts the mood minute by minute.

What you do

  • Bring transcripts you have the right to use (the notebook will not download anything unless you confirm that).
  • Label sentences with an instruction-tuned Qwen2.5-7B model, spot-check a sample by hand, and fix the labels.
  • Plot sentiment over time.

Good to know

  • Political talk can be sensitive: label the speech, not the people, and do not publish named individuals' "scores".
  • Machine labels are a starting point, not truth — that is why the manual check is built in.