Tag sentences in public comment with which bill, or minutes item, or more general topic they are discussing

[evamaxfield]

Idea / Feature

Tag sentences in public comment with the general topic they are discussing but if they are additionally classified as related to a specific minutes item or bill tag it as such.

Use Case / User Story

As a user: I want to see which public comments discussed a bill so I can quickly scan the public comments to see who commented about something and how they felt.

As a researcher: I want to quickly build a dataset of public comments about a certain topic.

Other Notes

[isaacna]

Is identifying public comments something we'll need speaker classification for or is this provided via something like Legistar?

[evamaxfield]

Legistar doesn't have this info from what I can tell and since all municipalities may not use Legistar regardless we shouldn't depend on it for this.

In my head this was releated to minutes item segmentation.

Segment minutes items -> great now we know which section of the transcript is public comments -> segment each public comment -> great now we have a dataset of individual public comments -> classify if the public comment is about minutes item by checking the semantic similarity of the whole public comment vs the minutes item title and description: https://huggingface.co/sentence-transformers/paraphrase-xlm-r-multilingual-v1