Make topic disambiguation optional?

[richardmillson]

One use case for toponymy is to have the LLM perform cluster-level classification with respect to some fixed label ontology. However since multiple clusters may be given the same label, this is in conflict with the disambiguation step.

The disambiguation step could be run conditional on a disambiguate_topic_names: Optional[bool] = True flag added to name_topics (and optionally to the methods that call it).

Does this sound like a use case toponymy should support?

[jc-healy]

Hi Richard, yep that makes a decent amount of sense for Toponymy to support (and it should be a pretty straight forward code change). Do you have enough cycles to submit a PR with a few unit tests?

[richardmillson]

Hi John, in trying to run the tests I ran into some issues and opened #113 .

I should be able to submit a PR implementing this feature now that I can run the tests.

I see that name_topics is marked with pragma: no cover, so it looks like there are no existing unit tests for this code?

[jc-healy]

Hi Richard, great job on issue #113, I expect to merge that PR very soon.

Regarding the name_topics my impression is that the pragma: no cover should just turn off coverage testing (and acknowledge that we are doing this on purpose). The real issue here is that the name_topics requires LLM integration to be done well. That is quite expensive and we haven't yet integrated external LLM api_keys into our continuous testing procedure (we are currently planning on getting this working sometime after April 1st).

That said, a few mocked tests certainly wouldn't go awry.