Profanity check SGID data

[gregbunce]

it would be helpful to have a function that looks through the data names and scans for derogatory names in the data - think trailheads, trail names, place names, etc.. This could be a good opportunity to leverage AI.

[gregbunce]

FYI: we do have a derogatory name in the trailheads data - it's a former name. I'm working on this now to clean it up.

[gregbunce]

a possible solution to look at:
https://github.com/surge-ai/profanity

[steveoh]

I think we'd probably stick to gcp or maybe aws.

https://cloud.google.com/natural-language/docs/moderating-text

[gregbunce]

Moving this FY25 Q1 and hopefully things will settle down a bit by then to make some progress on it.

[steveoh]

I submitted a google request to get the "s" word added. Are there any other words we know about in our data that needs to be replaced?

I tested the word and it's still not being flagged! Since I couldn't find the original request I created a new one (internal ref: 377718296).
Product let me know the way I submitted the feature request should work to add new words. It may still take time to implement but I'll be able to provide you updates.
Please let me know what other words we should be tracking.

[ZachBeck]

Removed squaw names from NHD Lakes, NHD Streams, and UGRC version of the GNIS.

[steveoh]

Good news, engineering rolled out a new version of the API and this now flags the derogatory word. To test you’ll need to include "modelVersion": "MODEL_VERSION_2" on the API call.

Using the examples from Wikipedia, it now flags under the “Derogatory” category.

Let me know if you have any questions or other examples.