inception-project / inception

INCEpTION provides a semantic annotation platform offering intelligent annotation assistance and knowledge management.
https://inception-project.github.io
Apache License 2.0
585 stars 149 forks source link

Error reading a gazeteer is not visible to the user #3545

Closed dagosgi closed 1 year ago

dagosgi commented 1 year ago

Describe the bug

I am encountering some issues with recommenders that should be informed by a gazetteer. Basically, I have tried all that would come to my mind to make it work, but the internal recommender "string matcher" does not seem to recognize the .txt file I uploaded as a dictionary from which to retrieve terms and matching tags for annotation. Always active or not, with or without evaluation, trying to retrain the recommender every time by starting to insert some terms manually...still, it behaves as if the gazetteer wasn't there. The text file is formatted correctly, I double checked, and no error message showed up any time I deleted and reuploaded the gazetteer, nor when I tried deleting and recreating the feature the recommendation is based upon, or the whole layer.

To Reproduce

No response

Expected behavior

Upon activation of a recommender system with an integrated gazetteer, I'd expect some recommendations to show up, based on the terms that can be found both in the dictionary and in the text. This doesn't happen.

Screenshots

image

Environment

Version and build ID: 25.1 (2022-10-11 21:15:38, build e85988ac) Operating system: e.g. Windows Browser: Chrome, Firefox

Additional context

Here's a sample project with selected files and the dictionary we are supposed to use in the original project. esg-factors-gazetteer1408288947898743567.zip

reckart commented 1 year ago

When trying to load the gazeteer, a message is logged that line 512 does not contain a tab character. After fixing that line, the project works.

dagosgi commented 1 year ago

Oh my...so the problem arised because that specific word at line 512 mistakenly didn't have any associated tag (?) I went and fixed that, now that you told me...and apparently the problem is solved for real.

I have a follow-up question, then: now I'm feeling dumb, but where could I have found out myself that this was the issue? I could not (and still cannot) find this information in the recommender log, nor (as I said before) any error message showed up upon upload of the dictionary. Where should I look? Thank you

reckart commented 1 year ago

The info was in the server/console log. But I'll fix it up so that the issue will also appear in the recommender log accessible from the UI in the next release.

So thanks for the report! 👍

dagosgi commented 1 year ago

Perfect, looking forward. Thank you for your help!