Closed jaguillette closed 1 year ago
lac
hasn't had updates since May 2021, so it isn't a recent update to that library, but one of its dependencies that my local environment picked up, paddlepaddle
got an update on November 17 of this year, so maybe that did it somehow? paddlepaddle
isn't in requirements/base.txt
though.
Makes sense to me. LAC was brought on expressly for a POC; it seemed best positioned when we were looking at it to just show that the functionality is possible, but if concerted effort was made to really get Chinese in a good, scalable and sustainable state, it would probably merit re-evaluating the library landscape.
LAC doesn't pin paddle
in any way:
https://github.com/baidu/lac/blob/master/python/setup.py
It's not clear to me from scanning the paddle
releases or the LAC issues whether it's fixable, or what's causing it, but LAC certainly doesn't look maintained at the moment, judging by the Issues activity; the only issues closed since early 2021 have been by the issue creators themselves.
Some relevant issues from LAC:
paddle 2.4.0
(the Nov 27 release), of the same issue we're seeing: https://github.com/baidu/lac/issues/256.
For some reason, the Python tests have started failing in GitHub actions. A failing run with logs can be seen here: https://github.com/Hedera-Lang-Learn/hedera/actions/runs/3603801082/jobs/6073350159. The issue seems to be with setting up
paddle
, which is being installed because oflac
, which is part of the lemmatizer for Chinese language text. Since we've removed Chinese language support from the app for the time being, this PR removeslac
from the requirements and comments out the tests that depend on it. The library and tests should be reinstated as part of bringing Chinese language support back in to the app, when that happens.