Conversation
Unique id's in the Form class
- drop earlier deprecated cols - consider "Seg." as deprecated - add CLTS_id
A more detailed log for entropy computations
Fallback strategy for feature names
* add POS CLI option * add doc * fix typo in log management * pos for entropies * add paralex dep
* add frequency management * add a small test package * add paralex to dependencies * normalized index names, new info function * Call logger Qumin --------- Co-authored-by: Sacha Beniamine <sacha@git.none>
* update doc and config * drop computations * replace by phon and edit & unify names * Unify find_patterns methods (a single one remaining) --------- Co-authored-by: Sacha Beniamine <sacha@git.none>
# Conflicts: # sphinx/changelog.rst
Deprec bipartite computations (closes #41)
…79) * Separate csv (machine readable) and novel md (human readable) export --------- Co-authored-by: Sacha Beniamine <sacha@git.none>
* rewrite Metadata class * add changelog and doc * improve docstrings * extending new interface to entropies * renamed importFile in importResults to keep clear semantics.
* Add options to select among overabundant entries. * Removes duplicate phon_forms for the same lexeme & cell * Unify overabundance config usage + doc --------- Co-authored-by: Sacha Beniamine <sacha@git.none>
* fix microclass computation * add integration tests * hotfix with inplace paradigms * Long to wide before using pdist * corrects overabundant to overabundant.keep * Allow import of overabundant patterns * Fix ParadigmPatterns from_file needs a boolean for overabundant --------- Co-authored-by: Jules Bouton <jules.bouton@u-paris.fr> Co-authored-by: Sacha Beniamine <sacha@git.none>
* Adds to ent_heatmap: zones graphics, distillation heatmap * few fixes, deprecated applymap, PEP conformity * Adding frequency colorbars when available * Hides debug heatmap if not verbose * add option to hide freq_margins * heatmap formatting * extract and save pair and predictor probabilities + add n_preds frequencies to heatmaps Note: sets vmin to 0 for heatmaps, including with number of pairs. The previous plot was treacherous as it used the full gradient for whatever variation we had: light colors looked like we had little data, when it could still mean thousands of pairs. This will be more interpretable at a glance (although it hides fine variation in number of pairs) This implementation easily scales to new metrics (as the probability of success) and to debug computations. It will also significantly shorten PR #67 --------- Co-authored-by: Sacha Beniamine <sacha@git.none> Co-authored-by: Jules Bouton <jules.bouton@u-paris.fr>
|
Also, doc run is failing on readthedocs, I need to check why. |
|
Oh, it's sphinx immaterial. I did like the previous theme, can you remind me why you wanted to switch to sphinx immaterial ? |
Sorry, do not hesitate to switch back to the default theme. I think the motivations where that I wasn't fond of this serif font + the API rendering is not that nice. But this is very subjective! Do the simplest for the deployment |
|
Ok, last decision: @julesbouton, do we want to use gitlab starting on version 3 ? I'll make a todolist with steps in #87 |
Yes, I think it would be nice! |
|
On it :) |
Here are a few todo to tidy everything before merging:
Do we want to use gitlab starting on version 3 ?
Once merged: