Improvements
- Processing the language models now performs a little faster by performing binary search on the language model NumPy arrays.
Bug Fixes
-
Several bugs in multiple languages detection have been fixed that caused incomplete results to be returned in several cases. (#143, #154)
-
A significant amount of Kazakh texts were incorrectly classified as Mongolian. This has been fixed. (#160)
Miscellaneous
-
A new section on performance tips has been added to the README.
-
All dependencies have been updated to their latest versions.