bitermplus: v0.9.1

• Fixed platform type mismatch issue (#40)

bitermplus: v0.9.0

• Python 3.13 support: added Python 3.13 compatibility, dropped Python 3.8
• Updated dependencies: upgraded Cython to 3.0, added Setuptools compatibility layer
• Enhanced stability: improved numerical stability and robustness
• Documentation improvements: rewritten docstrings with better examples and explanations
• Code cleanup: fixed typos and cleaned up code issues

bitermplus: v0.8.0

New Features

• Sklearn-style API: New BTMClassifier with familiar fit(), transform(), and fit_transform() methods
• ML Pipeline Integration: Native support for sklearn workflows, cross-validation, and grid search
• Enhanced Topic Interpretation: Built-in methods for extracting topic keywords and document assignments

Improvements

• A bit faster inference and improved memory efficiency
• Parameter validation with clear error messages
• Updated pyproject.toml with professional development tools

Bug Fixes

• Fixed serialization issues in BTM model state
• Corrected test assertions and parameter validation order
• Resolved coherence calculation boundary conditions

Backward Compatibility

All existing code using the traditional API continues to work unchanged. The new sklearn-style
API provides a simpler, more accessible interface for new users.

bitermplus: v0.7.0

This release introduces minor fixes and improvements of documentation and metrics calculation. Its packaging is now based on pyproject.toml and setuptools.

bitermplus: v0.6.12

This release contains some minor fixes and adds labels_ property to BTM model class (labels for the most probable topics for each of the documents). It also adds get_docs_top_topic method for creating DataFrames with documents and their labels.

bitermplus: v0.6.11

This release fixes the incompatibility error between bitermplus and scikit-learn.

bitermplus: v0.6.10

This release includes a number of minor fixes. Methods to select stable topics have been moved to tmplot package. Please see the updated tutorial in the documentation.

bitermplus: v0.6.9

This release introduces a function for Renyi entropy calculation (bitermplus.entropy) that can be used to estimate the optimal number of topics. For more details, read this paper.

bitermplus: v0.6.8

This release is an attempt to fix the issue with perplexity calculation yielding infinity values (#7).

bitermplus: v0.6.7

This release drops support for pyLDAvis in favor of tmplot that can be installed with pip (optional):

pip install tmplot