Changes in relatio
- Upload to PyPI
- New API with Preprocessor() and NarrativeModel() classes
- Support for SentenceTransformers and spaCy models
- Dimension reduction via PCA/UMAP
- Automatic selection of the "optimal" number of clusters based on well-known metrics for HDBSCAN / KMeans.
- Visualization functions for the cluster selection metrics (i.e., elbow method, silhouette score, and DBCV).
- Easy cluster visualization and inspection.
- Multi-lingual support via spaCy language models.
- Dependency parsing to extract SVOs (much much times faster than SRL and pretty decent output).
- Upload to PyPI
- Update documentation
- Fixed a bug with
top_n_entities
inbuild_narrative_model
. It was not saved in the narrative_model and thus did not filter for the top_n_entities when callingget_narratives
. - Set
lemmatize = True
as the default text preprocessing inbuild_narrative_model
. - Updated
tutorial.ipynb
- use
allennlp_models.structured_prediction.predictors.SemanticRoleLabelerPredictor
instead ofallennlp.predictors.predictor
- public release