Question: working with SpaCy

TeamHG-Memex / eli5

A library for debugging/inspecting machine learning classifiers and explaining their predictions

http://eli5.readthedocs.io

MIT License

2.75k stars 331 forks source link

Question: working with SpaCy #260

Open DonaldTsang opened 6 years ago

DonaldTsang commented 6 years ago

Does this library work with Textacy (SpaCy and Scikit-Learn)?

lopuhin commented 6 years ago

@DonaldTsang we never tested it with Textacy, so white-box (eli5.explain_prediction, eli5.explain_weights) is unlikely to work out of the box. Black-box (http://eli5.readthedocs.io/en/latest/blackbox/lime.html) approach should work though. Which models from Textacy are you interested in?

DonaldTsang commented 6 years ago

The main focuses are:

Text language classification (English, French, German, Spanish)
tokenization, text cleaning, POS tagging and OOV detection
Vectorization (internal SpaCy, Word2Vec and Fasttext with OOV)
LDA, PLSI/PLSA, LSI/LSA and PCA/NMF topic-model clustering
Sentiment analysis and context analysis

Regarding text and user classification, the main focus are Naive Bayes, SVM and Neural Networks.