empythy

empathyMachines

A standalone NLP sentiment classifier you can import as a module

Purposes

1. Offer a batteries-included NLP classifier you can use either on it's own, or to make sentiment predictions as part of a broder NLP project (for example, when classifying customer messages, whether the customer is angry or not might help you determine if this is a compensation request, or a request to adjust their address.)
2. Have the entire sentiment prediction process scaffolded so you can feed in your own training corpus, and easily train an NLP sentiment classifier.

How to use

1. pip install empythy

from empythy import EmpathyMachines
nlp_classifier = EmpathyMachines()
nlp_classifier.train()
nlp_classifier.predict(text_string)

Corpora included

NLTK Movie Reviews

The classic sentiment corpus, 2000 movie reviews already gathered by NLTK.

Assembling a custom Twitter sentiment corpus

CrowdFlower hosts a number of Twitter corpora that have already been graded for sentiment by panels of humans.

I aggregated together 6 of their corpora into a single, aggregated and cleaned corpus, with consistent scoring labels across the entire corpus. The cleaned corpus contains over 45,000 documents, with positive, negative, and neutral sentiments.

Train on your own corpus

Feel free to train a classifier on your own corpus!

Two ways to do this

1. Read in a .csv file with header row containing "sentiment", "text", and optionally, "confidence"
   • Pass the name of the .csv file to train, like so:
   • nlp_classifier.train(corpus='custom', corpus_path='path/to/custom/corpus.csv')
2. Pass in an array of Python dictionaries, where each dictionary has attributes for "sentiment", "text", and optionally, "confidence"
   • nlp_classifier.train(corpus='custom', corpus_array=my_array_of_texts)
   • Two important parts to this, both corpus='custom', and corpus_array=my_variable_holding_the_documents.

Advanced Usage

1. nlp_classifier.train(verbose=False) to turn off print status statements while training.
2. nlp_classifier.train(print_analytics_results=True) to print out results of training the classifier.