Lexicon based sentiment analysis.

Shapiro provides building blocks for sentiment analysis. The current implementation is a improved variant of code examples from a talk about Introduction to sentiment analysis with spaCy I gave at the EuroPython 2018 conference.

Shapiro's goals are:

• provide functions for common preprocessing tasks like replacing synonyms and unifying emojis/smiley codes
• provide means to read a sentiment lexicon and match tokens with it
• provide ways to model domain independent sentiment specific parts of Latin based languages (like negators and intensifiers)
• provide means to assign sentiment information to idioms
• (long term) provide ways to model a grammar for more advanced sentiment analysis similar to the one described Bing Liu (2015) "Sentiment Analysis: Mining Opinions, Sentiments, and Emotions".

Currently it provides basic support for:

• Users can provide lexicons (in CSV format) to assign topics and ratings to sentiment words
• Negators, intensifiers and diminishers (and combinations of them). Currently there is support for English and German. Adding more languages based on Latin should be possible.
• Lemmatization, part of speech tagging and basic language modelling by using spaCy.

This is a work in progress and features might be added and changed as seem fit. So the API is a moving target.

To use shapiro, you need Python and spaCy. The easiest way to get it is:

1. Download and install anaconda from https://www.anaconda.com/download/

2. Install spaCy:

   conda install -c conda-forge spacy

3. Install Tox:

   conda install -c conda-forge tox

4. Download the language models you need. For example to get English and German run:

   python -m spacy download en
   python -m spacy download de

For more information on installing spaCy visit https://spacy.io/usage/. For details on available free language models visit https://spacy.io/usage/models.