LyricPsych: towards a psychologically validated feature extractors for the lyrics data

This package contains a number of feature extraction methods on textual data, mostly expecting musical lyrics. As an work-in-progress development phase, we will continuously validate and update the package later on. Currently, several feature extraction methods including linguistic features, topic_modeling, psychology inventory based feature estimations, etc.

Getting Started

To install this package, we recommend you use the python virtual environment. Inside the virtualenv, installation is using pip and git.

$ pip install git+https://github.com/mmc-tudelft/lyricpsych.git

feature extractor lyricpsych-extract installed along with the package. The usage of the lyricpsych-extractor can be found in the -h option. For instance, you can extract personality, value, topic, linguistic features by using the example below:

$ lyricpsych-extract \
    /path/to/the/lyrics_data.csv \
    /path/for/output/ \
    --w2v glove-twitter-25 \
    --features linguistic value personality topic

Contributing

Currently it's in its alpha version. It means some extractors are not fully validated, and may have a unexpected behavior. We will continue to work on improving those aspects, but also we are more than welcoming contributions. We are open to take issues and pull request.

TODO list

• refactoring
  • split extractor to dedicated extractors
  • minor refactorings
  • clean up
    • unused functions
    • unused data files
    • unused scripts
  • restructuring
    • split task to the separate sub-module
    • separate fm and als_feat to the separate repo
• Documentation
  • docstrings
  • doc generation
• features
  • experimental run reproduction cli
• deploy
  • writing testings
  • CI [Travis integration]
  • register to PyPI

Authors

Jaehun Kim, Sandy Manolios

Reference

TBD

Acknowledgement

• Thanks to henryre, as our PLSA implementation of this API is extension of numba-plsa.