PyDelta

pydelta is a library to perform stylometric analyses like authorship attribution, and to evaluate methods for that. It originated as a commandline tool which implements three algorithms in the form described by Argamon in a paper on John Burrows Delta. (S. Argamon, "Interpreting Burrows’s Delta: Geometric and Probabilistic Foundations," Literary and linguistic computing, vol. 23, iss. 2, pp. 131-147, 2008.)

Delta is a measure to describe the stylistic difference between texts. It is used in computational stylistics, especially in author attribution. This implementation is for research purposes only, If you want to use a reliable implementation with a nice Gui and much more features you should have a closer look at the great R tool Stylo.

Documentation

Documentation for the _next_ branch is updated on each commit.

Different branches, different versions

There are currently three different long-running branches:

1. The master branch contains a rather old version of pydelta capable of running single experiments with a bunch of delta measures. Will be replaced sometime soon.
2. The develop branch is based on the script in the master version, but it contains more delta methods as well as some very unpolished command line scripts that we used, e.g., in order to produce the results presented at DH 2015 in Sidney.
3. The next branch is more or less a re-implementation that is not yet feature complete. It does not contain a command-line executable, and some of the methods that did not perform well have been left out.

Installation and Requirements

PyDelta requires Python 3.3 or newer. It has quite a set of dependencies (NumPy, Pandas, SciPy, SciKit-Learn, …), but it comes with a setup script that installs it with its dependencies.

If you only wish to use it, not to hack on it, pip can clone and install it for you:

pip install git+https://github.com/cophi-wue/pydelta@next

(Note that the Python 3 version of pip is sometimes called pip3).

Developers can clone the repo and run pip to install:

git clone -b next https://github.com/cophi-wue/pydelta
cd pydelta
pip install -r requirements.txt

This will also install a version of scikit-learn that features k-Medoids-Clustering which is not in the official version yet. You may need to install numpy, scipy, and Cython separately first.

---

Thanks go to

• Thorsten Vitt for his help with profiling some critical parts and general improvements
• Allan Riddell for advice on matplotlib