Stylometric (parallel) framework in Python for big data in clusters

• Parallelized (thus fast)
• Intended to integrate with a database-based corpus
• A variety of feature-generation techniques:
  • byte-ngrams
  • word-ngrams
  • readability metrics
  • simple statistics
  • part-of-speech tagging
  • part-of-speech ngrams
  • word/pos hybrids
• Plugs into a variety of stylometric techniques:
  • ppm-c (compression)
  • dmc (compression)
  • gvc (spam-filter)
  • sofia-ml (machine learning)
• Some graphing utilities to show performance

We also provide some plugs to transform existing corpora into database format. We also provide some plugs to export features into SVM-light sparse data format.

We assume you have lots of RAM or lots of time or lots of CPU cores or all 3.

• 30 million comments generally takes about a day to process 1 type of feature
• 3 million posts generally takes about an hour to process 1 type of feature