Before you begin: Be sure to download a word frequency data set to build the bigram model. This project uses the Brown data set (1961), but you can change that easily.

python -c "import nltk; nltk.download('brown')"

Overview

The scripts in this repo generate genre-tailored anagrams from a phrase that you provide.

The word search script tries to match your string with up to 3 words from one or more dictionaries. You can pass the output of the search script to a scorer, which has configurable genre profiles. Each profile tries to score the anagram words according to a seed vocabulary. The vocabulaty allows you to tune the scorers response by using word-count score and optional pattern-matched bonuses.

Usage

1. Generate a word list by using word-search.py. Be sure to add your phrase to word-search. I'll make this a command-line option at some point.
2. Use scorer.py to generate your top N candidates. I've included three profiles, which you can edit and tune to your liking. The medical profile includes a multiplier to boost the score of certain features.

# Add your phrase to word-search.py
python word-search.py > output.csv
python scorer.py output.csv --profile institutional --top 50