Following a thread: low-complexity experiments revolving around the Major Arcana and Text Classification
- tarot_scraper: Get raw data from web, perform simple sainification stuff.
- tarot_stats: Perform basic stats on data
- bayes_classify: Classification using a naive bayes classifier from NLTK
- dist_bayes: Different approach than bayes_classify- trains one binary bayes classifier for each card, computes likelihood for each card and uses softmax to select prediction. ~2x improvement
- More Data, always.
- Experiment with LTSM & other DL classification techniques.
- Improve data structure (currently uses a dir of txt files. no bueno.)
- Improve Code structure- DRY stuff.