Lexical Stylometrics

[dishmint]

For a given text T and lexical pattern LP

• Get number of times a particular LP match occurs.
• Use that as a feature encoding of that match.
• Could probably test if a particular lexical pattern comes from a particular author.

[dishmint]

Just like basic sentiment analysis in tweets is counting positive / negative, count the different patterns..

[dishmint]

One can use the counts of a set of salient lexical patterns to perform stylometrics. I can look at the literature and see what kinds of patterns are significant in identifying an author, then train a network on those features.