Surprisal

Surprisal classifies text by measuring the compression cost of input text against reference corpora using a hand-rolled LZ77 compression algorithm. It selects the language whose reference text encodes the input with the least bits. It requires no machine learning or training and is implemented as a single C file.