Shanghainese/Transducer

Github Repository: [1]

Evaluation

Total number of stems in transducer: 96

Current coverage: 6164 / 9221 (66.85%)

Current ambiguity: 10512 / 9221 (1.14)

List of unknown words: None, all tests passed

Notes

Number of tests passing: before was 56% and now 67% after adding all top unknown words.

(We know what all the words mean, adding them all would be laborious and make the code exponentially slower.)