Refactor as package

[cstorm125]

Thai-language specific metrics

Sequence classification

sklearn implementation

• [ ] accuracy
• F1
• precision
• recall
• prevalence

Token classification

seqeval at entity level

• [ ] accuracy
• F1
• precision
• recall
• support

Question answering

at character-, syllable-, newmm and deepcut word levels

• [ ] F1
• chrf
• exact match

Sequence-to-sequence

at character-, syllable-, newmm and deepcut word levels

• [ ] BLEU
• chrf
• ROUGE

Preprocessing functions

• [ ] token classification
• [ ] question answering
• [ ] seq2seq