VIthulan / travel-text-classification

Text classification

Text classififaction for question-answering on Travel domain is analysed here for 5000 questions. Here I have evaluated 5 methods.

1. Five Text features
2. FastText word embedding
3. LSTM
4. BERT

Summary

Model	Average accuracy (of 10 Folds)
SVM Classifier with 5 Text features	77.34%
SVM Classifier with only Lemmatized BoW	82%
SVM with FastText word embedding	82.6%
LSTM	81%
BERT	84.46%