A Comparative Study of Using Bag-of-Words and Word-Embedding Attributes in the Spoiler Classification of English and Thai Text

Abstract

© 2020, Springer Nature Switzerland AG. This research compares the effectiveness of using traditional bag-of-words and word-embedding attributes to classify movie comments into spoiler or non-spoiler. Both approaches were applied to comments in English, an inflectional language; and in Thai, a non-inflectional language. Experimental results suggested that in terms of classification performance, word embedding was not clearly better than bag of words. Yet, a decision to choose it over bag of words could be due to its scalability. Between Word2Vec and FastText embeddings, the former was favorable when few out-of-vocabulary (OOV) words were present. Finally, although FastText was expected to be helpful with a large number of OOV words, its benefit was hardly seen for Thai language.