Word Vector Space for Text Classification and Prediction According to Author
Abstract
Word embedings are intense vector representations of words. They capture the semantic order between words. In our study, we propose a new method for classifying text by author, using word embeds and author vector space. Using each author's books, we trained the data to create author-specific word vector space. Word vector spaces consist of vector representations of all the words used by the author in the training dataset. In our study, different word vector spaces were created for different authors. The aim here is answering the question "Which author is more likely to write any given text?". In a text, we propose a method to find out whether consecutive word vectors belong to that author's vector space. Our method is based on a basic principle used when constructing word embedding vectors. Extensive results on the data sets show that the proposed method performs better than the latest technology methods in terms of accuracy.
Keywords
Cite this article
Dönmez, İ., Pashaei , E., & Pashaei , E. (2019). Word Vector Space for Text Classification and Prediction According to Author. In International Congress on Human-Computer Interaction, Optimization and Robotic Applications (pp. 106–111). SET Technology. https://doi.org/10.36287/setsci.4.5.021
Article History
- Published
- 12 Oct 2019
License
© 2019 SET Technology. This article is distributed under the terms of the CC BY 4.0.