Natural language processing of text customer ratings in the banking sector.

The paper compares two approaches to the texts analysis of the online customer ratings. The first approach focuses on a simple statistical analysis of the mentioned words number in ratings with a score from 1 to 5. The second approach applies the Word2Vec algorithm for the text analysis. The first approach shows the most popular words found in ratings. The second approach illustrates non-obvious words that can also be useful in the textual analysis of customer ratings. As a result, the authors came to the following generalization. The first approach forms the latent factor, i.e., "professionalism" (associated with the interaction of clients with a staff). It is presented in ratings as a negative factor ("non-professionalism"). An "organizational" (associated with time and some actions) latent factor is formed on the basis of the second approach (Word2Vec). [ABSTRACT FROM AUTHOR]