Opinion Mining for Decision Making
thesisposted on 2019-08-06, 00:00 authored by Paolo Polimeno Camastra
This work is concerned with Text Clustering, Sentiment Classification and Transfer Learning in Sentiment Classification. We choose to work in the domain of online reviews because they are suitable for our purposes: the dataset comprises reviews’ texts and labels as well as various categories of products. Labels are needed in order to perform the Supervised Learning part of our work, i.e. Sentiment Classification, while the presence of different product categories enable us to have different domains, such as Electronics, Movies, Clothing, etc., which is a requirement for Transfer Learning. Our main contribution in Text Clustering is in terms of comparison between clusterings obtained after different text vectorizations: tf-idf+Singular Value Decomposition (SVD), LDA and word2vec.