Detecting positive and negative deceptive opinions using PU-learning

A novel method is proposed that with respect to its original version is much more conservative at the moment of selecting the negative examples from the unlabeled ones and consistently outperformed the original PU-learning approach in the detection of positive and negative deceptive opinions respectively.

Abstract

[EN] Nowadays a large number of opinion reviews are posted on the Web. Such reviews are a
\nvery important source of information for customers and companies. The former rely more
\nthan ever on online reviews to make their purchase decisions, and the latter to respond
\npromptly to their clients’ expectations. Unfortunately, due to the business that is behind,
\nthere is an increasing number of deceptive opinions, that is, fictitious opinions that have
\nbeen deliberately written to sound authentic, in order to deceive the consumers
\npromoting a low quality product (positive deceptive opinions) or criticizing a potentially
\ngood quality one (negative deceptive opinions). In this paper we focus on the detection of
\nboth types of deceptive opinions, positive and negative. Due to the scarcity of examples of
\ndeceptive opinions, we propose to approach the problem of the detection of deceptive
\nopinions employing PU-learning. PU-learning is a semi-supervised technique for building
\na binary classifier on the basis of positive (i.e., deceptive opinions) and unlabeled
\nexamples only. Concretely, we propose a novel method that with respect to its original
\nversion is much more conservative at the moment of selecting the negative examples
\n(i.e., not deceptive opinions) from the unlabeled ones. The obtained results show that
\nthe proposed PU-learning method consistently outperformed the original PU-learning
\napproach. In particular, results show an average improvement of 8.2% and 1.6% over the
\noriginal approach in the detection of positive and negative deceptive opinions
\nrespectively.
\n 2014 Elsevier Ltd. All rights reserved.