Thesis
Deceptive review detection using convolutional neural networks
Washington State University
Master of Science (MS), Washington State University
07/2019
DOI:
https://doi.org/10.7273/000004019
Handle:
https://hdl.handle.net/2376/124900
Abstract
Consumers prefer to consider reviews as a second thought before making online purchases. This gives a hidden opportunity for the sellers to manipulate them for obvious purposes, and hence online deceptive review detection is an ongoing research topic. However, it is difficult to obtain labeled data for deceptive reviews as identifying deceptive reviews requires human intervention, and even then it is a difficult task to identify deceptive reviews. As machine learning classifiers require large amount labeled data for training, semi-supervised approaches help in using the limited amount of label data to label the unlabeled data in order to have sufficient amount of data for training. Convolutional neural networks have produced excellent results for sentence classification with pre-trained vectors on multiple benchmarks. Thus, in this thesis we implement two semi-supervised algorithms, Positive Unlabeled (PU) algorithm and Co-training algorithm with the use of Convolutional Neural Networks (CNN). The techniques discussed in this thesis improve the state of the art performance for deceptive review detection to an F-score of 90.73.
Metrics
5 File views/ downloads
39 Record Views
Details
- Title
- Deceptive review detection using convolutional neural networks
- Creators
- Shubham Purushottam Adep
- Contributors
- Scott Wallace (Advisor) - Washington State University, School of Engineering and Computer Science (VANC)
- Awarding Institution
- Washington State University
- Academic Unit
- School of Engineering and Computer Science (VANC)
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Identifiers
- 99900890796101842
- Language
- English
- Resource Type
- Thesis