# Spam Review Detection Using Deep Learning

**Authors:** G. M. Shahariar, Swapnil Biswas, Faiza Omar, Faisal Muhammad Shah, Samiha Binte Hassan  
**arXiv:** [2211.01675](https://arxiv.org/abs/2211.01675)  
**Date:** 2022-11-03  
**Summary page:** https://www.summarizepaper.com/en/paper/2211.01675v1/

## Key points

- Online shopping is popular due to its convenience, but fake reviews can mislead customers.
- A reliable system for detecting spam reviews is needed.
- Machine learning techniques have been introduced to solve the problem of spam review detection.
- Traditional machine learning classifiers such as Nave Bayes (NB), K Nearest Neighbor (KNN), and Support Vector Machine (SVM) have been applied to detect spam reviews.
- Deep learning methods such as Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) have also been proposed for spam review detection.
- Previous studies attempted to mine and summarize all customer reviews of a product using natural language processing methods.
- Some researchers incorporated sentiment analysis or added psycholinguistic features in their models to improve performance in detecting fake or spam reviews.
- A hybrid approach was proposed that detected duplicate reviews first before creating a hybrid dataset with the help of active learning.
- Various CNN architectures composed of Topic Categorization tasks and Sentiment Analysis on various classification datasets were evaluated by researchers who achieved very good performance.
- In the first phase of the proposed model, a dataset of gold-standard deceptive opinion spam was produced using crowdsourcing through Amazon Mechanical Turk.
- Detecting spam reviews remains a critical issue in making online reviews reliable.

## Summary

In today's world, online shopping has become a regular thing for most people as it saves time and effort. However, one of the drawbacks of online purchasing is the prevalence of fake reviews or spam reviews that can mislead customers into making wrong decisions. To address this issue, a robust and reliable system for detecting spam reviews is needed. Prominent machine learning techniques have been introduced to solve the problem of spam review detection. The majority of current research has concentrated on supervised learning methods, which require labeled data - an inadequacy when it comes to online review. In this article, the focus is on detecting any deceptive text reviews using both labeled and unlabeled data. To achieve this goal, deep learning methods such as Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and a variant of Recurrent Neural Network (RNN) called Long Short-Term Memory (LSTM) have been proposed for spam review detection. Traditional machine learning classifiers such as Nave Bayes (NB), K Nearest Neighbor (KNN), and Support Vector Machine (SVM) have also been applied to detect spam reviews. Finally, performance comparison for both traditional and deep learning classifiers has been shown. Previous studies have attempted to mine and summarize all customer reviews of a product using natural language processing methods. Some authors classified spam reviews into three categories: non-reviews, brand-only reviews, and untruthful reviews while others used supervised learning and manually labeled reviews crawled from Epinions to detect product review spam. In addition to these approaches, some researchers incorporated sentiment analysis or added psycholinguistic features in their models to improve performance in detecting fake or spam reviews. A hybrid approach was also proposed that detected duplicate reviews first before creating a hybrid dataset with the help of active learning. Various CNN architectures composed of Topic Categorization tasks and Sentiment Analysis on various classification datasets were evaluated by researchers who achieved very good performance. Semantic clustering was introduced by adding an additional layer in the CNN architecture, and an efficient bag-of-words representation for input data was used to reduce the number of parameters for the network. In the first phase of the proposed model, a dataset of gold-standard deceptive opinion spam was produced using crowdsourcing through Amazon Mechanical Turk. Although part-of-speech n-gram features give a fairly good prediction on whether an individual review is fake, the classifier actually performed slightly better when psycholinguistic features were added to the model. Overall, detecting spam reviews remains a critical issue in making online reviews reliable.