Social Media Cross-Source and Cross-Domain Sentiment Classification

Paola Zola,Eugenio Brentari,Paulo Cortez,Costantino Ragno

doi:10.1142/s0219622019500305

Abstract

Due to the expansion of Internet and Web 2.0 phenomenon, there is a growing interest in sentiment analysis of freely opinionated text. In this paper, we propose a novel cross-source cross-domain sentiment classification, in which cross-domain-labeled Web sources (Amazon and Tripadvisor) are used to train supervised learning models (including two deep learning algorithms) that are tested on typically nonlabeled social media reviews (Facebook and Twitter). We explored a three-step methodology, in which distinct balanced training, text preprocessing and machine learning methods were tested, using two languages: English and Italian. The best results were achieved using undersampling training and a Convolutional Neural Network. Interesting cross-source classification performances were achieved, in particular when using Amazon and Tripadvisor reviews to train a model that is tested on Facebook data for both English and Italian.

Highlights

Technological advances, such as the Internet expansion, Web 2.0 phenomenon and massive mobile device adoption, have increased the availability of freely opinionated text
As for the learning algorithm, we propose a word embedded Convolutional Neural Network, which is compared with another deep learning model (Deep Feedforward Network) and two other classifiers (Support Vector Machines and Naive Bayes)
We address a novel cross-source cross-domain sentiment analysis, in which Web sources that contain easy labeled reviews (Amazon and Tripadvisor) are used to fit a sentiment analysis model, which is reused to predict the sentiment of two typically unlabeled social media platforms (Facebook and Twitter)

Summary

Introduction

Technological advances, such as the Internet expansion, Web 2.0 phenomenon and massive mobile device adoption, have increased the availability of freely opinionated text (e.g., blog reviews, social network comments) This big data source of unstructured texts enriches the value of sentiment analysis, termed opinion mining, which uses computational methods to automatically analyze human opinions, sentiments and evaluations towards entities (e.g., products, services, organizations).[1] several studies have analyzed opinion dynamics in social networks and their potential impact in decision making.[2,3,4] sentiment analysis is a key tool of modern decision support systems, helping to support decisions in several real-world applications, such as involving hotels,[5] stock markets,[6,7] and tra c accidents.[8]. These two issues can be handled by using a cross-domain sentiment analysis,[10,11] which is a recent transfer learning research trend that aims to reuse sentiment models, previously fitted to some domains (e.g., electronics), to predict the sentiment of texts from other domains (e.g., books)

Objectives

Methods

Results

Discussion

Conclusion