A comparative empirical study on social media sentiment analysis over various genres and languages

Abstract

People express their opinions about things like products, celebrities and services using social media channels. The analysis of these textual contents for sentiments is a gold mine for marketing experts as well as for research in humanities, thus automatic sentiment analysis is a popular area of applied artificial intelligence. The chief objective of this paper is to investigate automatic sentiment analysis on social media contents over various text sources and languages. The comparative findings of the investigation may give useful insights to artificial intelligence researchers who develop sentiment analyzers for a new textual source. To achieve this, we describe supervised machine learning based systems which perform sentiment analysis and we comparatively evaluate them on seven publicly available English and Hungarian databases, which contain text documents taken from Twitter and product review sites. We discuss the differences among these text genres and languages in terms of document-and target-level sentiment analysis. Keywords Natural language processing • Sentiment analysis • Data mining 1 Introduction Recently, the popularity of social media has increased. People post messages on a variety of topics, like products and political issues and a large amount of user generated data is created in textual form. Several applications have been developed for exploiting the knowledge and information present in user generated content, for example in predicting the results of elections (Sang and Bos, 2012), monitoring brands (Jansen et al, 2009) and disaster management (Varga et al, 2013). In this study, we focus on sentiment analysis (SA) whose task is to assign polarity labels (positive, negative and neutral) to textual elements.

A inteligência artificial na comunicação corporativa

“…Um dos principais gargalos desse tipo de análise é a classificação das postagens dos usuários em termos de sentimento em relação à marca (positivo, negativo ou neutro). Embora muitos softwares façam a classificação e coleta de forma automática, o processo está longe de ser perfeito 7 e necessita de uma revisão constante por parte de seres humanos 8 , principalmente quando trabalhamos com línguas que não possuem uma base de dados tão grande 9 , como é o caso do português 10 . Na BITES, toda a classificação é feita por analistas humanos, mas isso demanda grande volume de trabalho, o que, em momentos de grande uso das redes (proximidade de eventos importantes, crises, avaliação de campanhas recém-lançadas), pode levar um tempo, que é precioso para o cliente.…”

The Role of Artificial Intelligence in Community Planning

“…AI is already used by social media platforms to control the creation of content, suggest friends you may know and want to connect with, and determine the most relevant advertisements to show individual users (Kietzmann et al 2018). AI developed by the planning system has the potential to support streamlining communication with individuals on basic planning issues through chatbots, and support rapid sentiment analysis of social media posts relevant to a specific development proposal (Araujo 2018;Hangya and Farkas 2017).…”

A survey on sentiment analysis methods, applications, and challenges

“…In the work of Balaji et al (2021) conducted a thorough examination of the several applications of social media analysis utilizing sophisticated machine learning algorithms. Authors present a brief overview of machine learning algorithms used in social media analysis (Hangya and Farkas 2017).…”