Fuzzy rule based unsupervised sentiment analysis from social media posts

Highlights

• A fuzzy rule-based approach is proposed for sentiment analysis of tweets.

• Formulation of nine fuzzy rules to compute sentiment of each tweet.

• The proposed unsupervised approach is suitable for any sentiment lexicon.

• The approach can be adapted to both bipolar and tripolar sentiment analysis tasks.

Abstract

In this paper, we compute the sentiment of social media posts using a novel set of fuzzy rules involving multiple lexicons and datasets. The proposed fuzzy system integrates Natural Language Processing techniques and Word Sense Disambiguation using a novel unsupervised nine fuzzy rule based system to classify the post into: positive, negative or neutral sentiment class. We perform a comparative analysis of our method on nine public twitter datasets, three sentiment lexicons, four state-of-the-art approaches for unsupervised Sentiment Analysis and one state-of-the-art method for supervised machine learning. Traditionally, Sentiment Analysis of twitter data is performed using a single lexicon. Our results can give an insight to researchers to choose which lexicon is best for social media. The fusion of fuzzy logic with lexicons for sentiment classification provides a new paradigm in Sentiment Analysis. Our method can be adapted to any lexicon and any dataset (two-class or three-class sentiment). The experiments on benchmark datasets yield higher performance for our approach as compared to the state-of-the-art.

Introduction

Sentiment Analysis is a challenging research problem especially on social media. Users can freely express their views, opinions and feelings on different trending events, topics, etc. via social media posts. These posts need to be analysed to know what sentiment is conveyed through these posts. Sentiment Analysis, also referred as emotion AI, involves analyzing views from the written text so as to understand and gauge human emotions. The social media allows world-wide users to connect and interact with each other and express the opinions on general topics. Social Sentiment Analysis can be used to improve customer service and marketing and also serves as a measure of social media performance. In recent years, the impact of social media websites on daily life has become so considerable that even information on large and small incidents or disasters is gathered via social media sites. The users portray not only the content about events but also their feelings (Yoo, Song, & Jeong, 2018). The automated extraction of sentiment from these posts and classifying them into different polarities–positive, negative or neutral– has received extensive attention from researchers during the past decade.

Twitter is one of the popular social media and boasts of a respectful 255 million active monthly users. Some of the challenges in analysing tweets are: use of informal language, short forms, abbreviations, heavy use of emoticons and slangs. Twitter, also known as microblogging, has limited size of tweets that makes it difficult to compute the polarity. In this paper, we apply fuzzy rule-based unsupervised approach to process the tweets in such a way as to overcome the above challenges. We have implemented our approach on multiple public twitter datasets using multiple lexicons. The proposed fuzzy rule-based approach can compute sentiment for two-class and three-class sentiment datasets. Two-class datasets have only positive and negative sentiment while three-class have neutral sentiment as well.

Fuzzy logic is an extension of deterministic logic, i.e. the truth value has range from 0 to 1 rather than a binary value. The primary aim of the theory of fuzzy logic is turning a black and white problem into a grey problem (Zadeh, 2015). In the field of artificial intelligence, possibly the easiest way to represent the human knowledge is to transform it into natural language expressions in the format of IF-THEN rules. These rules are based on natural language representations and models, which are themselves based on fuzzy sets and fuzzy logic (Ross, 2010). Classification systems based on fuzzy rules are powerful and acknowledged tools for pattern recognition and classification. These systems can handle uncertainty, ambiguity or vagueness in a very efficient way due to the presence of fuzziness (López, Río, Benítez, & Herrera, 2015). We have used the concept of fuzzy rule-based system to create our own nine fuzzy rules to determine the sentiment of each tweet.

The main contributions of this paper are: i) formulation of nine fuzzy rules to compute sentiment of each tweet ii) the proposed unsupervised approach is suitable for any sentiment lexicon iii) also suitable for any dataset (two-class or three-class) iv) comparison of our proposed rule-based approach for Sentiment Analysis with four state-of-the-art methods for unsupervised sentiment classification and one state-of-the-art method for supervised machine learning. The rest of the paper is organized as follows. Section 2 describes the state-of-the-art on Sentiment Analysis from social media, while our proposed fuzzy rule-based system is presented in Section 3. Section 4 is about the experimental setup & implementation. Results are discussed in Section 5. The overall conclusions are drawn in Section 6.