Opinion Mining and Text Analytics of Literary Reader Responses: A Case Study of Reader Responses to KL Noir Volumes in Goodreads Using Sentiment Analysis and Topic

Introduction

Nowadays, large amounts of data are stored in intranets, internet, and databases. Some examples of electronic data that are available are the customer feedbacks and communications on social media platforms. As millions of users in social media platforms, or any other websites, are now able to share information, the amount of data produced daily is very large. The question is how this massive amount of data is handled and stored, as well as what medium or tools are being used. This is where text mining comes into light.

As the data quantity is rapidly increasing day by day, the amount of data that needs to be collected for many purposes has similarly increased. Using text mining to extract useful data from natural language texts is a new field that needs to be explored. Text mining is an important field of study that has proved beneficial for scholars of various disciplines. Literary scholars use text mining to examine the data produced by creative writers, literary readers, publishers and distributing companies (Ying et al., 2021). Text mining is a process of analyzing text in order to extract useful information for a specific purpose. Text is unstructured, ambiguous, and difficult to process when compared to the type of data stored in databases as text is also considered as one of the most suitable ways for the information changing (Kalra, 2013). Of this, the textual data is the data type that the text mining always deals with because analyzing the data with the text mining can identify the new insights. The goal of text mining is to discover new information that has not yet been found and written by anyone else. As the text mining is a relatively new field that aims to extract useful data from natural language text, numerous studies and research had been conducted (Malik et al., 2021; Ying et al., 2022). In recent years, social media has become a powerful data source in many ways such as an analytical platform to explore certain objectives.

To develop more understanding in social media mining, this paper aims to analyze the readers’ responses in Goodreads website to discover the attitude of the users toward certain books by extracting the book's information such as their rates and reviews. Goodreads is a platform where users can use its database of books, annotations, quotes, and reviews to conduct searches. To achieve the aim of this paper, we will use the four volumes of KL Noir that has been assigned to gain the trends and analyze users’ attitude towards the books. The books are anthologies that explore the darker side of Malaysian capital city, Kuala Lumpur. Cities and places have always intrigued literary writers, and in recent years, literary scholars have shown increasing interest in the study of settings and places in creative writings (Asl, 2021, 2022). By using text mining techniques, the aim of the first process is to recognize key phrases and relationships. It accomplishes this by searching for predefined text sequences, a process known as pattern matching. Next, key topics and words that appear in the same paragraphs or sentences are used to extract the main keywords. Then, using text mining techniques like topic modelling and others, the characteristics and frequency of the words are defined and analyzed. Sentiment analysis is also applied to measure the attitude of the users toward the review of the book which they describe in a text. Next, to scrap the reviews from Goodreads, we need to setup the Python environment in the Jupyter Notebook by installing required Python packages that we need to use such as pandas and selenium. Next, to gain insight, we will explore the text and study the frequent words using a bar chart and wordcloud. Prior to that, the text must be normalized and the process of tokenization, noise removal, stopwords removal and lemmatization will be done. After that, topic modelling model is developed. The goal is to investigate the major topics that have been discussed by the users toward the books. The topic modelling output is a series of keyword lists that have been classified with the coherence scores using Latent Dirichlet Allocation (LDA) (Jelodar et al., 2019) and Latent Semantic Indexing (LSI) (Adomavicius & Tuzhilin, 2005) algorithms. The next phase needs to contain the works of Vader algorithm and textblob from the sentiment analysis. There are many ways of classifying texts, using machine learning techniques. Sentiment analysis is one of the techniques where a system identifies whether the text being analyzed is positive, negative, or neutral based on its topics.