Sentiment analysis is one of the most important applications in the field of text mining. It computes people's opinions, comments, posts, reviews, evaluations, and emotions which are expressed on products, sales, services, individuals, organizations, etc. Nowadays, large amounts of structured and unstructured data are being produced on the web. The categorizing and grouping of these data become a real-world problem. In this chapter, the authors address the current research in this field, issues and the problem of sentiment analysis on Big Data for classification and clustering. It suggests new methods, applications, algorithm extensions of classification and clustering and software tools in the field of sentiment analysis.
TopProblem Of Sentiment Analysis On Big Data
Big data creates new challenges on data processing, data storage, data representation, pattern mining, visualization, etc (Gema Bello-Orgaz et al., 2016) in the field of data mining, machine learning, natural language processing, text mining, social networks, and sentiment analysis. The rapid growth of unstructured data in social networking, blogs, reviews, posts, comments, and tweets are the most important source for sentiment analysis (Changli Zhang et al., 2008). Further, sentiment analysis problems focus on different levels namely Document level, Sentence level, and aspect level (R Feldman, 2013). The document-level sentiment analyzes a piece of information. It represents multiple opinions, but not a single opinion view. Document level sentiment analysis hides insights and useful information by reducing a whole document into a single opinion. Sentence level expresses overall opinions of each sentence. Aspect level exactly expresses what people likes or dislikes. For instance, a single review of BMW 1 Series car is analyzed as shown in Figure 1. The sentences are numbered in parenthesis.
Figure 1. A single review sentiment detection at different levels
The classification and clustering techniques are required to provide meaningful information for those polarity data. The data can be classified and clustered into positive, negative, and neutral polarities. Sentiment analysis refers to a classification problem due to the prediction of polarity words such as positive, negative and neutral. The accuracy, data size, data sparsity, and sarcasm are the main issues in sentiment mining classification and clustering. Therefore, sentiment classification aims to mine the written documents (comments, posts, reviews, tweets, etc.) about a products or services and classifying the documents into positive or negative opinions (Qiang Ye et al., 2014, Yanghui Rao et al., 2014). Sentiment clustering aims to group the written documents. The general framework for sentiment analysis on Bigdata is shown in Figure 2.
Figure 2. A general framework for sentiment analysis on Bigdata