In recent years, data analysis has been widely applied in many different domains. Text data plays an important role in prediction of various insights. The data produced in the form of user reviews, satisfactory forms, movie reviews, after sales product reviews, and similar kinds of representations serve as inputs for textual data analysis. In previous years, however, companies relied on paper-based satisfactory surveys, agent reports, etc. for business development or product outreach development purposes. As these methods involve human intervention, there is always a very high chance of false inputs. Hence, the development of computational intelligence-based strategies such as textual and sentimental analysis have been of enormous help for such companies. Automated tools and software have helped various business organizations and firms to develop their business, find their faults and bugs, and relieve themselves from their limitations. This chapter discusses the basics of textual analysis, approaches for textual analysis, as well as the tools, solutions, and some limitations of applying textual analysis.
TopIntroduction
Text mining or analysis generally refers to the extraction of data or information from large volume of texts. Indeed, the extracted information will possess quality content which can be used for different purposes. Text mining or text analysis can be applied where there is a large volume of text content. It is noteworthy that text does not only mean books, manuscripts, and theses. Text in all forms that appear in various platforms such as internet, social media, messengers, mails etc. can also be used for text mining and analytics. In recent years, different organizations, firms, industries, and even small-scale business units have felt the growing need to collect and analyze customers' feedback in order to improve their services and productions. These kinds of inputs or data fed by the end user or customer offer genuine insights as they are given from users' perspective. Before the advancement in computational intelligence, the data on customer satisfaction or feedback were obtained in a paper format. As anyone could fill those forms, authenticity of the collected data was always a major constraint for the organizations to ensure the quality of the product/services. In some cases, based on customer’s thinking the interest towards products are distinguished. For instance, when a person is happy and he/she searches for a product and if he/she finds a suitable one, he/she would opt for purchasing the product. Under the same scenario in a mixed mindset, he/she would search for reviews, dialup friends or relatives who own a similar product and ask for their experience using that product. Under the same scenario if a person is under depression or sad, even though the product is good and excellent in the market, he/she would not opt for purchasing it. Thus, everything is based on the situation and emotions of the end user which makes them to decide what to do next, whether to purchase or not. Thus customer’s emotional incidence and customer feedback are very important for various decisions in business intelligence and product marketing. As customers decision may suddenly change for no reason, it is difficult to predict the exact outcome of a product based on surveys (Pang & Lee, 2008). Textual analysis gathers information from unstructured sources for the further perusal. User given input information plays a very important role in such textual analysis. In the current decade, people utilize ‘internet’ to express their thoughts and feelings. Various talk forums, social media hubs, blogspot(s), groups, etc. serve as platform for this purpose (Liu, 2010). This can be considered as general discussion forums where they share their opinions and feelings about a particular topic or a current happening. In the same way when it comes into a product or application, the inputs given by public impacts the real time sales of a ‘product’ or ‘application’. In e-commerce platforms or commodity selling sites, people search with ‘keywords’ pertaining to the product they look for. For instance, name of the ‘brand’, ‘capacity’, ’manufacturer’, ‘warranty’ etc are the common keywords in terms of ‘product’. ‘Vendor’, ‘version’, ‘validity’ etc can be ‘keywords’ for an application. On searching with the keyword people will be shown search results of various brands, of various varieties and capacities. People’s first noticeable constraint will be the ‘reviews’ of such product. In the world of e-commerce websites, previous customers' ‘reviews’ of a product play an important role in its sales. While negative reviews may discourage potential customers' from purchasing the product, positive reviews may have the opposite effect. Thus, the textual analysis plays a very important role in business analytics and sales prediction strategies. In this chapter, textual analysis will be carried out in 3 broader steps: The ‘Document level’, the ‘sentence level’ and the ‘feature level’. The ‘document’ and ‘sentence’ levels of analysis perform the sentimental opinion about the products. The third level performs the process of ‘features’ based on the product. In the process of evaluation or analysis, each review or comment is considered to be a ‘document’ (Godara & Kumar, 2019). There are various open-source tools that are used for text analytics which predict and perform genuine response detection from the text inputs from various end users. These tools work on the basis of machine learning, deep learning-based algorithms and extensions and they reflect as the genuine source of extracting customer’s sentiments and satisfaction measures. The Artificial Intelligence powered tools provide a broad space for analytics of data fed as input. This chapter discusses various such tools and their functionality as a review and further narrates how machine learning based systems have implications on text and emotional analytics and its future extensions.