Abstract
Electronic word of mouth (e-WOM) is rapidly becoming an empowering tool for consumers to express their experiences on services or products, on social media or other platforms. Beyond the obvious implications of such content to potential consumers, interest is also high among researchers, industry players, and other stakeholders who strive to analyze before-and-after sales expectations, emotions, and perceptions of customers. The need to find efficient ways of extracting and then analyzing online content rendered the reuse of tools and methodologies initially applied in other fields as well as the development of new approaches. In this chapter, the authors identify high-impact scientific work related to e-WOM and point out the analytical methods for analyzing e-WOM content. Furthermore, this chapter refers to the most relevant studies employing such methods and their findings. More specifically, it discusses clustering, sentiment analysis, supervised and unsupervised machine learning, lexicon-based approaches, corpus-based approach, summarization and predicting, and regression analysis.
TopIntroduction
Extensive use of social media (SM), defined as “a group of Internet-based applications that exist on the Web 2.0 platform and enable the Internet users from all over the world to interact, communicate, and share ideas, content, thoughts, experiences, perspectives, information, and relationships” (Perez-Aranda et al. 2017, pp. 53), lead to phenomenal increase of user-generated, online content. According to Munar and Jacobsen (2014), online content is mainly generated through wikis, blogs, microblogs, media-sharing sites, review and voting sites, and social media platforms.
A special subcategory of online content is the electronic word of mouth (eWOM), which can be traced to various channels as mention in Bickart (2005) and summarized below:
- 1.
Posted reviews, which are consumer opinions uploaded in public or private view by merchant websites, consumer review portals, or personal webpages like blogs
- 2.
Mailbags, which include comments posted online by customers and readers on news and magazines websites, and manufacturers and service providers pages.
- 3.
Discussion forums, which consist of bulletin, message and discussion boards or Usenet groups.
- 4.
Listservs, which consist of consumer opinions shared through group member email lists.
- 5.
Personal emails. These are shared opinions directly from one person to one or more other, through electronic mail.
- 6.
Chat rooms, which are areas on the internet or other computer networks where people and groups have real-time conversations.
- 7.
Instant messaging, where people converse online in real-time, usually one-to-one.
The first three communicating forms (posted reviews, mailbags, and discussion forums) are maintained online for longer periods of time; as a result, they are accessible by a potentially higher number of readers (Bickart 2005). Listservs, personal emails, chat rooms and instant messaging, are not publically available and therefore are accessible by far fewer users (Bickart 2005). The period a message is online and the number of readers who access it are two of factors that contribute to the impact of the message onto its readers. Other factors may also affect this process. Understanding, influencing or predicting human behavior through SM and eWOM analysis, is based on handling structured or unstructured user-generated content and derive value through its transformation into meaningful knowledge.
The focus of this chapter is to identify and discuss the main methods for analyzing eWOM content. To achieve this goal, we must systematically derive all relevant literature, as described in the following section.
TopEwom: Systematic Literature Review
In this section, we discuss the systematic literature review on impactful resources published up until October 2018. The systematic literature review focuses on regarding analysis methods of eWOM data. Based primarily on the methodology of Shaikh and Karjaluoto (2015) and the review approach of Sarma and Choudhury (2015), this endeavor start with identification of the keywords that will be used in our base search queries. The search terms include eWOM, online reviews, electronic word of mouth, online word of mouth, online consumer reviews. Then, we establish the inclusion criteria of our literature: we focus on resources published in English, peer-reviewed journals that are Scopus indexed. Based on the aforementioned criteria, the following search query has been formulated; it resulted in 1876 articles.
TITLE (((ewom) OR (word-of-mouth) OR (online AND reviews) OR (”online word of mouth”) OR (”online consumer reviews”))) AND (LIMIT-TO (SRCTYPE, “j”)) AND (LIMIT-TO (DOCTYPE, “ar”)) AND (LIMIT-TO (LANGUAGE, “English”))
Key Terms in this Chapter
Machine Learning: Machine learning is the application of algorithms and statistical models in order to perform a specific task automatically using artificial intelligence instead of explicitly providing instructions to the algorithm or model.
Cluster Analysis: Cluster analysis is the process of dividing data into groups (clusters), each object in a group has more similarities than the objects in other groups.
Ensembles Learning Technique: Ensemble learning technique is a machine learning technique that combines different algorithms in order to obtain improved predictive performance.
Topic Models: Topic models are statistical models that identify the main abstract themes (topics) that occur in corpus and discover hidden patterns in text.
High Dimensionality: High dimensionality is the phenomenon where the available information has so many dimensions that analysis becomes extremely difficult.
Classification Analysis: Classification analysis is the supervised process of assigning items to categories/classes in order improve the accuracy of our analysis.
Sentiment Mining: Sentiment mining or opinion mining is the process of identifying, extracting and analyzing text in order to identify emotional tone in the information gathered.