Sentiment Analysis: Past, Present and Future
Written by Claire Vishik, Trust In Digital Life
The interest in analyzing “sentiment” and making algorithmic conclusions from this analysis is as old as the idea of artificial intelligence (AI).
And, like AI, sentiment analysis has been defined and pursued - in several ways, leading to different conclusions. The very definition of “sentiment” itself depends heavily on the area where sentiment analysis is pursued. Medhat et al.[1] provided the following definition in 2014:
“Sentiment Analysis (SA) or Opinion Mining (OM) is the computational study of people’s opinions, attitudes and emotions toward an entity."
The entity represents individuals, events, or topics, and the definition requires for the human subjects to express the sentiment in a clear way prior to analysis. Hussein (2016)[2] reused a definition from 2015 stating that:
“Sentiment analysis (Basant et al., 2015[3]) uses the natural language processing (NLP), text analysis and computational techniques to automate the extraction or classification of sentiment from sentiment reviews”.
This definition relies on the automatic extraction of the elements of sentiment from various sources of information in different formats instead of relying on human subjects.
When sentiment analysis is used for marketing or opinion analysis, it is an efficiency engine, not a paradigm changer. In this case, greater understanding of a phenomenon could be obtained via collecting and processing existing utterances and other inputs, while also incorporating a useful and sometimes complex classification of sentiment or emotions, such as Plutchik’s pinwheel of emotions developed in 1980[4]. New pathways were found in advertising and marketing strategies based on a better understanding of the emotions driven by a product or a topic. Dozens of software packages exist to assist in linking emotions or a more general sentiment to an entity. Some are general and some domain-based, some require significant manual intervention via the creation of sentiment ontologies, and some are connected via APIs to AI-adapted large scale data repositories like Watson and ChatGPT.
With the advent of large-scale AI, there is a discussion about whether sentiment analysis could be a game changer and if it could lead to “sentiment-driven commerce.” Sentiment-driven commerce may be a direction allowing technologists to take advantage of large representative datasets without jeopardizing users’ privacy in order to create a paradigm where sophisticated sentiment analysis provides viable alternatives to technologies that are more invasive. These areas of application range from security assessment, dynamic pricing, technology roadmaps to decision support and personalized education.
As with other topics that our TDL blogs will consider, sentiment analysis includes multidisciplinary components that are highly technical as well as containing elements of psychology and policy considerations. We will continue to write about sentiment analysis and its applications, and we are interested in your opinion.
Do you think it is a rising area? What are the obvious applications of sentiment-driven commerce?
[1] Medhat, Walaa, Ahmed Hassan, and Hoda Korashy. "Sentiment analysis algorithms and applications: A survey." Ain Shams engineering journal 5.4 (2014): 1093-1113.
[2] Hussein, Doaa Mohey El-Din Mohamed. "A survey on sentiment analysis challenges." Journal of King Saud University-Engineering Sciences 30.4 (2018): 330-338.
[3] A. Basant, M. Namita, B. Pooja, Sonal Garg 2Sentiment Analysis Using Common-Sense and Context Information, Hindawi Publishing Corporation Computational Intelligence and Neuroscience (2015)
[4] R. Plutchik, 1980. A general psychoevolutionary theory of emotion, pp 3–33. Academic press, New York.