Sentiment Analysis and Emotion Detection
Understanding how customers like products as well as how technologies are perceived by operators is key for the success of manufacturing products and processes. Sentiment Analysis and Emotion Detection are one of the main exciting features developed within Artificial Intelligent (AI) laboratories that can be exploited for equipping computers and robots with human-like sentiment understanding. These technologies applied to the manufacturing domain can help to quickly detect the satisfaction level of operators and users and, therefore, to promptly make an action either to solve an issue or to support activities.
Modern implementations behind Sentiment Analysis and Emotion Detection are built on top of Transformers, deep learning models that have millions of features and that describe how words are used in speech or text. They can deeply understand how a sentence should be interpreted according to human-like understanding and suggest a score which represents the positiveness or negativeness of human expressions. Also, they can detect these scores for several characteristics of a product that users are talking about. For example, in a review about a product like “My car has a very good engine and brakes, however, the windows are quite small and limit the visibility” transformers might be used for doing the so-called Aspect Sentiment Analysis where first aspects are detected (for example “engine”, “breaks”, “windows”), and for each of them a sentiment score is associated. For example, “engine” and “brakes” can have a score of 0.8 associated with a transformer indicating that the customer is satisfied, while a score of 0.2 can be assigned to the “windows” to indicate that the customer is not satisfied with them. In doing so, a company can identify which actions have to be taken to improve the underlying product. At scale, companies can analyse, investigate, and explore large amounts of opinions expressed by customers and, hence, be able to evaluate the level of satisfaction of product utilizers. In the long term, detecting and solving issues in the designing and manufacturing of products as well as the identification of pitfalls in manufacturing processes will increase the customers’ satisfaction and will support the building of the customers’ loyalty.
As a consequence, Sentiment Analysis and Emotion Detection are two key technologies to be leveraged within several domains including manufacturing to enhance the companies-customers relationships, thus making a step ahead toward the user satisfaction and win-win situations. Many of the AI systems and components used in STAR require interaction with humans and machines, these interactions can benefit from Sentiment Analysis and Emotion Detection, to improve the communication with the end-users, but also in some internal use cases in which operators and machine collaborate, allowing to detect issues and recommend solutions.