Researchers have advanced the field of affective computing (AC) – the creation of computer systems that recognize, express and process human emotions – by proposing a new way to recognize emotion in text. Their development has significant potential for business applications.
In a world full of blog posts, tweets and emails, the implications for businesses able to identify emotions contained in written communications are clear: a greater understanding of their customers’ behaviours, motivations and satisfaction. By making use of AC in their ‘enterprise systems’, businesses that pay attention to ‘affective’ factors and build relationships with their customers may also obtain a competitive advantage. The current trend of ‘emotional marketing’ already acknowledges the important role emotions play in our decision-making processes. The research may also have implications for the analysis of human factors in other business functions and processes, such as supply chains.
Writing in the journal Enterprise Information Systems, Changqin Quan and Fuji Ren describe their approach – ‘multi-label textual emotion recognition’. It differs from other approaches to emotion recognition by taking into account the full emotional context of a sentence, rather than being purely ‘lexical’. Uniquely, Quan and Ren’s method allows its users to recognize indirect emotions, emotional ambiguity or multiple emotions in the subject text.
As they explain: ‘Our model generates an emotion vector for each emotional word in a sentence by analysing semantic, syntactic and contextual features. The emotion vector records basic emotions contained in the word.’ Each word is given an ‘emotional state’ represented by eight binary digits, each corresponding to one, or more, of eight key emotions: expectation, joy, love, surprise, anxiety, sorrow, anger and hate. The final ‘result’ is based on ‘the state of combined expressions of emotions’ in the sentence as a whole.
This article is a fascinating insight into an expanding area of research. It also lays a path for future research into the subject, including how speech and facial emotion analysis might also be thrown into the analytical mix. Marketing and customer service may never be the same again.