Cognitive computing approaches for human activity recognition from tweets—a case study of twitter marketing campaign

Abstract

Big data analytics is a growing area of research and has in the recent years introduced many applications for business sector to create value from data. Cognitive computing provides new approaches for analysing social media data, e.g. text analytics and machine vision, available via APIs and software libraries. This study focuses on cognitive computing approaches for recognizing and categorizing valence, arousal and human activity from tweets. The aim is to identify human activities from tweets and map these activities on pleasure and arousal dimensions using cognitive computing services in order to derive new insights for business. A case study of Twitter marketing campaign of a company providing stress management services is presented. In the marketing campaign participants submitted images and text on events that help them to recover from stress. These tweets are analysed with cognitive computing services to recognize activities that participants of the campaign performed to recover from stress and these are compared against qualitative analysis. The caption recognition was successful at detecting some of the activities of the users and overall results indicate that many participants cope with stress by e.g. performing outdoor activities. The two sentiment detection algorithms were also successful in detecting consumer sentiments from the text component of the tweets. Finally, implications for business are drawn from the results of analysis.