Topical Evolution and Regional Affinity of Tweets

Abstract

Business organizations are increasingly showing interest in Twitter content to know their consumers. Tracking popular tags and trends give some idea about what people are talking about. However, in order to act on the knowledge acquired, they need more detailed information like regional variability in content, exact location of discontent if any, regional affinities and influences etc. In this work, we present methods to identify topics of discussion in tweets using a LDA-based approach, which can identify emerging or evolving topics. Regional analysis of topics can provide interesting business insights about consumer expectation or behavioural variations. Further, regional distribution of topics are analysed to identify clusters of regions that tend to behave similarly over extended periods of time.