Machine Intelligence for Health Information: Capturing Concepts & Trends in Social Media via Query Expansion

Abstract

Introduction. We aim to improve retrieval of health information from Twitter. Background. The popularity of social media and micro-blogs has emphasised their potential for knowledge discovery and trend building. However, capturing and relating concepts in these short-spoken and lexically extensive sources of information requires search engines with increasing intelligence. Methods. Our approach uses query expansion techniques to associate query terms with the most similar Twitter terms to capture trends in the gamut of information. Results. We demonstrated the value, defined as improved precision, of our search engine by considering three search tasks and two independent annotators. We also showed the stability of the engine with an increasing number of tweets; this is crucial as large data sets are needed for capturing trends with high confidence. These results encourage us to continue developing the engine for discovering trends in health information available at Twitter.