Grant: Engage Grants Program

In recent years, Twitter has become a valuable source of information regarding the public perception of variousproducts, services, and events. Some researchers have explored methods for extractingpositive/neutral/negative sentiment from such information and correlating this against market performance,lending credence to the idea that public opinion expressed via Twitter can be used as a surrogate for consumeror investor research. While such work illustrates the potential for mining sentiment from Twitter and using thisas an indicator for market performance, the approaches have yet to be shown to be useful for investmentanalysts.The goal of this research is to develop methods to extract sentiment information from Twitter, and present thisin a visual and interactive manner that supports investors in analyzing the public opinion in relation to one ormore specific topics of interest (queries). In preliminary research, a proof-of-concept approach called VisualTwitter Analytics (Vista) was developed to extract data from Twitter, perform simple sentiment analysis on thetweets, and visualize this information in a timeline representation. Taking a visual analytics approach to thisdata analysis problem allows users to visually identify trends within the data, perform exploratory analysis, andinspect the specific tweets as needed. This grant will support the further development and refinement of allaspects of Vista, focusing specifically on how the approach can be used to support and enhance investmentanalysis tasks. In particular, a database system will be developed to aid the long-term monitoring and storage oftweets matching user-specified queries, machine learning methods will be developed to perform sentimentanalysis on this data by taking into account domain-specific terminology, and the interface to Vista will berefined and enhanced to include new visual and interactive elements that enable exploration, comprehension,and analysis of the temporally changing sentiment.