SBIR Phase II: Quantifying Consumer Rationale Expressed in Free Text Online Discussions

This Small Business Innovation Research Phase II project aims at creating novel methods for analyzing consumer behavior and product acceptance. This project will develop efficient, automated natural language processing (NLP) methods that perform at Web scale to analyze a broad spectrum of consumer experiences with products, as reported in conversations on social media, in consumer reviews, etc. Currently available representations of consumer commentary are relatively shallow. In this project more detailed representations will be built, advancing NLP technology and delivering state-of-the-art analysis of conversations that detail consumer decisions, motivations, and questions about products. In addition, the representations of consumer commentary will be integrated with third-party data, resulting in the development of predictive models of consumer behavior that exceed the accuracy of current models. This will allow product vendors to quantify the influencing factors behind product adoption and attrition, which is crucial to proving marketing hypotheses, predicting business outcomes, and improving customer awareness of product issues. The broader impact/commercial potential of this project lies in the ability to improve the analysis and use of consumers' perception and experience of products. Deeper analysis of millions of consumer experiences reported online will help the public, manufacturers, marketers, and regulators alike make better decisions about products. Currently, gaining representative insights about a product requires manual processing of an overwhelming range of information sources. Automatic, in-depth analysis of these disparate sources at Web scale will deliver an unbiased, authoritative stream of quantified product information and consumer opinion, improving consumer experience, product performance, and industry transparency. The richness of the resulting representation will enable an understanding of finer-grained aspects of product performance and consumer behavior, which cannot be extracted or analyzed using existing techniques such as sentiment analysis or topic-modeling. The technology developed in this project will translate to fundamental improvements in the conduct of market research, commercial campaigns and regulatory investigations, as well as enhancing the public's ability to make informed decisions about products.