Sequence Package Analysis: A New Natural Language Method for Mining User-Generated Content for Mobile Uses

A. Neustein and J.A. Markowitz (eds.), Mobile Speech and Advanced Natural Language Solutions, DOI 10.1007/978-1-4614-6018-3_5, © Springer Science+Business Media New York 2013 Abstract Paradoxically, in an era when cyber-postings proliferate on the Web, much of the valuable information that can be mined from user-generated content (UGC) still eludes most mining programs. One reason this massive amount of UGC is, for all practical purposes, “lost” in cyberspace has to do with the limitations inherent in existing approaches to natural language understanding. In this chapter, I will explore how Sequence Package Analysis (SPA), a new natural-language datamining method for text-based and spoken natural-language input, locates and extracts valuable opinion-related data buried in online postings—and makes such data available to mobile users. The SPA mining method can be used with existing SLM systems to assist in both supervised and unsupervised training. This chapter demonstrates that the advantage of SPA in such contexts is twofold: First, by breaking down unconstrained, open-ended natural-language input into relevant sequence packages, SPA can streamline the process of classifying a vast number of sentences (or spoken utterances); second, as the SPA algorithms become more robust, the process of collecting and classifying natural-language input can be automated entirely, thereby replacing human annotators with SPA-designed machine-learning. Using several examples, randomly selected from the TripAdvisor website, I illustrate how SPA can render the hidden attributes of online reviews (both positive and negative) more visible to the mobile user.