A Semantic Approach to “Local Search” Based on User Generated Content

If you are looking for a local restaurant recommendation, you are likely to end up on one of the many "restaurant review" sites on the Web. Popular restaurants have large numbers of user reviews and the not so popular ones have fewer or none. If you were looking in San Francisco, you'd likely end up with too many reviews and options to choose from and almost no user reviews to make a decision on if you were looking for a place in Gilroy, CA (50 miles south of San Francisco). Local businesses are fragmented by nature, and so is the content. BooRah (as in "boo" and hur"rah") takes a semantic approach to solving this fragmentation problem. BooRah's platform consists of a semantic crawler and a sentiment analyzer (natural language based).

The sentiment extraction technology processes vast amounts of plain English text from user generated content such as user reviews, blogs and professional reviews. It extracts specific user sentiments expressed within the content and generates scores, summaries and customizable search phrases relevant to consumers. The system performs sophisticated semantic and structural analysis on every sentence and identifies specific adjectives, verbs and nouns that help in quantifying the sentiment. A summary roll-up of all positive sentiments (Rahs) and negative sentiments (Boos) is provided to the consumer. Such a system prevents users from unduly influencing ratings since it actually analyzes written content. Ontology to support such a system is built using some publicly available Resource Definition Frameworks (RDF) and home grown knowledge databases.

Using the BooRah approach, if Bob and 100 other people talked about the excellent Pad Thai in a SF Bay Area Thai restaurant, the system can feature this restaurant when users are looking for Pad Thai. This is quite different from the more traditionally and widely used keyword search approach. Additionally, the system can generate related recommendations based on collaborative and semantic inference algorithms. This enables the system to generate nearby recommendations based on preference for "Food," "Service" and "Ambiance."

BooRah is currently in beta for restaurants in San Francisco, New York and Los Angeles metro areas and available at http://www.boorah.com.