Structured Queries Over Web Text.
Michael Cafarella, Oren Etzioni, Dan Mircea Suciu · 2006
Abstract The Web contains a vast amount of text that can only be queried using simple keywords-in, documents-out search queries. But Web text often contains structured elements, such as hotel location and price pairs embedded in a set of hotel reviews. Queries that process these structural text elements wouldbe much more powerful than our current document-centric queries. Of course, text does not contain metadata or a schema, making it unclear what a structured text query means precisely. In this paperwe describe three possible models for structured queries over text, each of which implies different query semantics and user interaction. 1 Introduction The Web contains a vast amount of data, most of it in the form of unstructured text. Search engines are thestandard tools for querying this text, and generally perform just one type of query. In response to a few keywords, a search engine will return a relevance-ranked list of documents. Unfortunately, treating Web text as nothingbut a collection of standalone documents ignores a substantial amount of embedded structure that is obvious to every human reader, even if current search engines are blind to it. Web text often has a rich, though implicit,structure that deserves a correspondingly-rich set of query tools. For example, consider a website that contains hotel reviews. Although each review is a self-containedtext that is authored by a single person, the set of all such reviews shows remarkable regularity in the type of information that is contained. Most reviews will contain standard values such as the hotel's price and generalquality. Reviews of urban hotels might discuss how central the hotel's location is, and reviews of resorts will mention the quality of the beach and pool. In other words, the hotel reviews have a messy and implicit "schema"that is not designed by any database administrator but is nonetheless present.