III-COR-Small: Beyond Keyword Search: Enabling Diverse Structured Query Paradigms over Text Databases

点击翻译按钮获取中文摘要

The text available on the Web and beyond embeds unprecedented volumesof valuable structured data, "hidden" in natural language. Forexample, a news article might discuss an outbreak of an infectiousdisease, reporting the name of the disease, the number of peopleaffected, and the geographical regions involved. Keyword search, theprevalent query paradigm for text, is often insufficiently expressivefor complex information needs that require structured data embedded intext. For such needs, users (e.g., an epidemiologist compilingstatistics, as reported in the media, on recent foodborne diseaseoutbreaks in a remote country) are forced to embark in labor-intensivecycles of keyword-based document retrieval and manual documentfiltering, until they locate the appropriate (structured) information.To move beyond keyword search, this project exploits informationextraction technology, which identifies structured data in text, toenable structured querying. To capture diverse user information needsand depart from a "one-size-fits-all" querying approach, which isinappropriate for this extraction-based scenario, this projectexplores a wealth of structured query paradigms: sometimes users(e.g., a high-school student in need of some quick examples andstatistics for a report on recent salmonella outbreaks in developingcountries) are after a few exploratory results, which should bereturned fast; some other times, users (e.g., the above epidemiologistinvestigating foodborne diseases) are after comprehensive results, forwhich waiting a longer time is acceptable. The project developsspecialized cost-based query optimizers for each query paradigm,accounting for the efficiency and, critically, the result quality ofthe query execution plans. The technology produced will assist a vastrange of users and information needs, by enabling efficient, diverseinteractions with text databases -- for sophisticated searching anddata mining -- that are cumbersome or impossible with today'stechnology. The research and educational components of the projectwill rely on -- and encourage -- a tight integration of threecomplementary Computer Science disciplines, namely, natural languageprocessing, information retrieval, and databases. The project willalso provide data sets and source code, for experimentation andevaluation, to the community at large over the Web (http://extraction.cs.columbia.edu/).