Query-index co-optimization executing query templates for complex text search

Nowadays, many complex text search systems, such as Entity Search or Topic Search, have been proposed to allow users to retrieve fine granularity units (e.g., entities or topics) inside documents directly. As those search systems target on more complex search tasks, the traditional query processing method purely based on an inverted index can not execute those search queries efficiently. New execution algorithms and index structures need to be proposed. In this paper, we study the problem of automatically deriving an efficient execution algorithm and indexes that support the algorithm for those systems. We take a relational view of the problem and model it as optimizing a query template with views. This query template optimization problem raises new challenges including \emph{enumerating plans with views} and \emph{selecting plans for answering a template} for a query optimizer. We present a novel optimization framework with a new set of transformation rules and an efficient selection strategy to deal with those two challenges. We systematically evaluate our framework in two concrete application settings. Experiments show that: (1) The derived algorithm and indexes significantly improve the efficiency the keyword-based baseline method. (2) Our framework can automatically derive plans and indexes that are manually optimized for a system. (3) Our approach is general enough to be applied to different search systems.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2011

Author dc:creator

• Li, Rui

Contributors dc:contributor

• Chang, Kevin C-C.

• Information Search
• Query Optimization

Statement dc:rights

• Copyright 2011 Rui Li

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/24508

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/24508