Loading...
A PROCESS-BASED SEARCH ENGINE
Liu, Yaling
Liu, Yaling
Citations
Altmetric:
Abstract
This dissertation introduces a novel approach to solving a certain type of Web search problem. The concept of Process-Based Web search is proposed and a prototype Process-Based Search Engine has been developed and evaluated to help users solve problems more effectively. Unlike traditional Web search engines showing a list of ranked hyperlinks as results, the proposed system searches and displays results based on processes. Therefore, users do not have to read, understand, and then search further to find all the information that they are seeking. In addition, different from business process systems, the proposed Web search engine is designed to be available to general public to solve everyday problems related to processes. It is based on the processes not having to be strictly defined to meet pre-requisite of applications such as business process execution engines. The process-based search engine extends traditional Web search engines, referred to as query-based search engines. The system integrates with an existing query-based search engine to provide process-based search, if necessary. The system is comprised of two sub-systems: the Process Base, and the Process-Based Search Engine. The Process Base provides knowledge of processes to the search engine. The Process Base is populated using both manual and automatic approaches. The manual approach includes a Web interface that is built based on the philosophy of Wiki. All Web users have access to retrieve, create, and update process data in the Process Base. The automatic approach crawls and extracts process data from the Web. The Process-Based Search Engine rewrites an original query into a sequence of sub-queries using the information in the Process Base. These sub-queries are then sent to a query-based search engine to perform Web search. Results from multiple queries are integrated and displayed in the format of process breakdown in one result page. This helps the user in gaining comprehensive understanding of the answer to the initial query. The system is implemented using Java programming language, specifically Java SE Development Kit (JDK), Servlets, and JavaServer Pages (JSP) by Sun Microsystems. Tomcat and Apache were used to build the Web servers. The server side of the system can run both on Linux or Microsoft Windows operating system. Common functionalities, such as input data validation, are implemented at client side using JavaScript, for efficiency. Users use a Web browser to access search services, the same as common Web search engines, such as Google. Process data are stored in a relational database to facilitate efficient management. Google, the Web search engine, is used as the query-based search engine. The Process-Based Search Engine invokes Google SOAP APIs to perform the query-based search. The system has been released on the University of Kansas Intranet for the experimental evaluation. A total of 19 human subjects participated in the evaluation. The Process-Based search engine was compared with related systems, including the MIT process handbook, eHow, wikiHow, and a query-based search engine. Experimental results have shown that the Process-Based Search Engine improves search quality for some types of complex queries, compared to traditional Web search systems.
Description
Date
2009-01-01
Journal Title
Journal ISSN
Volume Title
Publisher
University of Kansas
Collections
Archive Status
This item contains archived web content.
Files
Liu_ku_0099D_10502_DATA_1.pdf
Adobe PDF, 1.21 MB
- Embargoed until 2160-05-31
Research Projects
Organizational Units
Journal Issue
Keywords
Computer Science, process-based search, Web mining