Chinese Keyword Search Based On Knowledge Base Over Relational Databases

Inspired by the theory and technology in the field of Information Retrieval and Web Search Engine, free-style keyword search over relational database have been the new research focus of information processing. The methods of keyword search are suitable to Web database, and they need not users to know the database schema and the structured query language. The sorting methods and processing algorithms are key issues in keyword search. The existed keyword research technologies usually focus on English keyword search, in this thesis we will discuss Chinese keyword search in a relational database. We propose a new method to realize free-style Chinese keyword search over relational databases, and to avoid the problem of Chinese Word Segmentation and improved the searching accuracy.There are many differences between English and Chinese about grammar and basic information unit. It's difficult to directly apply the technologies of English keyword search to the Chinese keyword search. The main ideas of the method in this thesis are below: Firstly, a knowledge-base is built to store the related information of attributes and tuple word in a database, then a index will be created based on the knowledge-base, which be used to match the query word and the tuple word one by one to obtain the set of identifiers of candidate tuples. Secondly, according to the information of the tuple identifiers set, we design a ranking strategy to computer the similarity between the query and the candidate tuples, and sorted the tuples by the similarity. Finally, we adjust the similarity by phase-based method for Chinese keyword search and output the top-N result. A prototype is developed and experimental results on a large real dataset indicate that the performance of the method is efficient and effective.