Using Ontologies for information retrieval

 

M S Thesis, by R Sandhya, July 2008.

 

Meaningful information retrieval has been an important and interesting research concern

in the IR community. Here, the natural emphasis is on retrieving highly relevant

information for the given query. Existing keyword based search engines suffer from

the problems of retrieving the documents only in the presence of the keyword (along

with the statistically computed correlated words), and thus ignoring the documents

which do not contain the keyword although semantically relevant and related to the

query. On the other hand some of the retrieved documents that contain the query

terms may not be relevant. In this context, the user is sometimes overloaded with

irrelevant information and sometimes at a loss of data. Being the conceptual models

that capture domain knowledge, ontologies can be looked upon for aiding meaningful

information retrieval. There have been several efforts towards using the domain

knowledge contained in the ontologies for the said purpose.

 

This thesis explores the use of domain specific ontologies in information retrieval.

In this thesis, we present a query expansion mechanism which deploys the knowledge

of a domain represented in the form of an OWL Ontology to improve relevancy of

the retrieved documents. The proposed system fits the query terms in the ontology

graph in an appropriate way and exploits the surrounding knowledge to enhance the

query. The resulting enhanced query is given to the underlying basic keyword search

system. We also propose a way of ranking the results from the basic keyword search

system by using the domain ontology. The experiments we have conducted proved

a significant increase in the recall and precision when compared to that of keyword

search systems.