An Ontology Based Framework for Web Service Discovery from Natural Language User Queries

摘要

This study focus on providing a user friendly natural language query interface for web service discovery and maps it with both traditional and semantic web service description languages like WSDL, OWL-S etc and thus aiming to bridge the gap between various service description formats, by using a common registry schema. Our approach considers user requirements regarding input, output and purpose of web service and performs matches across all the three parameters on a web registry. The proposed framework uses an ontology based web service registry which integrates various web service related information like service description, classification, search keywords. The query processor performs basic NLP processing like stemming and tagging on the user query and converts it to an intermediate representation based on description logics which we call WS-QDL. The user query is then classified using classification ontology to limit the search process within applicable domains. The final preprocessing step optimizes the search process by replacing the query terms with the most related terms in the domain ontology. The search algorithm separates input, output and subject specifications from the user query and performs matching based on each criterion. The tests were performed based on OWL-TC3 dataset and the obtained results shows improvement in performance measures.