I-ANSWER: Intelligent Application of NLP and Semantics for Web Extracted Reporting

Abstract

Making web meaningful for computer will unleash possibilities beyond the horizon of present knowledge according to Tim Berners-Lee. While making data process able for computers we forgot that humans still require information from web. So it has reduced the effective utilization of information creating a gap between normal user and semantic web. In order to bridge that gap we have devised a trainable natural language engine based on template matching to convert the human language into an appropriate query language to extract results from ontologically formatted data. A natural language interface is built for naïve users to utilize the web effectively. It relieves the users to know the structure of ontology and offers a much more convenient and flexible interaction. Study of existing systems show that research on this issue is currently a hot issue. The researchers have emphasis on increasing the accuracy of the system without taking into account the issues like scalability, portability, domain independence and enhance sense. Our architecture caters both accuracy and issues concerning to adaptability and scalability. This system acts as a middleware between user and semantic web. It works intelligently by following our devised Rules which are not formally available for SPARQL parsing. Templates and rules can be added in the system at any time in order to make it scalable. The issue of domain independence is resolved by supplying a trainer in the system in order to make it adaptable to new ontologies introduced at some later stage. To evaluate the system 10 users from different domains were gathered who put 432 queries to the system in natural language. The system gave 84.5% accuracy. The accuracy level is very near to the systems previously developed like of QUERIX and GINSENG, in addition to the resolution of main architectural issues like domain and platform independence. From evaluation we have concluded that system needs a good training module which is under development and it needs more expansion because at this point the system does not handle queries involving complex graphs. In short I-Answer is one step forward in this domain. I-Answer main focus was to cater for architectural issues and query processing. It has not only solves the domain independence problem to a great extent but has also achieved accuracy level equal to previously developed systems.