Open Access Open Access  Restricted Access Subscription Access

Semantic Based Smarter Natural Language Interface for Database


Affiliations
1 Department of Computer Science and Information Technology, Singhania University, Pacheri Bari, Rajasthan, India
 

The paper presents semantics based search paradigm to be embedded in Natural Language Interface (NLI) systems. The classical Information Retrieval (IR) models were based on lexical mapping and approximation based searches which suffered from obvious weaknesses as follows -

1. The queries used predefined lexical mapping or approximations and would skip any direct or indirect references via semantic alternatives. Homonymous lexemes can give many meanings leading to ambiguous queries and failed processes or ambiguous results if user is using the hyper word query. No intelligent mechanism is present in the NLI by which it will interpret the query.

2. When we write the query, then each lexeme gives only individual meaning of the word but lexemes are related to each other and produce a collocated meaning of the entire sentence. The classical IR model does not consider this aspect of IR.

To get over with these inadequacies in the classical IR mode, the NLI has to be made smarter with adequate semantic capabilities. Therefore we will provide the inferential capability to the existing NLI by providing the knowledge base to the system. This knowledge base will consist of the facts, concepts, synonymy, homonymy, hypernymy, discourse, and the contextual information and will help in generating appropriate and accurate results.


Keywords

NLI, IR, QLE, Polysemy, Homonymy, Hypernymy, Discourse.
User
Notifications
Font Size

Abstract Views: 165

PDF Views: 0




  • Semantic Based Smarter Natural Language Interface for Database

Abstract Views: 165  |  PDF Views: 0

Authors

Keshav Niranjan
Department of Computer Science and Information Technology, Singhania University, Pacheri Bari, Rajasthan, India

Abstract


The paper presents semantics based search paradigm to be embedded in Natural Language Interface (NLI) systems. The classical Information Retrieval (IR) models were based on lexical mapping and approximation based searches which suffered from obvious weaknesses as follows -

1. The queries used predefined lexical mapping or approximations and would skip any direct or indirect references via semantic alternatives. Homonymous lexemes can give many meanings leading to ambiguous queries and failed processes or ambiguous results if user is using the hyper word query. No intelligent mechanism is present in the NLI by which it will interpret the query.

2. When we write the query, then each lexeme gives only individual meaning of the word but lexemes are related to each other and produce a collocated meaning of the entire sentence. The classical IR model does not consider this aspect of IR.

To get over with these inadequacies in the classical IR mode, the NLI has to be made smarter with adequate semantic capabilities. Therefore we will provide the inferential capability to the existing NLI by providing the knowledge base to the system. This knowledge base will consist of the facts, concepts, synonymy, homonymy, hypernymy, discourse, and the contextual information and will help in generating appropriate and accurate results.


Keywords


NLI, IR, QLE, Polysemy, Homonymy, Hypernymy, Discourse.