Research On Ontology And Rule Based System Model For Controlled Natural Language

Natural Language Processing (NLP) is an important branch and topic of artificial intelligence. Due to the characteristics of natural language itself, NLP can not be processed by the current computing technology as handled by human. Therefore, the application of natural language is limited. Controlled natural language (CNL) is a subset of NLP, which takes Linguistics, logic, classification theory of knowledge, psychology and information science, etc. as theoretical basis, reducing or eliminating the ambiguity and complexity of the language in a domain by restricting the vocabulary, grammar and meaning of natural language, and providing the application of natural language. Ontology is an important component of semantic web. Information is given well-defined semantics through ontology to facilitate works collaboratively between computers and humans. Rule is used to describe knowledge. Knowledge system based on rules can be implemented through logic programming. In this paper, in-depth study has been conducted to the controlled natural language, ontology and rule characteristics, and a model of controlled language system based on ontology and rules is proposed on from the study, which mainly includes the following aspects:Firstly, propose an ontology lexicon model of controlled language based on WordNet. WordNet is a semantic-oriented English Dictionary, and is similar to a traditional dictionary, and provides a rich level of semantic relations, including synonymous relations, antonym relations, and the context relationships, etc., which associate words and words together to form a word network. Ontology is a concept borrowed from philosophy category, providing formal language to capture relevant domain knowledge; and providing conventional understanding to the domain knowledge, and determining the common recognized words in a domain, and then giving a clear definition of relationships between the words from different levels of formal models. Thus, by integrating WordNet general lexicon and ontology domain lexicon, an ontology lexicon model of controlled natural language can be constructed.Secondly, propose an interpreter based on ontology lexicon. The interpreter has two components:Syntactic parser and Syntax tree interpreter. Syntactic parser, based on the theory of Probabilistic Lexicalized Context-Free Grammar, will parse the CNL sentences into the abstract syntax tree by combining WordNet, word part of speech defined by ontology domain lexicon, and the relationships between words. Syntax tree interpreter interprets abstract syntax tree into Discourse Representation Structure of an intermediate Expression Language, which describes semantic information of the sentences.Thirdly, propose a system model of controlled language reasoning based on ontology and rules. The logic foundation of this model is axioms of Semantic web OWL and SWRL. The controlled language sentences are partially made into DRS through language processing unit, and transformed into Semantic web OWL and rule knowledge through translation tools, and finally reasoning to knowledge is conducted by rules engine.Lastly, propose a user interface model of controlled language based on the ontology and rules. CNL can be used to describe the knowledge, but it is largely used as user interfaces in practice due to the limitation of using language to represent knowledge. The controlled language sentences are partially transformed into DRS through language processing unit, which contains the semantic information expressed in language; after analyzed, user input data can be gained from DRS; and then execution results will be returned after system execution actions triggered by rule reasoning.This paper, based on the theory of ontology and rules, puts forward two controlled natural language application models:one is a reasoning system based on ontology and rule, which conducts the reasoning by combining the axioms of Semantic Web ontology and rules; the other is an user interface system based on the ontology and rules, which executes the system actions and completes the business operations through analyzing the semantic information of controlled natural language. These two models basically implement the integration of ontology, rules and controlled natural language and provide the technology framework for the application of controlled language in domains. Experiments show that the proposed controlled natural language system based on ontology and rules in this paper is not only having the research value, but also has practical value. The study result in this paper could be served as the basis of reference and application for a related applied research in the controlled natural language area.