Research On The Method Of Interface For Natural Language Understanding Oriented Agriculture Expert System

With the widespread access to the information technology, people have got an increasing need for knowledge service, which is on the basis of the support of a large-scale knowledge bank. It must severely influence the capability and quality of knowledge service by only depending on mass human labor to construct the large-scale knowledge bank. Knowledge is the basis of intelligence, and most of man's existing knowledge is carried in texts. Consequently, it has been one of the tough problems in the field of knowledge engineering to enable computers to acquire knowledge automatically from texts. Correspondingly, the present urgency is to find out a generally used text knowledge acquiring method.Knowledge acquiring is an extremely difficult problem in exploring intelligence system. It mainly involves: experts in this field, materials, texts, database and knowledge engineer. However, the major knowledge source of most projects is experts in this field. In light of the current progress of knowledge acquiring research, it is far from enough on the text knowledge acquiring research, i.e. acquiring knowledge directly from texts with text literature as knowledge source. Text knowledge is an essential source of knowledge; acquiring knowledge directly from texts is closer to human knowledge acquiring method and hence has clearer advantage and prospects.In this paper we apply case grammar to organize the sentence structure, carry on knowledge representation with the semantic network, and put forward a method that helps acquire knowledge automatically from texts and turn into corresponding knowledge representation, solving the problem of constructing IUC(Interface of Understanding Chinese)in the expert system. The main efforts of this paper are as follows:(1)Download agricultural materials of about 500,000 words from the internet, then remove all markers to build up a corpora of Chinese agriculture.(2)Consult a large amount of literature to know about the current knowledge acquiring and knowledge representation methods.(3)Mark the participles and part of speech of the texts of the Chinese agricultural database and construct a case grammar graph of all the sentences containing verbs.(4)According to the constructed case grammar graph, put forward the method and rules to transform it semantic network, and form corresponding semantic network.(5)Extract and contract the rules by human and automatic means relatively.Finally, according to the method proposed above, we design and realize a knowledge acquiring system based on natural language. At the same time, we point out some problems existing in the model after analyzing the false examples in the testing results.By testing the system with 500,000 agricultural data, 225 reasoning rules are automatically extracted, among which 171 are effective; the theoretical result should be 242 rules. In accordance, effective rules cover almost 76.0% of the extracted total, and almost 70.7% of the theoretical one.