| Knowledge representation is the basis of artificial intelligence. The more reasonable and effective the knowledge representation method is, the better it can support all kinds of applications of artificial intelligence. Therefore, knowledge representation has been a research focus in the field of the artificial intelligence.The world consists of various objects. The object-oriented thought is the most natural way that the human understand the world. Based on the study of the various existing knowledge representation methods, this paper proposed a new method of object-oriented knowledge representation, gave the conceptual model of the new method, the basic framework and the realization method based on Java language. And it built a small knowledge library based on the new object-oriented knowledge representation method to verify the rationality and validity of the new method.In order to verify the validity of the new method, this paper applied the established knowledge library to the applications of the natural language processing, including the word similarity calculation and the sentence similarity calculation. Word similarity calculation is the basis of sentence similarity calculation. The most of existing Chinese word similarity algorithms were built on the basis of the existing semantic dictionary such as How Net. Based on the characteristics of the object-oriented knowledge library, this paper proposed a new word similarity algorithm, and applied it to the sentence similarity algorithm.In the calculation of sentence similarity, place name words have special meaning. But the existing sentence similarity algorithms can not reflect the importance of the place name to the sentence and the differences between the place names, so the calculation results are not very reasonable. To address this problem, this paper proposed a sentence similarity algorithm involving place names combining with the new object-oriented knowledge representationmethod and compared with the existing methods.In this paper, the main work and achievements are as follows:Firstly, it introduces some related concepts about knowledge representation. It mainly introduces the present research on knowledge representation and knowledge library, and analyzes the deficiencies existing in all kinds of knowledge representation methods.Secondly, this dissertation emphatically introduces and analyzes the research status of the object-oriented knowledge representation, and proposes a new method of the object-oriented knowledge representation, and established a small knowledge library based on this method for its application of natural language processing.Finally, this paper applied the new object-oriented knowledge representation method to the word similarity computation and sentence similarity calculation. After analyzing and comparing several typical algorithms, it puts forward a new word similarity calculation method. This paper established a geographic knowledge library based on the new object-oriented knowledge representation method. In addition to use the subordinate structure information and attribute information of the place name objects to calculate the place names similarity, it carried on the special treatment of the place names when calculating the sentence similarity. The place name words and the non-place name words were given different weights to reflect the importance of place name words. Then a complete sentence similarity method involved in place names in combination with the sentence structure similarity is proposed.This dissertation made the comparative experiments between the word similarity algorithm and the sentence similarity algorithm based on the new knowledge representation method and other mainstream algorithms based on How Net to verify the effectiveness of the new object-oriented knowledge representation method. |
PDF Full Text Request