| Natural language is the main tool of human communication,and the purpose of human communication is to exchange knowledge.Knowledge is the understanding and description of consciousness,so the study of natural language should start from the representation of knowledge,and then study how knowledge is transformed into the form of natural language that human beings can understand.Traditional knowledge representation methods cannot express the characteristics of self-consciousness,mutual representation,fuzziness and dynamics of human intelligence,but the AORBCO model established on the basis of Consciousness-Only Psychology can express multi-modal knowledge uniformly.In order to enable human beings to understand and interact with the knowledge expressed by AORBCO model,this paper mainly studies the conversion technology between AORBCO model description language and natural language from the following aspects:1.The design and implementation of word Knowledge Base in AORBCO Model: In view of the incomplete and unclear description of knowledge by existing knowledge representation technologies,the machine learning model cannot effectively utilize multi-modal knowledge.This paper studies natural language from the essence of knowledge,and analyzes the knowledge representation method of AORBCO model.It embodies mutual representation in the form of links,fuzziness in the weight of relations between concepts,and dynamics in the process of operation and evolution,and integrates it in Ego to express self-consciousness.It proposes a general expression of word knowledge and takes it as the basic part of subsequent research.The experimental results show that the natural language lexical and grammatical structure libraries proposed in this paper can obtain multi-granularity knowledge representation without losing the search efficiency.2.Research on natural language understanding in AORBCO model: Mainly studies the process,principle and technology of converting natural language into AORBCO model description language.The process mainly includes four stages: text processing,sentence processing,semantic understanding and knowledge representation.In the text processing stage,AORBCO model is completed to perceive and preprocess natural language,and Ego is divided from natural language perceived in the concrete world according to the definition of meaning paragraph in Chinese linguistics.In the statement processing stage,the words of the statement are divided.Combining with the characteristics of mutual table in the knowledge representation of AORBCO model,the processing method of DFS-FMM algorithm combined with the structure of two-word hash index lexicon is proposed,which can solve the uniqueness of the result of the original FMM algorithm and reduce the performance problem of searching the description language of AORBCO model.Semantic accuracy selection of lexical analysis results is completed in the semantic understanding stage.Knowledge blocks with mutual tabularity in the AORBCO model are regarded as weighted directed graphs,but there are negative weights in the directed graphs.Therefore,this paper optimizes SPFA algorithm from the perspective of limiting the number of nodes joining the team,so that it can maintain the same time complexity as the original algorithm.The negative circle digraph can also be calculated.In the knowledge representation stage,the AORBCO model description language representation of the semantic understanding result is completed.For non-existent knowledge blocks,it initializes them and establishes the relation and weight with other knowledge blocks;for existing knowledge blocks,it updates the relation weight with other knowledge blocks.3.Research on natural language generation in AORBCO model: Mainly studies the process,principle and technology of AORBCO model describing language conversion into natural language.The process mainly includes five stages: theme determination,knowledge block determination,knowledge block tailoring,sentence generation and text generation.In the topic determination stage,the topic words in the generated text are determined.The generation of natural language by Ego depends on the desire generated after communicating with acquaintances.By using the desire of customer acquaintance modeling to describe the context,the problem of context in the generative dialogue field of natural language processing technology is solved.In the stage of determining knowledge blocks,the problem of candidate knowledge blocks related to subject words when generating statements is solved.In this paper,the Page Rank algorithm is improved through the characteristics of mutual tabularity and fuzziness in the model,so as to form a calculation method that takes the entity in the AORBCO model as a directed weighted graph to obtain candidate knowledge blocks.In the knowledge block tailoring stage,the uniqueness problem of the selected knowledge is solved.By combining the above two stages of knowledge selection in the statement generation,this paper considers the link entity factor in the model,the weight of the common link object and the class factor to calculate the knowledge similarity of abstract words,and puts forward the weighted summation similarity calculation formula of overlap degree.In the statement implementation and text implementation stage,knowledge is organized according to the AORBCO model description language hierarchy,and finally forms the description form of natural language.4.Application Research of Natural Language Understanding in Intelligent WEB System Based on AORBCO Model: Aiming at the problems that customer Ego lacks information understanding and mesh Ego cannot organize content in the intelligent WEB system of the previous AORBCO model,natural language processing capability of AORBCO model is introduced to form an intelligent WEB system that can understand customer needs,provide personalized search and organize website content.At the same time,it also verifies the feasibility of the above AORBCO model natural language understanding research results. |
