Research On Semantic Cue Discovery Based On Random Walking In Open Domain Knowledge Network

Open domain knowledge network provides abundant structure, Internet, knowledge base for people to query, it is becoming an important foundation to support intelligent development in the information age. People can achieve large-scale knowledge more accurate semantic analysis and understanding, will also become one of the new ways of semantic analysis. However, there are many bottleneck problems in the realization of accurate semantic analysis of the knowledge network, which contains the massive knowledge and the large-scale and complicated knowledge link between the open domain knowledge network.Research in the field of semantic analysis has made great progress, there are a variety of excellent performance semantic correlation degree calculation. Currently, semantic dictionaries and corpus-based semantic relevance algorithm faced with low background knowledge coverage, high maintenance costs of the problem, and corpus-based open network domain knowledge in semantic inadequate refining of the background knowledge base. In addition, the above two methods to calculate semantic relevance feedback can only be a single numerical scale semantic intensity, text information semantic association between the depth of excavation.This dissertation analyzes the background knowledge of the characteristics of an open domain knowledge networks, semantic dictionary, corpus, etc. From construction costs, maintenance costs, organizational structure, knowledge coverage, analysis of future development perspectives open network compared to the semantic domain knowledge dictionary, massive Corpus and other traditional advantages of background knowledge on the semantic analysis application, to build a network concept to prepare. And as a basis to construct a weighted network model called the concept of the network, to determine the weight of their upper value information and text data according to the structure of knowledge networks. Semantic clue discovering has been a significant issue for mining the form of correlation among concepts or things(Knowledge Entities). The correlation is seen as an important reference for improving information retrieval, knowledge reasoning or making decisions. At present, most searches on semantics focused on semantic relatedness computing rather than semantic clues discovering. To discover semantic clues, we propose a novel semantic clues discovering method, named as Semantic Pheromone Walking(SPW) that utilizes the strategies of pheromone in ant colony algorithm and random walk algorithm to discover close semantic clues among knowledge entities. Our proposed SPW comprises two aspects: concept network construction and semantic clue discovering. Firstly, constructing a weighted network model named Concept Network(CN), the link weights of which are determined in accordance with the hyperlink structure and semantic relatedness of text data in ODKN contains abundant semantic information. Then, a concept network based Semantic Pheromone Walking method is addressed to discover semantic clues between knowledge entities by using Semantic Pheromone(SP) which is a digital signal reflecting the compactness of concept correlation, as heuristic information.Experimental results show that: the human cognitive information contained in the knowledge network can meet the need of the exploration of correlation form between things and our solution could find reasonable semantic clues based it.