Research Of Semantic Retrieval Methods Based On Conceptual Graphs

Traditional retrieval methods are based on the words matching. Due to the lack of semantic information and a single search tool, the existence of search results is often monotonous and the recall rate and precision rate on such issues are lack of precision. Therefore, the search for finding a new retrieval method become into a hot spot. The semantic retrieval is a system or way that is based on obtaining the semantic data or information and make the results of semantic search to meet the needs though clear semantic expression and processing. For the application of semantic retrieval, selecting a good tool for knowledge representation is essential.The conceptual graph is a said graphics which has the capacity to express the first order predicate logic. Compared to other knowledge representation tools, the conceptual graph is not only with a more direct mapping between natural languages but also with a more intuitive form of expression, a higher ability of the expression and a more simply reasoning. Because of these factors, conceptual graphs are applied to the field of artificial intelligence in all aspects, gradually. In response to the problem, this paper carries out a research of semantic retrieval based on conceptual graphs deeply. Chose conceptual graphs as a tool for semantic knowledge representation, created a semantic retrieval model and improved the matching process.In this paper, completed the following tasks:(1) Conceptual graphs as a tool for semantic knowledge representation are applied to the semantic retrieval, and designed and implemented the transformation from text into conceptual graphs.Compared with other knowledge representations, conceptual graphs are chosen as a tool for semantic knowledge representation because they have much higher ability of direct mapping to natural languages. At first, the text is turned into a syntax diagram by parsing. Then, using the table of the concept of interdependence with the basic relationship between the map corresponding, syntax graphs are turned into conceptual graphs. Sowa said that the conceptual graphs can show as a linear map, which inspired the plan that conceptual graphs can be expressed as trees. In this paper, a new matching method to the conceptual graphs is taken: merged the nodes which have the same sub node. After this adoption, the expression of concept graphs are much clearer and the calculations are simpler. It is useful for the foundation of giving weight of every node in the process of graphs matching. Through the above steps, information of resources are described in the form of conceptual graphs and stored into the database. In the query process, the query that formed of conceptual graphs for matching, compared with graphs of resources, and then retrieved the results.(2) Improved the process of the conceptual graphs'matching, to bring it closer to human'thought.The matching of conceptual graphs includes four steps: the calculation of the similarity between words, the calculation of the similarity between relations, the calculation of node weights and the calculation of the conceptual graphs similarity. For the calculation of the words'similarity, "How Net" is used for calculating the similarity between words by the distance of the concepts. And for the calculation of the similarity between relations, this article provides that the same relationship between the nodes is 1, and the different shall be 0. Psychological research shows that a certain part of the query description provides more details, then this part of the characteristics of the matching take a greater importance. Based on this, different relationships have different numbers of child nodes - it means that they describe the details of how much be given. In this paper, a rule is given that the node which has more brothers has higher weight than others, which emphasizes the parts have some more details. This distribution of weight cuts down the distance between the machine matched and the human mind. Finally, this article defines a formula of the conceptual graphs for calculating the similarity between the query graphs and the resource graphs. (3) The realization of a semantic retrieval model based on the conceptual graphs.In this work, based on the eventual realization of this paper, a semantic search model achieves to show the conceptual graph of query text and match the conceptual graphs and retrieve query results eventually. In order to compare and verify the validity of this model, another model which bases on the matching of keywords is built. The two models are used in the field of secondary education. The results of the test show that the semantic retrieval model based on the conceptual graphs is better than that of keywords'in the rate of recall rate and precision rate.Research of semantic retrieval, not only of great theoretical value, but also has broad application prospects. In this paper, this issue has been studied and discussed, a semantic retrieval model based on the conceptual graphs is realized, and by comparing the experiments to prove its effectiveness. How to make computers more accurate understanding of natural languages, how to achieve more accurate matching, how to provide users with personalized semantic recommendation, will serve as a continuation of the future work in this article.