Natural Language Representation Of Association Rules Based On Conceptual Graph

With the advent of the age of big data, the data is accumulating increasingly.Meanwhile, the data mining technology develops gradually, which has been widely used in many fields. The mining of association rules is an important part in data mining, as well as one of the main research directions. Its goal is to find some implicit, unknown or potential relationship, so as to find valuable information for the users. Therefore, how to express the mined association rules to the users is worthy of consideration.In recent years, as an effective way of expression, conceptual graphs have been applied to multiple fields. Its powerful representation of knowledge can interpret the abstract association rules into more concrete concept and conceptual relationship, as well as fully express the meaning of natural language, thus realize the generation of natural language.Conceptual graphs bring the users and association rules closer by taking advantage of people’s sensitivity to the image. Nevertheless, its means of expression still remains abstract,especially when the association rules are complex. And it is difficult for the ordinary users to understand. With the persistent development and wide application of the mining of association rules, the focus is how to generate the association rules into the way ordinary people can understand. As for the ordinary users, it is the easiest way to understand the association rules by the expression of natural language. It can enhance the necessity to the generating system of natural language at the same time. From the perspective of the current theory and technology, the high qualified natural language generation system is a long-term goal. But for the specific field, the system which can generate natural language can be realized. This thesis combines the conceptual graphs with the technology of generating the natural language to achieve the visual expression of association rules. The natural language makes the experts and the ordinary people understand the result of expression.The main content and innovation points of this thesis are as follows:(1) Propose the expression of association rules based on the conceptual graphs. Using the powerful expression of conceptual graphs and generation technology of natural language to interpret the association rules as a natural language. Firstly, identify the conceptand concept note. The main step is the predicate matching, this part is to match the predicate in the association rules with the predicate already existing in the domain knowledge base to determine the part of speech of each predicate of the association rule and carry on the annotation; the other step is to determine the relationship, which brings the predicate in the association rules to the domain knowledge base in order to determine the relationship among the predicate; then through the basic tuple concept node, concept map has been determined by the conceptual relations between these concepts of graph; finally show the predicate expression in association rules with the form of conceptual graphs for sake of further conversion of the association rules.(2) Propose the method of generating the conceptual graphs into natural language. The conceptual graphs have been basically clear to express the relationship between each part of a sentence, namely the concepts and concept relations. First of all, plan the document according to the background knowledge in the domain knowledge base in so as to determine the information and structure of the generated text. Conceptual graphs is similar to documentation plan to some extent, which sharply reduces the difficulty in transferring the conceptual graphs into document plans; then take the document plan as input, it can generate a tree structure through the micro planning through vocabulary selection,aggregation and the selected expression. Part of a leaf node already contains the last expression generated natural language sentences; finally realize the expression language and structure by the domain knowledge base, and sentence background knowledge aided generation of natural language. Users without the professional knowledge can understand the mining results, improving the usability of the system.(3) Design and implement the prototype system of the association rules based on conceptual graphs, which introduces the system architecture and flow chart. The present author carries out experiments and analysis of the system.