Research On The Key Techniques Of Knowledge Graph Construction For Contract Text

With the surge of business volume,contract management has attracted wide attention.However,contract management is facing the problems of high resource consumption and low labor efficiency,and the development of artificial intelligence provides a solution for this.The construction of knowledge graph for contract text plays an important role in the development of intelligent contract review system and the promotion of intelligent contract management.This paper proposes a technical scheme for constructing knowledge graph based on Chinese contracts,focusing on the use of natural language processing technology to extract contract knowledge from unstructured data for the construction of knowledge graph.For contract element extraction task,this paper proposes a contract element extraction model based on multi-semantic enhancement to solve the problems of complex element types and fine element granularity.The model first fuses the word-level enhancement provided by the pre-training language model,the word-level enhancement provided by the word-plus-word embedding,and the word-level enhancements provided by matching word sets as model inputs,and then uses a deep neural network to obtain the optimal labeling sequence.The experimental results show that the multiple semantic enhancements can improve the modeling ability of the model for the contract context,and the three enhanced structures can effectively improve the effect of factor extraction.For contract clause classification task,a contract clause classification model based on multi-task learning mechanism is proposed to solve the problem of nesting between terms and elements.The model first interacts with the two tasks of prime extraction and clause classification,and then uses the attention mechanism to further mine their relevance,and finally introduces vocabulary knowledge to improve the dynamic multi-task learning model for the complex semantic environment of text-level contracts.The experimental results show that the multi-task learning mechanism can fully capture the shared features between tasks and achieve better classification results than the single-task model,and the introduction of contract knowledge can effectively improve the results of clause classification.For contract relation extraction task,in order to solve the problem of contract relation overlap,this paper proposes a contract relation extraction model based on the guidance of relation attention.The model reverses the traditional "entity → relation" extraction mode,and introduces a multi-head attention mechanism to represent specific relations,and constructs a relation decoder to guide the identification of elements and terms.The experimental results show that the model achieves the best results on the contract data set.In addition,based on the above method,the contract knowledge map is constructed,and the construction process is described from three aspects of knowledge modeling,knowledge extraction and knowledge storage.To sum up,this paper focuses on the construction of knowledge map for contract text,and studies the method of automatic construction of contract knowledge from the perspective of element extraction,clause classification and relationship extraction,which is of great significance to intelligent contract management.