Research On Knowledge Inference And Verification For Open Information Extraction System

Open information extraction is the main way to mine knowledge from unstructured text.However,there are lot of noises in the results of information extraction,which have a great impact on knowledge discovery and knowledge base construction.To solve this problem,this paper proposes a verification method of knowledge inference based on probabilistic soft logic.In this method,the first-order logic language is use to transform the results of knowledge extraction,then inference process is carried out,and rules are introduced into the inference process for semantic constraints.In order to solve the problem that the current inference rules are too dependent on manual customization,this paper establishes an automatic learning mechanism for inference rules,which relieves the traditional knowledge inference from relying on artificially formulated rules,and realizes comprehensive automatic inference and verification of knowledge.The experimental results show that the inference model proposed in this paper has better algorithm performance than the comparison model,improves the efficiency of knowledge inference,and has a positive effect on verifying the semantic standardization and correctness of knowledge.At the same time,we designed three strategies to evaluate the quality of the learned rules.The first strategy is to compare with the rules obtained by the expert system.The second strategy is to put the obtained rules into inference model for verification.In order to prove that the rules obtained by the rule learning model have universal applicability,we design third strategy to place the rule learning method on two different data sets for learning,and introduce the obtained rules into the Trans series model,try improving its effectiveness in knowledge verification,and based on the final inference effect as an important basis for evaluating the universal applicability of the learning rules.