Named-Entity Recognition METHOD FOR PROVIDING POST-PROCESSING FOR IMPROVING THE ACCURACY OF NAMED-ENTITY RECOGNITION AND SERVER USING THE SAME

摘要

The present invention relates to a method for providing a post-processing process for improving the accuracy of Named-Entity Recognition and a server using the same. According to the present invention, (a) when input data is input, the server checks the value of the specific field included in the record of the input data by causing the entity name supplementary determination module to (i) the specific field If there is no value of, or (ii) when the value of the specific field is plural, determining a specific record including the specific field as a supplementary necessary record; (b) the server, a rule-based complement module, which is a module for determining a value of the specific field by expressing a value of the supplementary record and (i) a predetermined knowledge as a rule by a data configuration module for complement processing, ( ii) An ontology-based complement module, which is a module that determines the value of the specific field by searching for the specific ontology part that is most similar by comparing the supplementary record with the ontology DB, and (iii) the GT (Ground Truth) value of the supplementary record. Using the algorithm to learn the parameters of the classification algorithm to pass through to the machine learning-based complementary module, which is a module that determines the value of the specific field, the rule-based complementary module, the ontology-based complementary module, and the machine learning-based complementary module respectively Converting and transmitting the value of the supplementary record according to a specific format required; And (c) the server selects a correct answer candidate value of the supplementary necessary record according to output values of the supplementary entity name determination module, the rule-based supplementary module, the ontology-based supplementary module, and the machine learning-based supplementary module. And determining a correct answer value of the supplementary need record among the selected correct answer candidate values with reference to a predetermined condition.