Building domain knowledge graph construction method based on neural network self-adaptive optimization parameter adjustment

Abstract

The invention discloses a building domain knowledge graph construction method based on neural network self-adaptive optimization parameter adjustment. The method comprises the steps: firstly collecting building domain entities, entity attributes and corpora, and then carrying out the preprocessing, and obtaining a building entity recognition model annotation data set EntityData; training a self-adaptive entity recognition model based on BERT-BiLSTM-CRF by utilizing the data set EntityData, so as to obtain an entity recognition model EntityModel in the building field; preprocessing the buildingdomain relation data set to obtain a data set RelationData; training a self-adaptive relation extraction model based on the GRU by utilizing RelationData to obtain a RelationModel of a building domain relation extraction model; and finally, entities and attributes in the building field text data set are extracted through the EntityModel and the RelationModel respectively, and a building field knowledge graph is constructed. A user inputs a construction drawing review point through the Web platform, a hidden relation in the knowledge base is mined according to the construction drawing review point input by the user, and mined expert opinion information Recom is returned to the Web platform. According to the method, self-adaptive parameter adjustment is adopted, and compared with traditional manual setting, the knowledge graph of the building field is constructed more effectively.

Description

technical field [0001] The invention belongs to the technical field of knowledge graph construction, in particular to a method for constructing knowledge graphs in the construction field based on neural network adaptive optimization and parameter adjustment. Background technique [0002] Entity recognition, also known as named entity recognition or entity extraction, is to automatically identify named entities from raw data corpus. Since entities are the most basic elements in knowledge graphs, their extraction integrity, accuracy, and recall will directly affect the quality of knowledge graph construction. Therefore, entity recognition is an important method to acquire knowledge from semi-structured or unstructured data, and it is the basis and key step for building a knowledge graph. [0003] Knowledge extraction is to extract available knowledge units from natural language text or multimedia content documents through automatic or semi-automatic technology. Knowledge unit...

AI Technical Summary

Problems solved by technology

However, at present, there is no method to combine the entity recognition and relationship extraction in the construction field by using adaptive parameter tuning to train the neural network model to construct a knowledge graph in the construction field and apply it to the expert recommendation system.

Images