Knowledge base construction method based on multi-modal fusion

Abstract

The invention discloses a knowledge base construction method based on multi-modal fusion, which comprises the steps of extracting basic attributes of entities based on information box information of entries in a data source to establish a binary relationship, and generating a mapping dictionary of the information box attributes and the relationship; establishing a topic entity and relationship extraction model based on a word vector space and a conditional random field, and expanding the binary relationship to generate a multivariate relationship template; constructing a feature set of entities based on the extracted names of the binary relationships, performing hierarchical clustering division on the entities, constructing a classification system, and summarizing information of the classification system; carrying out attribute normalization on data in the data source according to the inherent attributes and attribute values of the entities; determining a fused data set, identifying equivalent entities in different data sources according to an entity alignment algorithm, establishing equivalent links of the same entity in the different data sources, and performing multi-source datafusion; and correcting and expanding the knowledge graph after multi-source data fusion to realize automatic knowledge completion.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a knowledge base construction method based on multimodal fusion. Background technique [0002] Big data has gradually penetrated into all aspects of scientific and technological development, national economy and people's daily life. Discovering knowledge and mining information from massive data has become a key technology. The rapid development of natural language processing provides feasibility for the construction and expansion of knowledge graphs, while knowledge extraction and unified representation provide theoretical and technical support for understanding and associating knowledge. In 2012, Google proposed the concept of knowledge graph, and took it as the key to the knowledge engine and the core of the next-generation search engine, realizing the conversion of knowledge and things in the objective world to data in the Internet. The proposal and rapid development of kno...

AI Technical Summary

Problems solved by technology

[0004] At present, there are two main problems in multi-source data fusion: (1) Data from different data sources are polymorphic and heterogeneous, and different data sources may be different at the ontology level, such as attribute names. Different expressions; (2) There may be conflicting data between different data sources, and data conflicts caused by some incomplete, wrong, and outdated problems may mislead the acquisition of information

Images