Method and system for automatic construction of knowledge graph for massive unstructured text

Abstract

The present invention relates to the technical field of computer software. Disclosed are a knowledge graph automatic construction method and system for massive unstructured text. A named entity recognition problem is abstracted into one sequence labeling problem: one sentence is given, each word in a sentence sequence is labeled; an effective feature is designed on the basis of training data, various classification models are learned, a trained classifier is used to predict relations; multiple pieces of existing knowledge are linked, one large-scale unified knowledge network is created from the top level; and entity information are grabbed from three major online encyclopedias, open websites, relevant knowledge bases or search engine logs and integrated. The present invention significantly increases the speed of constructing a knowledge graph, increases time efficiency, reduces costs for labor resource by 30% or more. At the same time, the present invention provides improved domain portability, when constructing the knowledge graph, optimization is only required for entities and a relational extraction algorithm in the present invention for rapid implementation.

Description

technical field [0001] The invention belongs to the technical field of computer software, and in particular relates to a method and system for automatically constructing a knowledge graph facing massive unstructured texts. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: Knowledge Graph aims to describe the entities of the objective world and the relationship between them. attribute composition. In 2012, Google first launched the knowledge graph and used it to enhance search results in search engines, which also marked the successful application of large-scale knowledge graphs in Internet semantic search. In other words, the knowledge graph is composed of a large amount of knowledge, and each piece of knowledge is represented by a triplet, for example: (China, capital, Beijing). At present, knowledge graphs are mostly extracted and constructed from encyclopedic structured data; knowledge graphs can serve cust...

AI Technical Summary

Problems solved by technology

[0003] To sum up, the problems existing in the existing technology are: At present, there are few automatic construction methods of knowledge graphs for massive unstructured texts, and the technical difficulty is relatively high.

The existing methods are mainly manual, with a low degree of automation, requiring heavy manual labor to construct, trim, deduplicate, process, and align the atlas. The entire process is highly professional, heavy workload, and difficult to maintain.

Images