Entity joint labeling relation extraction method and system based on probabilistic graph

Abstract

The invention discloses an entity joint labeling relation extraction method and system based on a probability graph, and belongs to the technical field of natural language processing. Comprising the steps of feature extraction; an entity extraction task is converted into a sequence labeling task, a sequence is input into a first model to obtain a first output feature, a prediction sequence is obtained after the first output feature is activated, and the starting position and the ending position of the entity are obtained through a set threshold value; the subject and the object are matched according to the principle of proximity, and similar entity heads and tails are marked and intercepted; and performing relationship classification: randomly extracting entity pairs, generating second output features according to the intermediate features of the first model, and inputting the second output features into a second model to obtain a corresponding classification relationship. According to the method, the correlation between the two sub-tasks is considered, so that a task extraction result does not excessively depend on an entity extraction result, and the problem of error accumulation and relation overlapping are avoided.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and in particular relates to a method and system for extracting entity joint annotation relations based on probability graphs. Background technique [0002] Relation extraction is an important sub-task of information extraction. The purpose is to extract structured data from unstructured text. The main task of relation extraction is to extract the entities in the text and the relationships between entities. These relationships are based on three Formal representation of tuples (subject, relation, object), which plays an important role in building knowledge graphs. The existing mainstream relation extraction techniques are divided into rule-based relation extraction, supervised relation extraction, unsupervised relation extraction and semi-supervised relation extraction. [0003] Rule-based relation extraction firstly extracts relational words based on rules and artificially, ...

AI Technical Summary

Problems solved by technology

However, because the clustering method itself has the problems of difficulty in describing the relationship and low recall rate of low-frequency instances, it is generally difficult to obtain good extraction results in unsupervised learning.

[0007] In the existing technology, most of the relationship extraction is regarded as a series of tasks, that is, the traditional pipeline method first performs entity recognition, and then predicts the relationship for each pair of entities. This method makes the task relatively simple and the division of labor is clear, but there is a problem. Series of problems: The correlation between the two subtasks is not considered in the solution of the two tasks, which leads to the result of the relationship extraction task heavily dependent on the result of the entity extraction, resulting in the problem of error accumulation

For the problem of overlapping relationships, the serial method cannot provide a better solution

Images