Unsupervised entity relationship extraction method based on zero-shot

Abstract

The invention discloses an unsupervised entity relationship extraction method based on zero-shot, and belongs to the field of computers. An entity relationship category is judged by extracting triplefeatures in text data and entity relationship type features in a domain knowledge graph and calculating the similarity between the triple features and the entity relationship type features, so that the dependence of a traditional entity relationship extraction method on manual annotation is reduced, and the accuracy of entity relationship extraction is improved. The method comprises the steps of data preprocessing, feature extraction, relation extraction network model training and entity relation classifier training. A convolutional neural network model which is good at capturing sentence information is adopted to respectively extract triple and relationship type features, and finally softmax is used to predict entity relationship type tags. In a model construction process, a sparse taggedcorpus can be used as a training set, and in a test process, the same parameters as those in the training process can also be used for predicting the type of an untagged triad.

Description

technical field [0001] The invention belongs to the field of computers and relates to a zero-shot based unsupervised entity relationship extraction method. Background technique [0002] In today's big data era, due to the rapid growth of data and the variety of types, the problem of information overload is becoming more and more serious. Therefore, how to quickly and accurately obtain the important information required is the main problem we are facing today. Information extraction technology extracts important information contained in the text by extracting specified types of factual information such as entities, relationships, and events from natural language texts. As an important subtask in information extraction technology, entity relationship extraction mainly identifies and classifies the relationship between concepts in sentences or texts. At the same time, it is also the basis for many tasks in the field of natural language processing, such as machine translation, q...

AI Technical Summary

Problems solved by technology

[0005] Although the above methods can complete the task of entity relationship extraction, the existing entity relationship extraction methods still have the following challenges: (1) In many fields, the relationship trigger words between entities can not only be described by the entity pair nearby verbs , can also be described by adjectives or adverbs in the sentence

(2) Due to the late start of information technology in my country, there is a lack of large-scale and complete labeled corpora in most fields, such as the medical field

Images