Text data processing method, apparatus and device, and medium

Abstract

The invention relates to a text data processing method, apparatus and device, and a medium, for improving the accuracy and efficiency of attribute value labeling, reducing the labor cost and avoidingerror accumulation. The text data processing method comprises the steps of converting a target text into a text vector sequence based on a predetermined attribute vector sequence used for representingan attribute sequence in the target text; and inputting the text vector sequence into a sequence labeling model used for labeling word attributes in a text, and labeling attributes of each word in the target text in a label form, the attributes being entity attributes of entities corresponding to subjects in the target text.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, in particular to a text data processing method, device, equipment and medium. Background technique [0002] In the context of the large-scale emergence of artificial intelligence technology and applications, the use of triples to represent knowledge is the basis of knowledge graphs and a powerful driving force for the development of artificial intelligence technology. A piece of knowledge expressed in the form of a triple (for example, Zhang San, date of birth, January 18, 1979) expresses the attribute of "date of birth" of the entity "Zhang San", and the three parts of the triple are called the subject , predicate, object. Introductory articles usually describe the same subject entity and densely introduce a large number of attributes about the subject entity. How to extract the attribute values ​​(knowledge triples) of structured representations from the natural la...

AI Technical Summary

Problems solved by technology

[0011] When using the above relationship classification method for attribute extraction, the object must be an entity that can be recognized by the NER model. If the model cannot recognize a certain type of entity, then the relationship related to this type of entity cannot be detected.

Moreover, errors will accumulate in the NER process. If there is a recognition error, the result of the relationship classification must be wrong. If the traditional machine learning model is adopted, it is also necessary to extract features. The information that can be contained in the features depends on manual experience.

Images