Automatic generation method of product description text based on pre-training model

Abstract

The invention discloses an automatic generation method of a product description text based on a pre-training model. The method is characterized in that a framework pre-training model method is adopted, an auxiliary database and a pre-trained BERT model weight are introduced, a product description text is automatically generated according to given product function class attribute characteristics. The method specifically comprises the steps of constructing a dictionary, a co-occurrence relation library and a retrieval database, and generating a model and a text. Compared with the prior art, themethod has the advantages that a section of product description text which is smooth, readable and rich in information amount is generated according to the given product function type attribute characteristics, the three problems of small data size, generated text length control and function type attribute characteristic controllability are effectively solved, and the generation quality and text diversity of the product description text are improved.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a method for automatically generating product description text based on a pre-trained model. Background technique [0002] The automatic generation of product description is a technology that automatically generates a piece of descriptive text that contains given feature information, is fluent and readable, and has a controllable text length based on given product attribute characteristics. Taking an air conditioner as an example, given the attribute characteristics "quiet, defrost, 5 horsepower", a description text containing the above characteristics will be automatically generated. Writing product descriptions is a tedious and time-consuming job for people, but product descriptions are undoubtedly important, because it is an important way for consumers to understand products, and to a certain extent determines whether consumers buy products or not. no. Au...

AI Technical Summary

Problems solved by technology

[0008] 1) A small amount of data can easily lead to underfitting and affect the quality of text generation. The acquisition of data, especially the acquisition of high-quality data, requires cost. How to generate smooth, readable and informative texts with a small amount of data? product description for

[0009] 2) Text length control, how to generate a product description with a controllable length, the information expressed by too short text length is insufficient, and too long text length will reduce consumers' desire to read

Images