Model training method and device based on word embedding, electronic equipment and storage medium

Abstract

The invention is suitable for the technical field of artificial intelligence, and provides a model training method and device based on word embedding, electronic equipment and a storage medium, and the method comprises the steps: carrying out the word segmentation based on a query item and a data matching item matched with the query item, and obtaining a model training word list; inputting a target word in the model training word list into a to-be-trained model, and respectively obtaining vector representations associated with the target word from an initialized word vector matrix, a query item vector matrix and a data matching item vector matrix in the to-be-trained model; splicing the vector representations, and performing feature cross fusion through a to-be-trained model to obtain a target word vector representation; and based on the target word vector representation, obtaining vector representation parameters in the query item vector matrix and the data matching item vector matrix in the to-be-trained model, and obtaining a trained model containing the vector representation parameters. According to the scheme, the accuracy of text matching results can be improved, and text matching requirements are met.

Description

technical field [0001] The present application belongs to the technical field of artificial intelligence, and in particular relates to a word embedding-based model training method, device, electronic equipment and storage medium. Background technique [0002] Text matching has always been one of the hot tasks in the field of information retrieval such as search engines and recommendation systems. It is mainly aimed at query items given by users, using models to match documents with high similarity from the database and returning them to the user. The whole process includes two modes, recall and sort, which are used to match and filter the text and sort the recall results. [0003] In the recall mode, the existing word embedding-based recall method mainly uses the word2vec model to obtain the vector representation of each word, and then sums all the word vectors in the query and document to obtain the average value, and then obtains the vector representation of the query and...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a word embedding-based model training method, device, electronic equipment, and storage medium to solve the problem that only literal matching can be achieved between query items and matching content in the prior art, but the core ideas are far from each other. The accuracy of the matching results is not good enough to meet the text matching requirements

Images