Recall model training method, search and recall method, device and computer equipment

Abstract

The present application relates to a recall model training method, search and recall method, device, computer equipment, and storage medium. The embodiments can be applied to various scenarios such as cloud technology, artificial intelligence, and intelligent transportation. The method includes: obtaining training search sentences, corresponding forward recall sentences and reverse recall sentence sets; the forward recall sentences are the recall sentences corresponding to the target operation state in the target recall sentence sets corresponding to the training search sentences, and the reverse recall sentence sets Including the first reverse recall statement, the second reverse recall statement and the third reverse recall statement, the first reverse recall statement is randomly selected from the candidate recall statement set, and the second reverse recall statement is the target recall statement set The recall sentence corresponding to the non-target operation state, the position difference between the first position corresponding to the third reverse recall sentence and the second position corresponding to the forward recall sentence satisfies the preset conditions, and the target recall model can be improved by training the target recall model through the above data. Model prediction accuracy.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to a recall model training method, search and recall method, device, computer equipment and storage medium. Background technique [0002] With the rapid development of Internet technology, the functions of search engines are becoming more and more powerful. Recall is an important link in search, and information related to search sentences can be recalled from massive data according to the search sentences entered by users. [0003] In traditional technology, a machine learning model can be trained to determine the similarity between search sentences and candidate information, so that target information can be selected from a large number of candidate texts based on the similarity as the recall result. Then, random sampling is usually used to obtain negative samples in the model training samples. The model trained based on such negative samples can only learn the differe...

AI Technical Summary

Problems solved by technology

Then, the negative samples in the model training samples are usually obtained by random sampling. The model trained based on such negative samples can only learn the difference in coarse-grained, and there is a problem of low prediction accuracy.

Images