OTA search deep learning sorting method, terminal and storage medium thereof

Abstract

The invention relates to an OTA search deep learning sorting method, a terminal and a storage medium thereof, and belongs to the field of search sorting, and the method comprises the following steps: cleaning resources selected by a user and storing the resources in a user log; regularly acquiring resources recorded in the user log, and performing data cleaning on the recorded resources; training a neural network model according to the cleaned data; and replacing the original neural network model with the trained neural network model. The method has the advantages that the rough arrangement candidate resources obtained through rough arrangement are subjected to fine arrangement, fine arrangement candidate resources are obtained, calculation needs to be conducted through the neural network model in the fine arrangement process, the neural network model is trained according to the recorded user selection resources, and therefore the more accurate neural network model is obtained; and replacing the original neural network model with the neural network model obtained after training, thereby achieving the effects of improving the search accuracy and effectively reducing the problem of relatively large errors.

Description

technical field [0001] The present application relates to the field of search sorting, in particular to an OTA search deep learning sorting method, a terminal and a storage medium thereof. Background technique [0002] In the past ten years, Internet search technology has developed rapidly. In terms of technical architecture, the search process can be divided into three modules: semantic recognition, recall, and sorting. Semantic recognition is responsible for parsing user queries, recall is responsible for recalling a small number of resources that meet user requirements from the massive resource library, and the sorting module is responsible for sorting the recalled resources. Finally, relevant resources are presented to the user according to the sorting results. As a module that directly interfaces with users, the sorting module has a huge impact on user experience. From the perspective of technological evolution, the ranking model has gone through the evolution proces...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that the error accumulates to the next step, which leads to the gradual amplification of the downstream error, and finally leads to the problem of low search accuracy, this application provides an OTA search deep learning sorting method, a terminal and its storage medium

Images