Reinforcement learning-based sorting learning method and server

Abstract

The invention discloses a reinforcement learning-based sorting learning method. The method comprises the steps that a server obtains a historical query word and obtains M documents corresponding to the historical query word; the server performs sorting on the M documents to obtain a target document sorting list; the server obtains a sorting effect assessment value of the target document sorting list; the server takes the historical query word, the M documents, the target document sorting list and the sorting effect assessment value as a training sample, and puts the training sample into a training sample set; and when a quantity of training samples in the training sample set is greater than a preset quantity, the server performs training on the training sample set to obtain a sorting model. By adopting the technical scheme provided by the method, a sorting index can be accurately optimized, so that the satisfaction of a user to a sorting result of the document sorting list corresponding to the query word is improved.

Description

technical field [0001] The invention relates to the field of ranking learning, in particular to a ranking learning method and server based on reinforcement learning. Background technique [0002] With the rapid development of the Internet, information has exploded. How to quickly find out the data that users need from massive amounts of information has become the focus of information retrieval research. At present, to find out the required data from massive information is mainly to use the search engine to search and sort the search results. [0003] However, with the development of servers, more and more factors affect the ranking of search results, and it is no longer possible to manually fit the ranking model. At this time, machine learning is very suitable. Sorting learning algorithm is a very important machine learning algorithm at present. [0004] Learning to rank algorithms are a class of ranking algorithms based on supervised learning, which have been widely used...

AI Technical Summary

Problems solved by technology

[0005] Existing models based on ranking learning algorithms need to rely on the correlation annotation data between query words and documents for training, but cannot use the data obtained by users to evaluate the sorting effect of the document ranking list corresponding to the query words. Improve user satisfaction with sorting effects

Images