Query recommendation method and system based on improved VHRED and reinforcement learning

Abstract

The invention relates to a query recommendation method and system based on improved VHRED and reinforcement learning, and the method comprises the following steps: A, collecting a user query log record of a search engine, carrying out the preprocessing of the user query log record data, and constructing a user query log training set TS; b, training a query recommendation deep learning network model based on VHRED with time characteristics and reinforcement learning by using the user query log training set TS; and C, the query recommendation system receives a query statement input by the user,inputs the query statement into the trained query recommendation deep learning network model, and outputs a matched query recommendation. The method and the system are beneficial to generating query recommendations meeting the requirements of users.

Description

Technical field [0001] The invention relates to the field of natural language processing, in particular to a query recommendation method and system based on improved VHRED and reinforcement learning. Background technique [0002] Query suggestions provide suggested queries for sessions entered by the user. Query suggestions can enable search engines to better understand the user's query intent, so as to better optimize the user's query. Therefore, this task has received considerable attention in the past decade. [0003] Cao et al. proposed a context-aware query suggestion framework that considers the entire query sequence in the conversation instead of only considering the last query. They used query clusters to build a concept sequence suffix tree for efficient and effective context-aware query suggestions. The query sequence can also be modeled by a mixed variable memory Markov model. Context-aware queries suggest that more user actions should be considered in the conversati...

AI Technical Summary

Problems solved by technology

[0004] The previous models only used the HRED model as the generative model, and the effect of the model still has room for improvement

Moreover, most models ignore the time characteristics of the query, and the time characteristics also have a great impact on the generation effect of the model.

If only the generator model is used to generate queries, it cannot guarantee that the generated queries can be prepared to be close to the queries generated by users. The generated queries have obvious traces of machine generation and cannot express the user's query intent well.

Images