An extended search method and device

Abstract

The purpose of the present invention is to provide an extended search method and device. Wherein, the interest word model of each user is established; the expansion word model of each fourth search sequence is established; -according to the current search sequence of the current user, based on the interest word model of the current user and the expansion word of the current search sequence A model is used to determine corresponding extended search words; and to provide corresponding search results to the current user according to the current search sequence and the extended search words. The present invention can realize simpler and more effective personalized search, and has advantages in satisfying long-tail requirements of users (that is, increasing recall of personalized results), reducing calculation and storage costs, versatility, and practicability.

Description

technical field [0001] The present invention relates to the technical field of search, in particular to a technology of extended search. Background technique [0002] Personalized search technology can improve the accuracy of predicting users' true intentions by introducing users' long-term interests and short-term intentions as factors for recalling and sorting retrieval results, making search results more in line with users' needs. Existing personalized search methods are mainly realized by personalized rearrangement of top-n natural search results (ie, search results obtained according to the search sequence submitted by users), and there are many limitations in practical applications. [0003] There are following problems in prior art: [0004] 1) There are great limitations in the recall. The main purpose of rearranging natural search results according to user interests is to highlight results that match user interests while ensuring relevance. The above method is mo...

AI Technical Summary

Problems solved by technology

For image search and other application scenarios where the content description data of search results is not rich and the amount of search results is huge, it is difficult to meet the above two requirements

[0006] 3) The user interest model based on topic classification cannot fully meet the actual application requirements

Although the subject classifications established by manual methods (such as Open Directory Project) are of high quality, they are expensive to build and update, and have poor cross-domain migration capabilities.

Using machine learning algorithms (such as LDA) for automatic text topic classification has problems such as low accuracy and poor effect on short texts.

Images