Training method and device for rewritten word generation model

A technology for generating models and training methods, applied in the field of model training, can solve the problems of low coverage of long-tail traffic, low query rewriting accuracy and coverage, and low accuracy of long-tail traffic, etc., to improve the accuracy and coverage. Effect

Active Publication Date: 2021-10-15
BEIJING SANKUAI ONLINE TECH CO LTD
View PDF11 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] At present, the commonly used query rewriting methods are: data mining method, which mines through user behavior. Usually, the accuracy rate of mining on popular query keywords with relatively dense behavior is relatively high, but the accuracy rate of mining on long-tail traffic with sparse user behavior is accurate. relatively low rate
In use, mining data with relatively high accuracy is usually used, so the coverage on long-tail traffic is usually low, which leads to low accuracy and coverage of query rewriting

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Training method and device for rewritten word generation model
  • Training method and device for rewritten word generation model
  • Training method and device for rewritten word generation model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0069] refer to figure 1 , shows a flow chart of steps of a training method for rewriting a word generation model provided by an embodiment of the present disclosure, as figure 1 As shown, the training method of the rewritten word generation model may specifically include the following steps:

[0070] Step 101: Obtain an initial rewritten word generation model.

[0071] The embodiments of the present disclosure can be applied to a scenario where model training is performed in combination with search keywords, rewrite keywords, and deviation scores between the two keywords.

[0072] The initial rewritten word generation model refers to the rewritten word generation model obtained through pre-training and needs further training. The initial rewritten word generation model can be obtained by combining similar query keywords collected in advance to train the rewritten word generation model to be trained For the model obtained later, the specific training process for the initial ...

Embodiment 2

[0088] refer to figure 2 , shows a flowchart of steps of another training method for rewriting a word generation model provided by an embodiment of the present disclosure, as figure 2 As shown, the training method of the rewritten word generation model may specifically include the following steps:

[0089] Step 201: Obtain a set of query keywords according to the search log.

[0090] The embodiments of the present disclosure can be applied to a scenario where model training is performed in combination with search keywords, rewrite keywords, and deviation scores between the two keywords.

[0091] Search logs refer to the logs generated by users searching on various search platforms during the historical period.

[0092] When it is necessary to train the rewritten word generation model, the search log can be obtained, and query keywords can be obtained according to the search log, and similar query keywords can be formed into a query keyword set.

[0093] After the query ke...

Embodiment 3

[0149] refer to image 3 , which shows a schematic structural diagram of a training device for rewriting a word generation model provided by an embodiment of the present disclosure, as shown in image 3 As shown, the training device 300 of the rewritten word generation model may specifically include the following modules:

[0150] The initial generation model acquisition module 310 is used to obtain the initial rewritten word generation model;

[0151] A rewriting keyword acquisition module 320, configured to input search keywords into the initial rewriting word generation model, and obtain rewriting keywords corresponding to the search keywords output by the initial rewriting word generation model;

[0152] A deviation score acquisition module 330, configured to acquire a deviation score between the search keyword and the rewritten keyword;

[0153] The target generation model acquisition module 340 is configured to train the initial rewritten word model according to the de...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The embodiment of the invention provides a training method and device for a rewritten word generation model. The method comprises the steps of obtaining an initial rewriting word generation model; inputting a search keyword into the initial rewriting word generation model, and obtaining a rewriting keyword corresponding to the search keyword output by the initial rewriting word generation model; obtaining a deviation score between the search keyword and the rewriting keyword; and training the initial rewriting word model according to the deviation score, the search keyword and the rewriting keyword to obtain a target rewriting word generation model. According to the embodiment of the invention, the accuracy and coverage rate of query rewriting can be improved.

Description

technical field [0001] The embodiments of the present disclosure relate to the technical field of model training, and in particular to a training method and device for rewriting a word generation model. Background technique [0002] Query rewriting (also known as query expansion) is a method of rewriting the query words searched by users in search engines to improve the recall results of search engines. In the Meituan Dianping search engine, there are usually multiple expressions for a merchant or a dish; for example, "hot pot" is also called "shabu-shabu". Sometimes different words express the same user intent; for example, "wedding photography" and "wedding photos", "glasses" and "optical shop", etc. [0003] Search engine query rewriting can be regarded as recall optimization for search engines. Without changing user intent, recall as many search results that meet user intent as possible to improve user search experience. [0004] At present, the commonly used query rew...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/33G06F40/30G06F40/289
CPCG06F16/3338G06F40/30G06F40/289
Inventor 杨俭王宗宇谢睿武威
Owner BEIJING SANKUAI ONLINE TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap