Resource searching method, device and equipment

A resource search and resource identification technology, which is applied in the field of resource search methods, devices and equipment, can solve the problems of different names of video resources, failure to search for video resources, low recall rate, etc., and achieve the effect of improving the recall rate

Pending Publication Date: 2021-06-01
BEIJING IQIYI TECH CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, in the above solution, the video resource can only be searched if the text to be searched is the same as the name of the video. However, the text to be searched input by the user usually has irregularities, such as continuous typos in the text to be searched entered by the user , the text order is reversed, the input name is an abbreviation, etc. These circumstances make the name of the text to be searched different from the stored video resource, and the corresponding video resource cannot be searched
Therefore, this search scheme has a low recall

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
  • Resource searching method, device and equipment
  • Resource searching method, device and equipment
  • Resource searching method, device and equipment

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach

[0122]Figure 4 In the middle, Album (Resource Identification), Positive Query (Positive Text Corresponding Text) and Negative Query Enter the neural network to obtain E_A (Embed Embedded Feature Vector), E_Q_POS ( Resource Identification The corresponding forward text embedded feature vector), E_Q_NEG (Embedding feature vector corresponding to the resource identification), using the Triplet LOSS (triple-state loss function), based on E_A, E_Q_POS, and E_Q_NEG calculations; judgment If the loss function converges, if not, the parameters of the neural network are reversely propagated, the parameters of the neural network are reversed, and the resource identity, the resource ID corresponds to the negative text corresponding to the resource ID to the update parameters. Neural network; if it is, the training is completed to obtain a feature extraction network.

[0123]Figure 5 In, resource Identification 1 ... The resource Identification n is a resource ID in each resource information store...

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 resource searching method, device and equipment, in the scheme, a pre-trained feature extraction network is used for extracting embedded features of a to-be-searched text to serve as target embedded features; and the target embedding feature is matched with embedding features of a plurality of pre-stored resource identifiers, and a resource address corresponding to the embedding feature of the resource identifier matched with the target embedding feature is determined as a search result. The embedded features can express semantics of the to-be-searched text and the resource identifier, searching by using the embedded features can be understood as searching through semantics, and even if the to-be-searched text input by the user is not standard, the matched resource identifier can be searched through semantic searching, and in this way, The situation that part of resources cannot be searched due to the nonstandard to-be-searched text is reduced, and the recall rate of the scheme is increased.

Description

Technical field[0001]The present invention relates to the field of intelligent analysis, and in particular, to a resource search method, apparatus, and equipment.Background technique[0002]Currently, resource search needs in some scenarios. For example, users can search for video resources they want to watch in the video website; users can search for topic resources that you want to pay attention to on your social platform.[0003]Searching for video resources as an example, search schemes typically include: obtaining information to be searched by user input; matching the name of the text to the stored video resource, only the same name is the same as the stored video resource In the case, the video resource corresponding to the video name is determined as the search results.[0004]However, in the above scheme, only the video resource can only be searched in the case of the video name to be searched, but the user input to the text usually has an irregular situation, such as the user ent...

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): G06F16/78G06F16/73G06F16/68G06F16/63G06F16/33
CPCG06F16/3334G06F16/63G06F16/686G06F16/73G06F16/7867
Inventor 代嘉慧
Owner BEIJING IQIYI 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