Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Model generation method and system, semantic recognition method and system, equipment and storage medium

A technology of model generation and semantic recognition, applied in the field of semantic recognition, can solve the problems of inflexibility and high labor costs, and achieve the effect of improving accuracy, user experience and click conversion rate

Pending Publication Date: 2019-09-03
BEIJING JINGDONG SHANGKE INFORMATION TECH CO LTD +1
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is to overcome the inflexibility of semantic recognition using Stanford regular matching templates to extract semantic information in the prior art, and the need to design more and more regular templates resulting in high labor costs, and to provide a flexible Accurately extract key information from user voice input information and then realize semantic recognition based on LSTM, CNN and conditional random field user semantic recognition model generation, semantic recognition method, system, device and storage medium

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
  • Model generation method and system, semantic recognition method and system, equipment and storage medium
  • Model generation method and system, semantic recognition method and system, equipment and storage medium
  • Model generation method and system, semantic recognition method and system, equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0074] Such as figure 1 As shown, the method for model generation provided in this embodiment includes the following steps:

[0075] Step 101, acquiring historical data.

[0076] Step 102, perform word segmentation and labeling on each piece of historical data to obtain a corresponding first historical sequence.

[0077] In step 102 of this embodiment, each of the first history sequences includes words after word segmentation and tagged tags corresponding to each word after word segmentation; the tags include the central word of the item and the modifiers of the item , inquiry range, brand and channel number of the item, each of the tags also includes position information of the corresponding word after word segmentation in the corresponding historical data, and the position information includes a start position, an intermediate position and / or or end position. In this embodiment, the tag adopts One-Hot (one-hot code) encoding.

[0078] Step 103, using CNN to perform chara...

Embodiment 2

[0093] Such as figure 2 As shown, the system of model generation of this embodiment includes:

[0094] The first acquisition module 1 is used to acquire historical data.

[0095] The first processing module 2 is used to segment and label each piece of historical data to obtain a corresponding first historical sequence, and each of the first historical sequences includes word after word segmentation and each word after word segmentation The corresponding label after the annotation.

[0096] The second processing module 3 is configured to use CNN to perform character-level feature extraction on each of the first historical sequences to obtain a second historical sequence, and the second historical sequence includes the information of the first historical sequence and the corresponding The extracted character-level features.

[0097] The third processing module 4 is used to use bidirectional LSTM to perform word-level feature extraction on each of the second historical sequen...

Embodiment 3

[0107] image 3 A schematic structural diagram of a model generation device provided by Embodiment 3 of the present invention. image 3 A block diagram of an exemplary model generation apparatus 30 suitable for implementing embodiments of the invention is shown. image 3 The shown model generating device 30 is only an example, and should not impose any limitation on the functions and scope of use of the embodiments of the present invention.

[0108] Such as image 3 As shown, the model-generating device 30 may take the form of a general-purpose computing device, which may be a server device, for example. Components of the model generation device 30 may include but not limited to: the at least one processor 31 , the at least one memory 32 , and the bus 33 connecting different system components (including the memory 32 and the processor 31 ).

[0109] The bus 33 includes a data bus, an address bus, and a control bus.

[0110] The memory 32 may include a volatile memory, such...

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 invention discloses a model generation method and system, a semantic recognition method and system, equipment and a storage medium. The model generation method comprises the following steps: obtaining historical data; performing word segmentation and labeling on each piece of historical data to obtain a corresponding first historical sequence; performing character-level feature extraction on each first historical sequence by adopting a CNN to obtain a second historical sequence; performing word-level feature extraction on each second historical sequence by adopting LSTM to obtain a third historical sequence; and performing model training on the third historical sequence by adopting a conditional random field learning algorithm to determine parameters of the conditional random field model. According to the model generation method and system, the semantic recognition method and system, equipment and the storage medium provided by the invention, compared with the traditional thought of template matching, the training thought of fusing the traditional conditional random field CRF and deep learning is more flexible and covers more users, and the user experience and click conversionrate of voice related services can be improved.

Description

technical field [0001] The present invention relates to the field of natural language processing, in particular to a model generation, semantic The identification method, system, equipment and storage medium. Background technique [0002] Speech recognition and semantic understanding are a trend in the development of Internet sites in the future. When a user speaks a sentence to a machine, when the speech is converted into text, how to accurately grasp the user's intention based on the text is becoming more and more important. In the existing technology, the Stanford regular matching template is used to extract semantic information. This implementation method is relatively rigid, and only the words specified in the template can be matched. With the expansion of application requirements, more and more regular templates need to be designed, which wastes manpower. , the effect is not flexible enough. Contents of the invention [0003] The technical problem to be solved by t...

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): G06F17/21G06F17/27G06N3/08G10L15/26
CPCG06N3/08G10L15/26G06F40/117G06F40/289G06F40/30G06F16/36
Inventor 王颖帅李晓霞苗诗雨
Owner BEIJING JINGDONG SHANGKE INFORMATION TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com