User intention recognition method and device, and model construction method

A user intent and recognition model technology, applied in the field of information search, can solve the problem of low recognition accuracy, and achieve the effect of avoiding syntax noise, reducing demand, and improving accuracy

Pending Publication Date: 2021-04-13
北京欧拉认知智能科技有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem that the existing user intention recognition module has low recognition accuracy when the voice is ambiguous, the present invention provides a user intention recognition method, device and model building method, which parses the corpus to be searched into nodes containing entities and relationships The language is used as the input of the user intent recognition model, which can avoid the existence of grammatical noise in the recognition process and greatly improve the accuracy of intent recognition

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
  • User intention recognition method and device, and model construction method
  • User intention recognition method and device, and model construction method
  • User intention recognition method and device, and model construction method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0038] Such as figure 1 A method for building a user intent recognition model is shown, including the following steps:

[0039] Collect training set corpus and parse it into node sentences containing entities and relationships;

[0040] constructing and training a translation model comprising an encoder network and a decoder network;

[0041] Add attention and fully connected network to the trained encoder network to form a user intent recognition model;

[0042] The parameters of the encoder network are fixed, and the user intent recognition model is trained using the parsed training set.

Embodiment 2

[0044] Based on the principle of the above construction method, this embodiment discloses a specific implementation manner.

[0045] Collect and analyze the first training set corpus: collect the corpus to be searched by users and store and parse it, and parse it into node sentences containing entities and relationships. Taking "Yao Ming's height" as an example, after parsing, the entity is "Yao Ming" and the relationship is " Height", for example, "the height of Yao Ming's wife", after analysis, its entity is "Yao Ming", and the relationship is "wife" and "height". This step can be completed before training the user intent recognition model.

[0046] Collect the second training set corpus: collect the translation corpus, the translation corpus includes the source language and the corresponding target language, and perform word segmentation on the translation corpus, where the source language can be Chinese, English, German and other languages, and the target language can also...

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 user intention recognition method and device and a model construction method, and aims at solving the problem that grammar noise exists in the recognition process and greatly improving the accuracy of intention recognition by analyzing a corpus to be searched into node languages containing entities and relations and taking the node languages as input of a user intention recognition model. The model construction method comprises the following steps: collecting training set corpora and analyzing the training set corpora into node statements containing entities and relationships; constructing and training a translation model, wherein the translation model comprises an encoder network and a decoder network; adding attention and a full connection network to the trained encoder network to form a user intention recognition model; and fixing encoder network parameters, and training a user intention recognition model by using the analyzed training set.

Description

technical field [0001] The invention belongs to the technical field of information search, and in particular relates to a user intention recognition method, device and model building method. Background technique [0002] As the amount of data on the Internet surges, as an effective tool for organizing, extracting, and searching information resources, search engines can help users quickly locate desired Internet resources. The effectiveness of search engines depends not only on whether search users can transform their intentions into accurate query words, but also on how well the query words are understood by search engines. When these two aspects can be well completed, The quality of the returned results can be greatly improved. The current search engine is very weak in the recognition of user search intent. Especially in the question and answer, due to the uneven level of the questioners and the strong casualness of the spoken language, there may be a lot of missing conte...

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): G06F16/36G06F16/33G06F40/279
CPCG06F16/367G06F40/279G06F16/3344
Inventor 王绪刚
Owner 北京欧拉认知智能科技有限公司
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