Method for predicting multiple criminal names by using sequence generation network based on multilayer attention

A technology of sequence generation and attention, applied in biological neural network models, neural learning methods, special data processing applications, etc., can solve problems such as large distance between relevant information and required information, loss of key information, and inability to connect relevant information. , to achieve the effect of improving modeling ability and forecasting effect, and strengthening information flow

Active Publication Date: 2020-04-17
SHANDONG UNIV +1
View PDF7 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

When we want to predict the unknown word in "I grew up in France...I speak fluent French.", which is "French", according to the previous information "Ispeak fluent", we can know that the next word should be a language, but which A language must get more information from the previous sentence "I grew up in France", and the distance between the relevant information and the position where the information is needed is very large. When the distance increases, the RNN becomes unable to connect the relevant information and loses a lot of key information

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
  • Method for predicting multiple criminal names by using sequence generation network based on multilayer attention
  • Method for predicting multiple criminal names by using sequence generation network based on multilayer attention
  • Method for predicting multiple criminal names by using sequence generation network based on multilayer attention

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0109] A method for multi-crime prediction using a multi-layer attention-based sequence generation network, such as figure 2 shown, including the following steps:

[0110] (1) Data preprocessing:

[0111] Because the data set is the original data set and does not meet the input requirements of the model, the data needs to be preprocessed. Screen the original data, the original data is the judgment document, extract the description of the criminal facts contained in the judgment document using the method of regular expression matching, and perform Chinese word segmentation to obtain all the data sets of the judgment document; After the data set is scrambled, it is divided into several parts, set as N, N-1 as the training data set, and the remaining 1 as the test data set;

[0112] (2) Training word vectors to obtain semantic information, semantic information refers to word vectors:

[0113] Input the training data set that above-mentioned step (1) obtains into skipgram neur...

Embodiment 2

[0124] According to a method of multi-crime prediction using a multi-layer attention-based sequence generation network described in Embodiment 1, the difference is that:

[0125] After step (3), extract legal articles, including: first, use the article extractor to select the first k legal articles, and then obtain the feature vectors of the k legal articles to represent semantic information, and send the feature vectors to the attention mechanism middle.

[0126] The legal article extraction part is set according to the content of the data set. As mentioned in the later experiments, the CJO data set contains legal information, and the legal article extraction module can be added. There is no legal information in the CAIL data set, and the legal article extraction is not added to the model. part;

[0127] The present invention also adds a section for extracting legal articles, using the information of legal articles in the data as an auxiliary means to predict related crimes....

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 relates to a method for predicting multiple criminal names by using a sequence generation network based on multi-layer attention, which better realizes context content dependence betweentexts on the basis of fusing a neural network and an attention mechanism so as to more accurately extract multiple criminal names of text contents. An original data set is transformed based on a multi-criminal-name prediction model of a multi-layer attention mechanism (nested word-level and sentence-level attention mechanisms), and then association information among criminal names is fused into the model through logical connection among criminal law criminal names. According to the method, a legal provision extractor and a legal provision text encoder are added, legal provision information isintroduced, text information irrelevant to a criminal name is filtered out from an original text through attention operation, information representation of a text corresponding to the criminal name to be predicted is enhanced, and therefore the prediction precision of the model on the criminal name to be predicted is improved.

Description

technical field [0001] The invention relates to a method for predicting multiple crimes by using a multi-layer attention-based sequence generation network, which belongs to the technical field of natural language processing. Background technique [0002] With the support of artificial intelligence and big data technology, legal research is moving towards intelligence and automation. Informatization has undergone a transformation of legal retrieval, and the digitization of legal documents such as legal texts and judgment documents has supported a huge legal database market. In 2014, the "Regulations of the Supreme People's Court on the Publication of Judgments by the People's Courts on the Internet" was officially implemented. The Supreme Law established the China Judgment Documents Network on the Internet to uniformly publish the effective judgment documents of the people's courts at all levels. [0003] Judgment documents contain a large amount of data. At present, the big...

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): G06F16/35G06F16/33G06F40/30G06K9/62G06N3/04G06N3/08
CPCG06F16/35G06F16/3344G06N3/084G06N3/047G06N3/045G06N3/044
Inventor 李玉军马宝森朱孔凡马浩洋
Owner SHANDONG UNIV
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