Unlock instant, AI-driven research and patent intelligence for your innovation.
A method of sentence prediction based on multi-task artificial neural network
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A technology of artificial neural network and prediction method, which is applied in the field of sentence prediction based on multi-task artificial neural network, can solve the problems of large deviation of actual prediction results, ignoring useful information association, ignoring correlation relationship, etc., and achieve the effect of improving accuracy
Active Publication Date: 2019-08-13
SHANDONG UNIV
View PDF3 Cites 0 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
[0004] Although the sentence prediction method based on the artificial neural network has realized the intelligent mining and utilization of judgment document information to a certain extent, the traditional sentence prediction method based on the artificial neural network directly predicts the sentence according to the description of the crime facts, and has not been fully utilized. The large amount of information contained in the judgment documents ignores the correlation between the sentence and various information in other dimensions. The sentence prediction method based on the single-task artificial neural network only takes the facts of the crime as the input and the sentence as the output, ignoring the fact in the judgment documents. The correlation between useful information in various dimensions, such as charges, basic information of suspects, and attribute information described by criminal facts, leads to problems such as poor convergence of model training and large deviations in actual prediction results, making it difficult to meet practical application requirements
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
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0067] A method of sentence prediction based on multi-task artificial neural network, such as figure 1 shown, including the following steps:
[0069] Extract the required information, realize data structuring, and construct structured data sets;
[0070] (2) Training stage:
[0071] Randomly divide the structured data set into two parts, with a ratio of 8:2. The large part of the data set is divided into N parts after being disrupted. Each time, N-1 parts are taken for training, 1 part for verification, and N times of cross-validation. Evaluate the performance of the model, and use a small part as the test data set; obtain the training data required for the current training stage, perform word segmentation processing and word vector mapping on the training data in turn, input the model, and obtain the output;
[0072] model such as figure 2 As shown, the model includes word vector embedding layer, bidirectional LSTM layer, maximum ...
Embodiment 2
[0086] According to a kind of sentence prediction method based on multi-task artificial neural network described in embodiment 1, its difference is:
[0087] In step (2), for the classification task of predicting crimes and intermediate attributes, the classification error in the form of cross-entropy is used to calculate the error between the output and the target; the cross-entropy calculation formula is shown in formula (I):
[0088]
[0089] In formula (I), y' i is the i-th value in the label, y i is the corresponding predictive component, when the cross-entropy is smaller, the classification is more accurate. h y′ (y) refers to the cross entropy;
[0090] In step (2), for the sentence regression task, the mean square error is used to calculate the error, and the mean square error between the target and the actual sentence is calculated; the mean square error calculation formula is shown in formula (II):
[0091]
[0092] In formula (Ⅱ), y' i is the i-th value i...
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
Login to View More
Abstract
The invention relates to a method for predicting the term of sentence based on a multi-task artificial neural network, which is used for data analysis and mining of judgment documents, and for predicting the term of sentence for newly obtained descriptions of criminal facts and basic information of suspects. This method is guided by the focus on the attributes of legal clauses quoted in referee documents, and constructs a multi-task neural network model with mutual support effects by fully utilizing the multi-dimensional relevant data in referee documents, and trains with the preprocessed referee document data. A high precision and practical sentence prediction method is obtained. This method takes 21 legal articles concerned attributes such as crime, degree of injury, and amount involved as auxiliary tasks, and classifies life sentence and death penalty separately, so as to realize accurate prediction of sentence. Compared with a single-task neural network model that does not use such attributes, the prediction accuracy of the method proposed by the present invention is higher, indicating that the method proposed by the present invention is effective and practical.
Description
technical field [0001] The invention relates to a method for predicting a sentence based on a multi-task artificial neural network, which belongs to the technical field of natural languageprocessing. Background technique [0002] The sentence prediction problem is one of the important basic problems in the information mining and analysis of judgment documents. Its goal is to predict the sentence that will be imposed on the basis of the crime facts based on the law and other relevant information based on the description of the crime facts. It can be used for subsequent automatic trials and legal intelligent consultation. In recent years, the development and application of natural languageprocessing technology based on neural network has greatly promoted the development of referee document information processing and mining. The traditional sentence prediction method based on artificial neural network is based on the description of criminal facts and directly predicts the se...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to View More 
Login to View More