A system and method for improving the accuracy of public opinion warning using deep learning

Through the combination of deep learning technology, named entity recognition and Transformer model, the problem of inability to automatically extract named entities and lack of multi-classification in the existing technology is solved, and the accurate identification and detailed extraction of news and public opinion events is achieved, and the public opinion warning is improved Accuracy and level of detail.

CN114492439BActive Publication Date: 2024-11-05SICHUAN XW BANK CO LTD
View PDF 12 Cites 0 Cited by

Patent Information

Application Number
CN202111634848.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-27
Publication Date
2024-11-05
Estimated Expiration
2041-12-27

AI Technical Summary

Technical Problem

The existing public opinion warning technology cannot automatically extract named entities and lacks the identification of specific categories of news and public opinion events. The existing solutions mainly rely on binary classification, and the output results are vague and cannot meet actual business needs.

Method used

Using deep learning technology, combining named entity recognition and Transformer model, realize automatic extraction of named entities in news text and multi-classification model training of public opinion risk events, output named entity matrix, its confidence probability and risk event categories and confidence probability, using Generate early warning signal.

Benefits of technology

It greatly improves the accuracy and detailedness of the identification of news risk events, realizes the classification probability prediction and structured extraction of dozens of news events, and outputs the classification probability matrix and the event category with the highest confidence level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114492439B_ABST
    Figure CN114492439B_ABST
Patent Text Reader

Abstract

The present invention relates to a system and method for improving the accuracy of public opinion warning by applying deep learning, and belongs to the field of computer technology. The present invention includes a named entity recognition terminal: used to obtain news text from a data input port and output a named entity matrix and its confidence probability matrix P1; a news event classification terminal: used to obtain news text from a data input port and output a risk event category E2 and its corresponding confidence probability P2; a news public opinion warning terminal: used to prepare warning rules, receive a named entity matrix and its confidence probability matrix P1, a risk event category E2 and its corresponding confidence probability P2, and execute warning rules to output warning signals. The present invention can automatically extract named entities from news texts, and can simultaneously predict the classification probabilities of dozens of news events, output a classification probability matrix and the event category with the highest confidence, and greatly improve the accuracy and detail of news risk event identification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for improving the accuracy of public opinion warning by applying deep learning. Background Art

[0002] With the rapid development of new technologies such as mobile Internet and Internet of Things, human beings have entered the data age. The information storm brought by data is changing our lives, work and thinking, and has also had a profound impact on the management of online public opinion. With the advent of the new public opinion era, while paying close attention to public opinion risks, organizations need to deeply explore the value of public opinion big data and listen to the broad voices of society, the market and users.

[0003] There are three types of existing technical solutions for news public opinion warning. The first is the traditional keyword extraction and rule matching mode; the second is to use technologies such as Word2Vec (Word Embedding) to structure news text into word vectors, and then combine manually annotated risk event labels to use machine learning models to train public opinion risk event classification models, and finally output public opinion warning signals based on model prediction results; the third is the mode of applying deep learning models. The two-stage solution uses deep learning models to extract structured features of news texts and train machine learning classification models based on this. The end-to-end solution directly uses deep learning models to train the binary classification model of public opinion warning.

[0004] There are two shortcomings in the existing public opinion warning methods: the first is that it is impossible to automatically extract named entities such as enterprises, time, and places (in fact, most of them are implemented using keyword retrieval and template / regular matching methods); the second is that although deep learning models are applied, the specific categories of news public opinion events are not fully identified (the existing solutions are all binary classification). The existing technologies are classified through sentiment classification models (binary classification, positive-negative sentiment) and public opinion risk models (binary classification, whether it belongs to a public opinion event). The actual output classification results are relatively vague, which is far from enough for application in real business scenarios. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned prior art, the present invention discloses a method and system for improving the accuracy of public opinion warning using deep learning, the purpose of which is to improve the accuracy and detail of identifying news risk events.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A system for improving the accuracy of public opinion warning by applying deep learning, comprising:

[0008] Named entity recognition terminal: used to obtain news text from the data input port and output the named entity matrix and its confidence probability matrix P1;

[0009] News event classification terminal: used to obtain news text from the data input port and output the risk event category E2 and its corresponding confidence probability P2;

[0010] News and public opinion warning terminal: used to prepare warning rules, receive the named entity matrix and its confidence probability matrix P1, the risk event category E2 and its corresponding confidence probability P2, and execute the warning rules to output warning signals.

[0011] After adopting this technical solution, the present invention can automatically extract named entities such as enterprises, time, and place from news texts, improve the structured extraction of news entity elements, and realize multi-classification model training and prediction of news public opinion risk events. It can simultaneously predict the classification probability of dozens of news events, output the classification probability matrix and the event category with the highest confidence, greatly improving the accuracy and detail of news risk event identification.

[0012] Preferably, the named entity recognition terminal includes an entity element labeling module for labeling entity elements in news text, a named entity recognition model training module for training a named entity recognition model, and a named entity element extraction module for extracting a named entity matrix and its confidence probability matrix P1.

[0013] Preferably, the news event classification terminal includes a news event labeling module for labeling news events, a Transformer training module for training the Transformer model, and a news event classification module for outputting risk event categories E2 and their corresponding confidence probabilities P2.

[0014] Preferably, the news public opinion early warning terminal includes:

[0015] Warning rule configuration module: used to configure warning rules;

[0016] Warning signal generation module: used to output warning signals based on the named entity matrix and its confidence probability matrix P1, risk event category E2 and its corresponding confidence probability P2.

[0017] Preferably, step 1: obtaining news text, the named entity recognition terminal and the news event classification terminal respectively mark entity elements and risk events in the news text to obtain an entity element dataset D1 and a risk event dataset D2;

[0018] Step 2: Use the entity factor dataset D1 and the risk event dataset D2 to train the named entity recognition model and the Transformer model respectively, and obtain the trained named entity recognition model M1 and the trained news public opinion risk event classification model M2;

[0019] Step 3: Input the news text acquired in real time into the trained named entity recognition model M1 and the news public opinion risk event classification model M2 respectively. The named entity recognition model M1 outputs the named entity matrix and its confidence probability matrix P1, and the news public opinion risk event classification model M2 outputs the risk event category E2 and its corresponding confidence probability P2;

[0020] Step 4: The news public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3, executes the warning rules, and outputs a warning signal.

[0021] Preferably, the step 1 specifically includes:

[0022] Step 1.1: Get the news text, split the news text at the sentence level, use the shuffle function to randomly sort the news text at the sentence level, and extract a set proportion of the news text to manually annotate the entity elements. The remaining news text is used for the initial version model to predict pseudo labels. The news text with high confidence probability is marked as the true label, and the news samples with low confidence probability are manually corrected and annotated to obtain the entity element dataset D1;

[0023] Step 1.2: Obtain news text, split the news text at the sentence level, use the shuffle function to randomly sort the news text at the sentence level, and extract a set proportion of the news text to manually annotate the risk events. The remaining news text is used for the initial version model to predict pseudo labels (pseudo label), among which news text with high confidence probability is marked as the true label (true label), and news samples with low confidence probability are manually corrected and annotated to obtain the risk event dataset D2.

[0024] Preferably, the step 2 specifically comprises the following steps:

[0025] Step 2.1: Import the entity feature dataset D1 into the named entity recognition model. The named entity recognition model uses the BERT+IDCNN+CRF architecture to tokenize the text, divide the entity feature dataset D1 into a training set and a validation set, and then load the pre-trained model, fine-tune it on the entity feature dataset D, set the appropriate initial learning rate (lr) and batch size (batch size), and obtain the trained named entity recognition model M1 after iteration;

[0026] Step 2.2: Import the risk event dataset D2 into the Transformer model. The Transformer model adopts the BERT architecture, tokenizes the text, divides the risk event dataset D2 into a training set and a validation set, and then loads the BERT pre-trained model, initializes the learning rate (lr) and batch size (batch size), and performs fine-tuning (Finetune) on the risk event dataset D2. After iteration, the trained news public opinion risk event classification model M2 is obtained.

[0027] Preferably, the step 3 specifically comprises the following steps:

[0028] Step 3.1: Obtain news text from the data input port in real time and input it into the trained named entity recognition model M1. The named entity recognition model M1 outputs a named entity matrix and its confidence probability matrix P1; the named entity matrix includes corporate entities, events, time and place;

[0029] Step 3.2: Obtain news text from the data input port in real time and input it into the trained news public opinion risk event classification model M2. The news public opinion risk event classification model M2 outputs the classification probability matrix of news public opinion events, and calculates the risk event category E2 with the highest confidence and its corresponding confidence probability P2 based on it.

[0030] Further preferably, the step 4 specifically comprises the following steps:

[0031] Step 4.1: Configure warning rule R1 on the rule configuration page of the news and public opinion warning terminal;

[0032] Step 4.2: The news public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3;

[0033] Step 4.3: Execute the configured warning rule R1. If the threshold in the warning rule R1 is triggered, a warning signal is output, otherwise no warning signal is output.

[0034] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method are implemented.

[0035] The beneficial effects of the present invention include:

[0036] 1. The present invention introduces a named entity recognition (NER) model, which can automatically extract named entities such as enterprises, time, and places from news texts, and improves the structured extraction of news entity elements.

[0037] 2. The present invention applies the Transformer model architecture to realize the multi-classification model training and prediction of news public opinion risk events. It can simultaneously predict the classification probabilities of dozens of news events, output the classification probability matrix and the event category with the highest confidence, and greatly improve the accuracy and detail of news risk event identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the framework of the present invention;

[0039] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0040] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0041] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0042] like Figure 1 As shown, a system for improving the accuracy of public opinion warning by applying deep learning includes:

[0043] Named entity recognition terminal: used to obtain news text from the data input port and output the named entity matrix and its confidence probability matrix P1;

[0044] News event classification terminal: used to obtain news text from the data input port and output the risk event category E2 and its corresponding confidence probability P2;

[0045] News and public opinion warning terminal: used to prepare warning rules, receive the named entity matrix and its confidence probability matrix P1, the risk event category E2 and its corresponding confidence probability P2, and execute the warning rules to output warning signals.

[0046] Specifically, the named entity recognition terminal includes an entity element labeling module for labeling entity elements in news text, a named entity recognition model training module for training a named entity recognition model, and a named entity element extraction module for extracting a named entity matrix and its confidence probability matrix P1.

[0047] The news event classification terminal includes a news event labeling module for labeling news events, a Transformer training module for training a Transformer model, and a news event classification module for outputting a risk event category E2 and its corresponding confidence probability P2.

[0048] The news public opinion early warning terminal includes: an early warning rule configuration module and an early warning signal generation module.

[0049] The warning rule configuration module is used to configure warning rules; the front-end interface provides feature processing script input, rule configuration page (features, thresholds, logical relationships, combined rule decision trees), and the configured warning rule R1 enters the next module for execution.

[0050] The warning signal generation module is used to output a warning signal according to the named entity matrix and its confidence probability matrix P1, the risk event category E2 and its corresponding confidence probability P2. The warning signal generation module receives news text data in real time, calls the named entity element extraction module and the news event classification module in parallel, receives the news entity elements (enterprise entity, time, location) and the risk event category and confidence probability returned by the interface, executes the configured warning rule R1, and outputs a warning signal if the threshold is triggered, otherwise no output.

[0051] like Figure 2 As shown, the present invention also discloses a method for improving the accuracy of public opinion warning by applying deep learning, the method comprising:

[0052] Step 1: Acquire news text, and the named entity recognition terminal and the news event classification terminal respectively mark entity elements and risk events in the news text to obtain an entity element dataset D1 and a risk event dataset D2;

[0053] As a further preferred embodiment of this solution, step 1 specifically includes:

[0054] Step 1.1: Get news text, split the news text at the sentence level, use the shuffle function to randomly sort the sentence-level news text, and extract 50% of the news text to manually annotate the entity elements (business entity, event, time, place), and use the remaining news text for the initial version model to predict pseudo labels (pseudolabel), where news text with high confidence probability is marked as true label (truetrue label), and news samples with low confidence probability are manually corrected and annotated to obtain the entity element dataset D1; the annotated entity element dataset D1 contains news text and named entity labels (business entity, event, time, place).

[0055] Step 1.2: Get the news text, split the news text at the sentence level, use the shuffle function to randomly sort the news text at the sentence level, and extract 50% of the news text to manually annotate the risk events (the news public opinion risk event labels include dozens of event categories, such as bankruptcy reorganization, bond default, regulatory penalties, etc.), and use the remaining news text for the initial version model to predict pseudo labels (pseudo labels), where news texts with high confidence probabilities are marked as true labels, and news samples with low confidence probabilities are manually corrected and annotated to obtain the risk event dataset D2. The annotated risk event dataset D2 contains news text and risk event category labels.

[0056] Step 2: Use the entity factor dataset D1 and the risk event dataset D2 to train the named entity recognition model and the Transformer model respectively, and obtain the trained named entity recognition model M1 and the trained news public opinion risk event classification model M2;

[0057] As a further preferred embodiment of this solution, step 2 specifically includes:

[0058] Step 2.1: Import the entity feature dataset D1 into the named entity recognition model. The named entity recognition model adopts the BERT+IDCNN+CRF architecture, tokenizes the text, divides the entity feature dataset D1 into a training set and a validation set, and then loads the pre-trained model. Fine-tunes (Finetune) the entity feature dataset D, sets the appropriate initial learning rate (lr) and batch size (batch size), and obtains the trained named entity recognition model M1 after reaching the set number of iterations. Compared with the traditional BERT+BILSTM+CRF, this model structure can be trained in parallel, the speed is greatly accelerated, and holes are added to the convolution, which can expand the receptive field without pooling and obtain better long-distance information or even global information.

[0059] Step 2.2: Import the risk event dataset D2 into the Transformer model. The Transformer model adopts the BERT architecture, tokenizes the text, divides the risk event dataset D2 into a training set and a validation set, and then loads the BERT pre-trained model, initializes the learning rate (lr) and batch size (batch size), and performs fine-tuning (Finetune) on the risk event dataset D2. After reaching the set number of iterations, the trained news public opinion risk event classification model M2 (multi-class) is obtained.

[0060] Step 3: Input the news text acquired in real time into the trained named entity recognition model M1 and the news public opinion risk event classification model M2 respectively. The named entity recognition model M1 outputs the named entity matrix and its confidence probability matrix P1, and the news public opinion risk event classification model M2 outputs the risk event category E2 and its corresponding confidence probability P2;

[0061] As a further preferred embodiment of this solution, step 3 specifically includes:

[0062] Step 3.1: Obtain news text from the data input port in real time and input it into the trained named entity recognition model M1. The named entity recognition model M1 outputs a named entity matrix and its confidence probability matrix P1; the named entity matrix includes corporate entities, events, time and place;

[0063] Step 3.2: Obtain news text from the data input port in real time and input it into the trained news public opinion risk event classification model M2. The news public opinion risk event classification model M2 outputs the classification probability matrix of news public opinion events, and calculates the risk event category E2 with the highest confidence and its corresponding confidence probability P2 based on it.

[0064] Step 4: The news public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3, executes the warning rules, and outputs a warning signal.

[0065] As a further preferred embodiment of this solution, step 4 specifically includes:

[0066] Step 4.1: Configure warning rule R1 on the rule configuration page of the news and public opinion warning terminal;

[0067] Step 4.2: The news public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3;

[0068] Step 4.3: Execute the configured warning rule R1. If the threshold in the warning rule R1 is triggered, a warning signal is output, otherwise no warning signal is output.

[0069] In addition, in combination with the training method of the public opinion warning model and the public opinion warning method in the above embodiments, the embodiments of this specification may provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the training methods of the public opinion warning model or the public opinion warning method in the above embodiments is implemented.

[0070] The present invention introduces named entity recognition (NER) - a named entity recognition model, which can automatically extract named entities such as enterprises, time, and place from news texts, and improve the structured extraction of news entity elements; secondly, it applies the cutting-edge Transformer model architecture to realize the multi-classification model training and prediction of news public opinion risk events, and can simultaneously predict the classification probabilities of dozens of news events, output the classification probability matrix and the event category with the highest confidence, greatly improving the accuracy and detail of news risk event identification.

[0071] It should be clear that the present specification is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, a detailed description of the known methods is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present specification is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present specification.

[0072] It should also be noted that the exemplary embodiments mentioned in this specification describe some methods or systems based on a series of steps or devices. However, this specification is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0073] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A method for improving the accuracy of public opinion warning by applying deep learning, characterized in that: The following steps are involved: Step 1: Obtain news text, and the named entity recognition terminal and the news event classification terminal respectively mark the entity elements and risk events in the news text to obtain the entity element dataset D1 and the risk event dataset D2; Step 2: Use the entity factor dataset D1 and the risk event dataset D2 to train the named entity recognition model and the Transformer model respectively, and obtain the trained named entity recognition model M1 and the trained news public opinion risk event classification model M2; Step 2.1: Import the entity feature dataset D1 into the named entity recognition model. The named entity recognition model uses the BERT+IDCNN+CRF architecture to split the text, divide the entity feature dataset D1 into a training set and a validation set, and then load the pre-trained model, fine-tune it on the entity feature dataset D, set the appropriate initial learning rate and batch size, and obtain the trained named entity recognition model M1 after iteration; Step 2.2: Import the risk event dataset D2 into the Transformer model. The Transformer model uses the BERT architecture to split the text and divide the risk event dataset D2 into a training set and a validation set. Then load the BERT pre-trained model, initialize the learning rate and batch size, and perform fine-tuning on the risk event dataset D2. After iteration, the trained news public opinion risk event classification model M2 is obtained. Step 3: Input the news text acquired in real time into the trained named entity recognition model M1 and the news public opinion risk event classification model M2 respectively. The named entity recognition model M1 outputs the named entity matrix and its confidence probability matrix P1, and the news public opinion risk event classification model M2 outputs the risk event category E2 and its corresponding confidence probability P2; Step 4: The news and public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3, executes the warning rules, and outputs a warning signal.

2. According to claim 1, a method for improving the accuracy of public opinion warning by applying deep learning is characterized in that: The step 1 specifically includes: Step 1.1: Get the news text, split the news text at the sentence level, use the shuffle function to randomly sort the news text at the sentence level, and extract a set proportion of the news text to manually annotate the entity elements. The remaining news text is used for the initial version model to predict pseudo labels, among which the news text with high confidence probability is marked as the real label, and the news samples with low confidence probability are manually corrected and annotated to obtain the entity element dataset D1; Step 1.2: Obtain news text, split the news text at the sentence level, use the shuffle function to randomly sort the news text at the sentence level, and extract a set proportion of the news text to manually annotate the risk events. The remaining news text is used for the initial version model to predict pseudo labels, among which news text with high confidence probability is marked as the real label, and news samples with low confidence probability are manually corrected and annotated to obtain the risk event dataset D2.

3. According to claim 1, a method for improving the accuracy of public opinion warning by applying deep learning is characterized by: The step 3 specifically comprises the following steps: Step 3.1: Obtain news text from the data input port in real time and input it into the trained named entity recognition model M1. The named entity recognition model M1 outputs a named entity matrix and its confidence probability matrix P1; the named entity matrix includes corporate entities, events, time and place; Step 3.2: Obtain news text from the data input port in real time and input it into the trained news public opinion risk event classification model M2. The news public opinion risk event classification model M2 outputs the classification probability matrix of news public opinion events, and calculates the risk event category E2 with the highest confidence and its corresponding confidence probability P2 based on it.

4. According to claim 1, a method for improving the accuracy of public opinion warning by applying deep learning is characterized in that: The step 4 specifically comprises the following steps: Step 4.1: Configure warning rule R1 on the rule configuration page of the news and public opinion warning terminal; Step 4.2: The news public opinion warning terminal receives the named entity matrix and its confidence probability matrix P1 and the risk event category E2 and its corresponding confidence probability P2 in step 3; Step 4.3: Execute the configured warning rule R1. If the threshold in the warning rule R1 is triggered, a warning signal is output, otherwise no output is given.

5. A system for improving the accuracy of public opinion warning using deep learning, characterized in that: A method for improving the accuracy of public opinion warning by applying deep learning as described in any one of claims 1 to 4 is applied, comprising a named entity recognition terminal: used to obtain news text from a data input port and output a named entity matrix and its confidence probability matrix P1; News event classification terminal: used to obtain news text from the data input port and output the risk event category E2 and its corresponding confidence probability P2; News and public opinion warning terminal: used to prepare warning rules, receive the named entity matrix and its confidence probability matrix P1, the risk event category E2 and its corresponding confidence probability P2, and execute the warning rules to output warning signals.

6. According to claim 5, a system for improving the accuracy of public opinion warning by using deep learning is characterized by: The named entity recognition terminal includes an entity element labeling module for labeling entity elements in news text, a named entity recognition model training module for training a named entity recognition model, and a named entity element extraction module for extracting a named entity matrix and its confidence probability matrix P1.

7. According to claim 5, a system for improving the accuracy of public opinion warning by using deep learning is characterized by: The news event classification terminal includes a news event labeling module for labeling news events, a Transformer training module for training a Transformer model, and a news event classification module for outputting a risk event category E2 and its corresponding confidence probability P2.

8. According to claim 5, a system for improving the accuracy of public opinion warning by using deep learning is characterized by: The news public opinion early warning terminal includes: Warning rule configuration module: used to configure warning rules; Warning signal generation module: used to output warning signals based on the named entity matrix and its confidence probability matrix P1, risk event category E2 and its corresponding confidence probability P2.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of a method for applying deep learning to improve the accuracy of public opinion warnings as described in any one of claims 1-4 are implemented.

Citation Information

Patent Citations

  • Semi-supervised Chinese named entity recognition method based on deep learning

    CN108959252A

  • Product information security risk monitoring method based on semantic analysis

    CN110008311A

  • Event trigger word recognition method and device and electronic equipment

    CN110162771A

  • Event recognition method, model training method and device and storage medium

    CN110209807A

  • Enterprise news risk classification method based on target entity

    CN110502638A