WO2022001517A1 - 基于谣言预测模型的信息发送方法、装置和计算机设备 - Google Patents

基于谣言预测模型的信息发送方法、装置和计算机设备 Download PDF

Info

Publication number
WO2022001517A1
WO2022001517A1 PCT/CN2021/096236 CN2021096236W WO2022001517A1 WO 2022001517 A1 WO2022001517 A1 WO 2022001517A1 CN 2021096236 W CN2021096236 W CN 2021096236W WO 2022001517 A1 WO2022001517 A1 WO 2022001517A1
Authority
WO
WIPO (PCT)
Prior art keywords
rumor
preset
node
text
knowledge
Prior art date
Application number
PCT/CN2021/096236
Other languages
English (en)
French (fr)
Inventor
梁天恺
Original Assignee
平安国际智慧城市科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安国际智慧城市科技股份有限公司 filed Critical 平安国际智慧城市科技股份有限公司
Publication of WO2022001517A1 publication Critical patent/WO2022001517A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/334Query execution
    • G06F16/3344Query execution using natural language analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35Clustering; Classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • G06F40/289Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295Named entity recognition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/30Semantic analysis

Definitions

  • the present application relates to the technical field of artificial intelligence, and in particular, to an information sending method, device, computer equipment and storage medium based on a rumor prediction model.
  • Internet rumors are artificially created, inconsistent with the facts, and artificially promoted to spread widely and confuse the facts.
  • Internet information existing in the form of Internet articles.
  • online rumor discovery scheme is only after the large-scale dissemination of online rumors and found that it is inconsistent with the facts, so that it can be identified as online rum, so it has shortcomings such as poor timeliness.
  • extended rum referring to other rum extended from the initial online rum
  • the identification of extended rumors is less timely.
  • the existing network rumor detection scheme cannot detect network rumors in a timely manner, and it is impossible to avoid the secondary damage of extended rumor.
  • the present application proposes a method for sending information based on a rumor prediction model, comprising the following steps:
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the specified knowledge node is used as the salvage base point to perform salvage processing on the knowledge node, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • the present application provides an information sending device based on a rumor prediction model, including:
  • the click volume detection unit is configured to detect the click volume of the online article in the preset website, and determine whether the increase of the click volume of the online article within the preset time is greater than the preset growth threshold;
  • a text vector matrix obtaining unit configured to perform vectorization processing on the online article according to a preset text vectorization method if the increase in the click volume of the online article within a preset time is greater than a preset growth threshold, Thus, the text vector matrix is obtained;
  • a prediction probability value obtaining unit configured to input the text vector matrix into a preset rumor prediction model, so as to obtain a prediction probability value output by the rumor prediction model, wherein the prediction probability value refers to the online article being an online rumor the probability value;
  • a predicted probability value judgment unit configured to determine whether the predicted probability value is greater than a preset probability threshold
  • a keyword extraction unit configured to perform keyword extraction processing on the online article if the predicted probability value is greater than a preset probability threshold, so as to obtain the keywords of the online article;
  • a specified knowledge graph retrieval unit used for retrieving a specified knowledge graph from a preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • a salvage tool generation unit used for generating a knowledge node salvage tool according to the specified knowledge node by using a preset salvage tool generation method
  • a node set acquiring unit configured to use the knowledge node salvaging tool to perform salvage processing of knowledge nodes with the specified knowledge node as a salvage base point, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • the early warning information sending unit is used for sending early warning information to a preset server, and the early warning information is attached with the network article and the node set.
  • the present application provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements an information sending method based on a rumor prediction model when the processor executes the computer program, wherein the rumor-based
  • the information sending method of the prediction model includes the following steps:
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the specified knowledge node is used as the salvage base point to perform salvage processing on the knowledge node, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements a method for transmitting information based on a rumor prediction model, wherein the information transmission method based on the rumor prediction model
  • the method includes the following steps:
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the specified knowledge node is used as the salvage base point to perform salvage processing on the knowledge node, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • the information sending method, device, computer equipment and storage medium based on the rumor prediction model of the present application realize the identification of initial network rumors and the prevention of secondary rum.
  • FIG. 1 is a schematic flowchart of a method for sending information based on a rumor prediction model according to an embodiment of the present application
  • FIG. 2 is a schematic block diagram of the structure of an information sending apparatus based on a rumor prediction model according to an embodiment of the application;
  • FIG. 3 is a schematic structural block diagram of a computer device according to an embodiment of the present application.
  • Online rumor rumors have at least two characteristics, one is non-authenticity, and the other is large-scale dissemination.
  • this application adopts the detection of the degree of diffusion of network information and the verification of its authenticity (implemented by using a rumor prediction model). More special salvage tools are used to salvage in the knowledge map, so as to obtain extended rumor disasters.
  • extended rumor are based on initial network rumors, such as secondary rum prepared for artificial maliciousness, or for initial network rumors during the propagation process, due to the natural evolution of information during the propagation process and natural artificial correction (this is It is a secondary rumor that is caused by the natural attributes of human beings, so-called three people become tigers.
  • the present application is especially suitable for the prevention of extended rumors, which is a major feature of the present application.
  • the present application can be applied in any feasible field, for example, in the field of medical rum.
  • an embodiment of the present application provides a method for sending information based on a rumor prediction model, including the following steps:
  • S1 detect the click volume of the online article in the preset website, and determine whether the increase in the click volume of the online article within a preset time is greater than a preset growth threshold
  • S6 call the specified knowledge graph from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the number of clicks on the online articles in the preset website is detected, and it is determined whether the increase in the clicks of the online articles within a preset time is greater than a preset growth threshold; If the increase of the click volume within the preset time is greater than the preset growth threshold, then according to the preset text vectorization method, vectorization processing is performed on the network article, thereby obtaining a text vector matrix.
  • the purpose of detecting the number of clicks on the web article in the preset website is to determine whether the web article has a wide range of dissemination.
  • the preset text vectorization method may be any feasible method, for example, by querying a preset word vector library to obtain word vectors corresponding to each word in the online article, thereby constructing a text vector matrix; or,
  • the TF-IDF+LSA algorithm is used for text vectorization, that is, TF-IDF (which is a statistical learning method, which determines the feature value of a word by measuring the importance of a word in a certain text in a text database) ) algorithm, calculates the feature value in the network article, and extracts the word whose feature value is greater than the preset feature threshold, and denote it as the designated word. Then use the LSA algorithm to calculate the text vector matrix of all the specified words.
  • the main idea of the LSA algorithm is semantic analysis.
  • this application uses the LSA algorithm to calculate the text vector matrix for subsequent rumor identification.
  • the text vector matrix is input into the preset rumor prediction model, so as to obtain the prediction probability value output by the rumor prediction model, wherein the prediction probability value refers to the fact that the online article is an online rumor probability value.
  • the rumor prediction model is used to predict whether the input text (that is, the network article) is a rumor. It can perform supervised learning on rumors and non-rum, and identify the difference between the two and their respective feature to predict whether the incoming text is a rumor or not.
  • the rumor prediction model can be any feasible model, for example, it is obtained by training a TextCNN model based on a text classification algorithm.
  • TextCNN uses the sliding window to perform convolution processing on the input text vector through the convolution layer, compresses the feature matrix, and then further extracts the key feature points of the text through the maximum pooling layer to distinguish rumors and non-rumor texts. Finally, Through the softMax layer, the final rumor prediction result is obtained, and a rumor predictor with better performance can be formed through multiple training of the network.
  • the value range of the softMax function is [0, 1], and the output is the probability that the text is a rumor, that is, the output is a predicted probability value, where the predicted probability value refers to the probability value of the online article being an online rumor.
  • the rumor prediction model in this application can also adopt other models besides the TextCNN model.
  • the predicted probability value is greater than the preset probability threshold; if the predicted probability value is greater than the preset probability threshold, keyword extraction processing is performed on the online article, so as to obtain Keywords of the web article.
  • the predicted probability value is greater than a preset probability threshold, it indicates that the online article is a rumor.
  • the keyword extraction process is performed on the online article, so as to obtain the keywords of the online article.
  • the keywords at this time are not only the content of the initial online rumors that need early warning, but also the basis for the prediction of secondary rum.
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article; the preset salvage tool is used to generate method, generating a knowledge node salvage tool according to the specified knowledge node; using the knowledge node salvage tool, taking the specified knowledge node as a salvage base point to salvage knowledge nodes, thereby correspondingly obtaining a node set; wherein the node set At least the specified knowledge node is included.
  • This application adopts the design of the knowledge base, and more importantly, adopts a special knowledge node salvaging tool to obtain a node set, which is actually a set of initial network rumors and secondary rum.
  • the knowledge graph is a visual map, which is composed of knowledge entities (knowledge nodes) and connection relationships (entity relationships) between knowledge entities.
  • the application is preset with a knowledge graph to reflect the relationship between rum in the knowledge graph, so that when an initial rumor appears, the secondary rumor associated with it can be found in the knowledge graph, such as the initial rumor (for example, The purpose of the rumor is to affect the stock price) is that the senior management of a company is seriously ill and unable to manage (for example, the identified knowledge node includes the senior management), then the reflection of the secondary rumor in the knowledge graph is, for example, the matter that the senior management is responsible for , the corresponding other high-level and so on.
  • the present application enables identification of primary rumors and secondary rum.
  • early warning information is sent to a preset server, and the early warning information is accompanied by the network article and the node set.
  • the server is used to manage network information, for example, to limit the current, shield the network articles, etc., so as to reduce the harm of network rum.
  • the network article and the node set are attached to the early warning information, so that the present application not only realizes the damage reduction of the initial rumor (ie, the network article), but also prevents the secondary rumor in advance.
  • this application also includes: extracting place names from online articles that have been determined to be online rumors and performing descending sorting processing according to the number of occurrences of place names, to obtain a place name descending list, and sending the place name descending list to the server again, so that Prevention and control of local rum.
  • the text vector matrix is input into a preset rumor prediction model, so as to obtain a prediction probability value output by the rumor prediction model, wherein the prediction probability value means that the online article is an online rumor Before step S3 of the probability value, include:
  • E represents the expected value
  • X is the real rumor text in the rumor text set
  • pdata(x) is the distribution of the real rumor text
  • p(z) is the noise distribution
  • D is the generation model
  • G is the Describe the discriminant model
  • the real rumor text and the simulated rumor text are used together as training data for the rumor prediction model.
  • the number of non-rumor texts is much more than the number of rumor texts, so it is difficult to collect enough rumor texts as training data. Therefore, this application adopts the adversarial network model to expand the rumor text.
  • the adversarial network is mainly composed of two parts, that is, the adversarial network model includes a generative model and a discriminant model, the generative model is used to receive the input rumor text, so as to obtain the simulated rumor text, and if the discriminant model cannot Once the simulated rumor text is identified, the simulated rumor text can be used as a supplement to the rumor text to expand the training data.
  • E represents the expected value
  • X is the real rumor text in the rumor text set
  • pdata(x) is the distribution of the real rumor text
  • p(z) is the noise distribution
  • D is the generation model
  • G is the The discriminant model is described, even if the discriminative model's ability to discriminate between true and false rumors is getting worse and worse, that is, to maximize the error of the discriminant model D, and at the same time, it is hoped that the gap between the simulated rumors and the real rumors is getting smaller and smaller, that is, minimized (min) Error of generating model G.
  • the adversarial network model can output the simulated rumor text that is correspondingly fraudulent with the original rumor text; and then the real rumor text and the simulated rumor text are used together as the training data of the rumor prediction model, and the training data can be realized. Extensions to avoid model inaccuracy due to insufficient training data.
  • the step S5 of performing keyword extraction processing on the online article to obtain keywords of the online article includes:
  • the keyword extraction process is implemented on the network article, so as to obtain the keywords of the network article.
  • the present application determines the volume of the online article by calculating the total number of characters of the online article, and judging whether the total number of characters is greater than a preset threshold of the number of characters. If the volume of the online article is small, that is, the online article is short, then the online article is directly processed, or all words or most of the words in the online article can be used as keywords; otherwise, the Further processing of web articles is required. Therefore, if the total number of characters is not greater than the preset number of characters threshold, then according to the formula: Calculate the initial vector (Y1, Y2,...,Yi,...,Yn).
  • the calculation formula of the initial vector measures the importance of all words in the preset text library. Then from all the sub-vectors of the initial vector, select the designated sub-vector whose value is greater than the preset screening threshold, so the word corresponding to the designated sub-vector is an important keyword, so the word corresponding to the designated sub-vector is recorded. is the keyword of the online article.
  • the retrieving a specified knowledge graph from a preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is a keyword of the online article before step S6, including:
  • the present application may use any feasible knowledge graph construction tool, such as open source SPSS, VOSviewer, and the like.
  • the structure of the knowledge graph is the triple structure of entity-entity relationship-entity, so the entity is the node in the knowledge graph, and the entity relationship in the knowledge graph is the connection between the entity and the entity.
  • the process is, for example, line segmentation, so as to obtain a word sequence composed of multiple words, and input the word sequence into a preset sentence structure model, so as to obtain a temporary entity in the word sequence. It should be noted that this application uses a special knowledge graph.
  • the common knowledge graph does not have a high-dimensional spatial coordinate system, and this application introduces a high-dimensional spatial coordinate system. , so that in addition to the entity relationship between different knowledge nodes, there is also a spatial distance relationship, and then the constructed knowledge map is stored in the knowledge map library, so that the application can use the salvage tool to obtain more accurate knowledge from the knowledge map. fuller information.
  • the step S7 of generating a knowledge node salvage tool according to the specified knowledge node by using a preset salvage tool generation method includes:
  • Ai is the value of the i-th coordinate in the first coordinate (A1,A2,...,An)
  • Bi is the second coordinate (B1,B2,...,Bn) in the value
  • Ci is the value of the ith-dimensional coordinate in the third coordinate (C1, C2, ..., Cn)
  • a is the preset equalization parameter, a is less than 1 and greater than 0;
  • the bounded multidimensional space is a symmetrical space, the center of the bounded multidimensional space is a salvage base point, and the center of the bounded multidimensional space is away from any boundary of the bounded multidimensional space
  • the distances of the points are all equal to the salvage distance M;
  • a preset method for generating a salvage tool is implemented, and a knowledge node salvage tool is generated according to the specified knowledge node.
  • Common knowledge graphs can only rely on entity relationships to obtain associated knowledge nodes, but this application is different.
  • This application has a special design, that is, by generating a knowledge node salvage tool, so that the utilization rate of the knowledge graph is higher, and the obtained data is more accurate and comprehensive.
  • the first coordinates (A1, A2, . . .
  • the knowledge nodes obtained by using the knowledge node salvage tool include not only the knowledge nodes directly connected to the specified knowledge nodes, but also the knowledge nodes whose distance is less than the salvage distance M in the high-dimensional space, thus realizing the supplement of the non-directly connected knowledge nodes ( Because the knowledge nodes that are close to each other may not have a direct relationship, the degree of correlation is still high enough, and the possibility of secondary rumors is also high enough).
  • the value of the equalization parameter a is 0.8-0.99, preferably 0.85.
  • the information sending method based on the rumor prediction model of the present application detects the number of clicks on online articles in a preset website, and determines whether the increase in the clicks of the online articles within a preset time is greater than a preset growth threshold; If the increase in the number of hits of the online article within a preset time is greater than a preset growth threshold, then vectorize the online article to obtain a text vector matrix; input the text vector matrix into a preset rumor prediction model , so as to obtain the predicted probability value; if the predicted probability value is greater than the preset probability threshold, obtain the keywords of the online article; retrieve the specified knowledge map; generate a knowledge node salvage tool; use the knowledge node salvage tool , take the designated knowledge node as the salvage base point to perform the salvage processing of the knowledge node, so as to obtain the node set correspondingly; send early warning information to the preset server, and the early warning information is accompanied by the network article and the node set.
  • an embodiment of the present application provides an apparatus for sending information based on a rumor prediction model, including:
  • the click volume detection unit 10 is configured to detect the click volume of the online article in the preset website, and judge whether the increase of the click volume of the online article within the preset time is greater than the preset growth threshold;
  • a text vector matrix obtaining unit 20 configured to perform vectorization processing on the online article according to a preset text vectorization method if the increase in the click volume of the online article within a preset time is greater than a preset growth threshold , so as to obtain the text vector matrix;
  • the predicted probability value obtaining unit 30 is used to input the text vector matrix into a preset rumor prediction model, so as to obtain the predicted probability value output by the rumor prediction model, wherein the predicted probability value refers to the network article being a network The probability value of the rumor;
  • a predicted probability value judgment unit 40 configured to determine whether the predicted probability value is greater than a preset probability threshold
  • the keyword extraction unit 50 is configured to perform keyword extraction processing on the online article if the predicted probability value is greater than a preset probability threshold, so as to obtain the keywords of the online article;
  • the specified knowledge graph retrieval unit 60 is used to retrieve the specified knowledge graph from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • a salvage tool generation unit 70 configured to generate a knowledge node salvage tool according to the specified knowledge node by using a preset salvage tool generation method
  • the node set acquisition unit 80 is configured to use the knowledge node salvaging tool to perform knowledge node salvage processing with the specified knowledge node as a salvage base point, so as to obtain a node set correspondingly; wherein the node set at least includes the specified knowledge node. ;
  • the early warning information sending unit 90 is configured to send early warning information to a preset server, and the early warning information is accompanied by the network article and the node set.
  • the apparatus includes:
  • a model calling unit for calling a preset rumor text set and calling a preset adversarial network model; wherein the adversarial network model includes a generative model and a discriminant model;
  • a rumor data augmentation model acquisition unit used to train the adversarial network model by using the rumor text set to obtain a rumor data augmentation model; wherein the Among them, E represents the expected value, X is the real rumor text in the rumor text set, pdata(x) is the distribution of the real rumor text, p(z) is the noise distribution, D is the generation model, G is the Describe the discriminant model;
  • a simulated rumor text obtaining unit configured to input the real rumor text in the rumor text set into the rumor data expansion model, so as to obtain the simulated rumor text output by the rumor data expansion model;
  • a training data labeling unit configured to use the real rumor text and the simulated rumor text together as training data for the rumor prediction model.
  • the keyword extraction unit includes:
  • a total number of characters calculation subunit used to calculate the total number of characters of the online article, and to determine whether the total number of characters is greater than a preset number of characters threshold;
  • the initial vector calculation subunit is used for if the total number of characters is not greater than the preset number of characters threshold, according to the formula: Calculate the initial vector (Y1, Y2,...,Yi,...,Yn); wherein, Q is the total number of occurrences of all words in the online article, Qi is the ith in the online article The number of occurrences of words, W is the total number of texts contained in the preset text library, Wi is the number of texts containing the i-th word in the text library, and the online article contains n different words in total;
  • the keyword acquisition subunit is used to select a specified sub-vector whose value is greater than a preset screening threshold from all sub-vectors of the initial vector, and record the word corresponding to the specified sub-vector as the Key words.
  • the apparatus includes:
  • the temporary entity acquisition unit is used to perform entity identification processing on the pre-collected data by using a preset knowledge graph construction tool, so as to obtain a plurality of temporary entities;
  • a synonymous disambiguation unit configured to perform synonymous disambiguation processing on the multiple temporary entities to obtain multiple final entities with different semantics
  • the final entity mapping unit is used to construct a high-dimensional space coordinate system, and according to a preset coordinate point mapping method, all final entities are mapped to coordinate points in the high-dimensional space coordinate system;
  • the knowledge graph construction unit is used to propose the entity relationship between the plurality of final entities from the pre-collected data, and in the high-dimensional spatial coordinate system, the final entity-entity relationship-final entity
  • the structure forms a triple, so that a knowledge graph is constructed with the triple having high-dimensional spatial coordinate points as the basic structure, and the constructed knowledge graph is stored in the knowledge graph library.
  • the salvage tool generating unit includes:
  • the first coordinate obtaining subunit is used to obtain the first coordinates (A1, A2, . . . , An) corresponding to the specified knowledge nodes in the high-dimensional space coordinate system, wherein the high-dimensional space coordinate system is an n-dimensional coordinate system;
  • the first associated node obtaining subunit is used to obtain all the first associated nodes directly connected to the specified knowledge node in the specified knowledge graph, and select the specified knowledge node from all the first associated nodes the nearest nearby node and the farthest distant node from the specified knowledge node;
  • the salvage distance calculation subunit is used to obtain the second coordinates (B1, B2, ..., Bn) of the near nodes and the third coordinates (C1, C2, ..., Cn) of the far nodes , and according to the formula:
  • Ai is the value of the i-th coordinate in the first coordinate (A1,A2,...,An)
  • Bi is the second coordinate (B1,B2,...,Bn) in the value
  • Ci is the value of the ith-dimensional coordinate in the third coordinate (C1, C2, ..., Cn)
  • a is the preset equalization parameter, a is less than 1 and greater than 0;
  • the bounded multidimensional space generating subunit is used to generate a bounded multidimensional space, the bounded multidimensional space is a symmetric space, the center of the bounded multidimensional space is the salvage base point, and the center of the bounded multidimensional space is away from the The distance of any boundary point of the bounded multi-dimensional space is equal to the salvage distance M;
  • the knowledge node salvage tool generating subunit is used to generate the knowledge node salvage tool; wherein the knowledge node salvaged by the knowledge node salvage tool is equal to the knowledge node in the bounded multi-dimensional space and the knowledge node directly connected to the salvage base point union of .
  • an embodiment of the present invention further provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in the figure.
  • the computer device includes a processor, memory, a network interface, and a database connected by a system bus.
  • the processor of the computer design is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the nonvolatile storage medium stores an operating system, a computer program, and a database.
  • the memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer device is used for storing data used for the information sending method based on the rumor prediction model.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer program is executed by the processor to implement an information transmission method based on a rumor prediction model.
  • the above-mentioned processor executes the steps of the above-mentioned information sending method based on the rumor prediction model:
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the specified knowledge node is used as the salvage base point to perform salvage processing on the knowledge node, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • An embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium may be non-volatile or volatile, and a computer program is stored thereon, and the computer program is implemented when executed by a processor
  • the method for sending information based on a rumor prediction model shown in any of the above exemplary embodiments, the method for sending information based on a rumor prediction model includes the following steps:
  • the specified knowledge graph is retrieved from the preset knowledge graph library; wherein the specified knowledge node in the specified knowledge graph is the keyword of the online article;
  • the specified knowledge node is used as the salvage base point to perform salvage processing on the knowledge node, thereby correspondingly obtaining a node set; wherein the node set at least includes the specified knowledge node;
  • the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function, and the like; The data created by the use of the node, etc.
  • the blockchain referred to in the present invention is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • the underlying platform of the blockchain can include processing modules such as user management, basic services, smart contracts, and operation monitoring.
  • the user management module is responsible for the identity information management of all blockchain participants, including maintenance of public and private key generation (account management), key management, and maintenance of the corresponding relationship between the user's real identity and blockchain address (authority management), etc.
  • account management maintenance of public and private key generation
  • key management key management
  • authorization management maintenance of the corresponding relationship between the user's real identity and blockchain address
  • the basic service module is deployed on all blockchain node devices to verify the validity of business requests, After completing the consensus on valid requests, record them in the storage.
  • the basic service For a new business request, the basic service first adapts the interface for analysis and authentication processing (interface adaptation), and then encrypts the business information through the consensus algorithm (consensus management), After encryption, it is completely and consistently transmitted to the shared ledger (network communication), and records are stored; the smart contract module is responsible for the registration and issuance of contracts, as well as contract triggering and contract execution.
  • contract logic through a programming language and publish to On the blockchain (contract registration), according to the logic of the contract terms, call the key or other events to trigger execution, complete the contract logic, and also provide the function of contract upgrade and cancellation;
  • the operation monitoring module is mainly responsible for the deployment in the product release process , configuration modification, contract settings, cloud adaptation, and visual output of real-time status in product operation, such as: alarms, monitoring network conditions, monitoring node equipment health status, etc.
  • the present application can be applied in the field of smart cities, thereby promoting the construction of smart cities.
  • Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请涉及人工智能技术领域,可应用于智慧城市领域中,揭示了一种基于谣言预测模型的信息发送方法、装置、计算机设备和存储介质,检测网络文章的点击量;若网络文章的点击量的增幅大于预设的增长阈值,则进行向量化处理,从而得到文本向量矩阵;将所述文本向量矩阵输入谣言预测模型中,从而得到预测几率值;若预测几率值大于几率阈值,则得到关键词;调取指定知识图谱;生成知识节点打捞工具;利用知识节点打捞工具进行打捞处理,从而对应得到节点集合;向服务器发送预警信息,预警信息上附带有节点集合。从而实现了对初始网络谣言的识别,以及对二次谣言的预防。本申请还涉及区块链技术,所述谣言预测模型可存储于区块链中。

Description

基于谣言预测模型的信息发送方法、装置和计算机设备
本申请要求于2020年07月01日提交中国专利局、申请号为202010618415.0,发明名称为“基于谣言预测模型的信息发送方法、装置和计算机设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域,特别是涉及一种基于谣言预测模型的信息发送方法、装置、计算机设备和存储介质。
背景技术
网络谣言是人为制造的,与事实不符,且有人为推动以达到大范围扩散,造成混淆事实的目的的网络信息(以网络文章方式存在)。对于网络谣言,发明人意识到,现有的网络谣言发现方案,是在网络谣言大规模传播之后,发现其与事实不符,才能认定其为网络谣言,因此存在时效性差等缺点。并且,网络谣言存在多次传播,在多次传播的过程中还会逐渐滋生出与初始网络谣言不同的延伸谣言(指由初始网络谣言延伸出来的其他谣言),而现有的网络谣言发现方案对延伸谣言的识别时效性更差。
技术问题
现有的网络谣言发现方案无法及时进行网络谣言发现,更无法避免延伸谣言的二次伤害。
技术解决方案
本申请提出一种基于谣言预测模型的信息发送方法,包括以下步骤:
检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
判断所述预测几率值是否大于预设的几率阈值;
若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
本申请提供一种基于谣言预测模型的信息发送装置,包括:
点击量检测单元,用于检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
文本向量矩阵获取单元,用于若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
预测几率值获取单元,用于将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
预测几率值判断单元,用于判断所述预测几率值是否大于预设的几率阈值;
关键词提取单元,用于若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
指定知识图谱调取单元,用于从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
打捞工具生成单元,用于采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
节点集合获取单元,用于利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
预警信息发送单元,用于向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
本申请提供一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,所述处理器执行所述计算机程序时实现一种基于谣言预测模型的信息发送方法,其中,所述基于谣言预测模型的信息 发送方法包括以下步骤:
检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
判断所述预测几率值是否大于预设的几率阈值;
若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
本申请提供一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被处理器执行时实现一种基于谣言预测模型的信息发送方法,其中,所述基于谣言预测模型的信息发送方法包括以下步骤:
检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
判断所述预测几率值是否大于预设的几率阈值;
若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
有益效果
本申请的基于谣言预测模型的信息发送方法、装置、计算机设备和存储介质,实现了对初始网络谣言的识别,以及对二次谣言的预防。
附图说明
图1为本申请一实施例的基于谣言预测模型的信息发送方法的流程示意图;
图2为本申请一实施例的基于谣言预测模型的信息发送装置的结构示意框图;
图3为本申请一实施例的计算机设备的结构示意框图。
本发明的最佳实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
网络谣言至少具有两种特性,一是非真实性,二是大范围传播性,这是区别为网络真实信息的特点。为了将网络谣言准确识别出来,本申请采用对网络信息的扩散程度的检测,以及对其真实性的核实(利用谣言预测模型实现),来完成。更采用特别的打捞工具,在知识图谱中进行打捞处理,从而得到延伸谣言,从而减少谣言灾害的扩散。其中,延伸谣言是基于初始网络谣言而来的,其例如为人工恶意准备的二次谣言,或者为初始网络谣言在传播过程中,由于传播过程中信息的自然演变与自然的人为修正(这是人的自然属性所造成的,所谓三人成虎)而形成的二次谣言。本申请尤其适于对延伸谣言的预防,这是本申请的一大特点。本申请可应用于任意可行领域,例如应用于医疗谣言领域。
参照图1,本申请实施例提供一种基于谣言预测模型的信息发送方法,包括以下步骤:
S1、检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
S2、若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
S3、将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
S4、判断所述预测几率值是否大于预设的几率阈值;
S5、若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
S6、从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
S7、采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
S8、利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
S9、向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
如上述步骤S1-S2所述,检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵。其中,检测预设网站中的网络文章的点击量,其目的在于确定该网络文章是有具有大范围传播性。一般而言,若只是由于非恶意的对事实的误解,不会进行大力扩散,并且在文章发布者核实正确信息后,能够得以纠正,因此不具有大范围传播性的网络文章不被判定为网络谣言。其中,所述预设的文本向量化方法可为任意可行方法,例如为,通过查询预设的词向量库,从而获取网络文章中各个单词对应的词向量,从而构建为文本向量矩阵;或者,采用TF-IDF+LSA算法进行文本向量化,即先采用TF-IDF(其是一种统计学习方法,其通过衡量某词语在文本库的某文本中的重要性,来确定该词语的特征值)算法,计算出所述网络文章中的特征值,并将特征值大于预设的特征阈值的单词提取出来,记为指定单词。再采用LSA算法计算出所有的指定单词的文本向量矩阵,其中,所述LSA算法主要思想是语义分析,通过构建语义结构上隐含的上下文关系,找到这种潜在的语义关系,即某个词语通常与哪些词语同时出现,或者某个词语附近经常会出现哪些词语,从而本申请利用LSA算法计算出文本向量矩阵,以备后续谣言识别使用。
如上述步骤S3所述,将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值。其中,所述谣言预测模型用于对输入文本(即网络文章)进行是否为谣言的预判,其可通过对谣言以及非谣言进行有监督学习,并从中识别出两者的差异性和各自的特征,以实现对新进文本进行是否为谣言的预判。所述谣言预测模型可为任意可行模型,例如为基于文本分类算法TextCNN模型训练得到。TextCNN对输入的文本向量借助滑动窗口,通过卷积层进行卷积处理,将特征矩阵进行压缩,再通过最大池化层,进一步提取出文本的关键特征点,以区分谣言和非谣言文本,最后通过softMax层,得到最终的谣言预测结果,通过该网络多次的训练,即可形成性能较好的谣言预测器。其中softMax函数取值范围是[0,1],输出的是该文本是谣言的概率,即输出的是预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值。当然,本申请中的谣言预测模型还能采用除TextCNN模型之外的其他模型。
如上述步骤S4-S5所述,判断所述预测几率值是否大于预设的几率阈值;若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词。其中,若所述预测几率值大于预设的几率阈值,表明所述网络文章为谣言。此时,以传统的谣言预警方法而言,其将直接进行预警,但无法对二次谣言的伤害进行预防。而本申请通过进一步的设计,防止了二次谣言的出现。首先,对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词。此时的关键词,不仅是初始网络谣言需要预警的内容,更是二次谣言预测的基础。
如上述步骤S6-S8所述,从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点。本申请采用知识图库的设计,更重要的是,采用了特别的知识节点打捞工具,从而得到节点集合,而该节点集合实际上是初始网络谣言和二次谣言的集合。其中,知识图谱是一种可视化的地图,其由知识实体(知识节点)与知识实体之间的连接关系(实体关系)构成。本申请通过预设设置有知识图谱,以将谣言之间的关联关系反应在知识图谱中,从而当一个初始谣言出现后,与其关联的二次谣言能够在知识图谱中发现,例如初始谣言(例如该谣言的目的在于影响股价)为某公司管理高层病重无法理事(此时识别出的知识节点例如包括该管理高层),那么二次谣言在知识图谱中的反映例如为该管理高层负责的事项、对应的其他高层 等。从而,本申请能够进行初始谣言和二次谣言的识别。
如上述步骤S9所述,向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。其中,所述服务器用于对网络信息进行管理,例如对网络文章进行限流、屏蔽等,从而减少网络谣言的伤害。并且,所述预警信息上附带有所述网络文章和所述节点集合,以使本申请不仅实现了对初始谣言(即网络文章)的伤害减免,更对二次谣言进行了预先防备。进一步地,本申请还包括:对已被确定为网络谣言的网络文章进行地名提取并根据地名出现次数进行降序排列处理,以得到地名降序表,再次所述地名降序表发送给所述服务器,以便于地区谣言的防控。
在一个实施方式中,所述将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值的步骤S3之前,包括:
S21、调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
S22、利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述
Figure PCTCN2021096236-appb-000001
Figure PCTCN2021096236-appb-000002
其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
S23、将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
S24、将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
如上所述,实现了将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。在现实中,非谣言文本数量远多于谣言文本数量,因此要采集到足够的谣言文本作为训练数据难以实现。因此,本申请采用对抗网络模型进行谣言文本的扩充。其中,所述对抗网络主要由两部分构成,即所述对抗网络模型包括生成模型和判别模型,所述生成模型用于接收输入的谣言文本,从而得到模拟谣言文本,并且若所述判别模型无法识别出模拟谣言文本,则可将模拟谣言文本作为谣言文本的补充,从而扩展训练数据。其中,所述
Figure PCTCN2021096236-appb-000003
Figure PCTCN2021096236-appb-000004
其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型,即使判别模型对真假谣言的判别能力越来越差,即最大化(max)判别模型D的误差,同时又希望让模拟谣言与真实谣言的差距越来越小,即最小化(min)生成模型G的误差。从而对抗网络模型能够输出与原有的谣言文本具有相应欺诈性的模拟谣言文本;再将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据,即可实现训练数据扩展,以避免训练数据不足造成模型不准确的缺陷。
在一个实施方式中,所述对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词的步骤S5,包括:
S501、计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
S502、若所述总字符数不大于预设的字符数量阈值,则根据公式:
Figure PCTCN2021096236-appb-000005
计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
S503、从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
如上所述,实现了对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词。本申请通过计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值,以确定所述网络文章的体量。若所述网络文章的体量较小,即所述网络文章短小,那么对所述网络文章直接进行处理,或者将所述网络文章的所有单词或者大部分单词作为关键词即可;反之,则需要对网络文章进一步处理。 因此,若所述总字符数不大于预设的字符数量阈值,则根据公式:
Figure PCTCN2021096236-appb-000006
计算出初始向量(Y1,Y2,...,Yi,...,Yn)。其中,初始向量的计算公式衡量所有单词分别在预设的文本库的重要性。再从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,因此指定分向量对应的单词则为重要关键词,所以将所述指定分向量对应的单词记为所述网络文章的关键词。
在一个实施方式中,所述从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词的步骤S6之前,包括:
S51、采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
S52、对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
S53、构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
S54、从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
如上所述,实现了将构建得到的知识图谱存入所述知识图谱库中。本申请可采用任意可行的知识图谱构建工具,其例如为开源的SPSS、VOSviewer等。知识图谱的结构是实体-实体关系-实体的三元组结构,因此实体即是知识图谱中的节点,实体关系在知识图谱中为实体与实体的连接关系。其过程例如为:行分词处理,从而获得由多个词构成的词序列,将所述词序列输入预设的语句结构模型,从而在所述词序列中获取暂时实体。需要注意的是,本申请采用的是特别的知识图谱,其相对于普通的知识图谱最重要的一点在于,普通的知识图谱不具有高维空间坐标系,而本申请引入了高维空间坐标系,以使不同知识节点之间除了实体关系之外,还具有空间距离关系,再将构建得到的知识图谱存入所述知识图谱库中,从而使本申请采用打捞工具能够从知识图谱获取更准确更充分的信息。
在一个实施方式中,所述采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具的步骤S7,包括:
S701、在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
S702、在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
S703、获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
Figure PCTCN2021096236-appb-000007
计算出打捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
S704、生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
S705、生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
如上所述,实现了采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具。普通的知识图谱只能依靠实体关系获取关联的知识节点,而本申请不同。本申请通过特别的设计,即通过生成知识节点打捞工具,使得知识图谱的利用率更高,且获得的数据更准确且更全面。具体地,在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An);所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
Figure PCTCN2021096236-appb-000008
计算出打捞距离M;生成有界多维空间;生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。从而利用知识节点打捞工具得到的知识节点,不仅包括与指定知识节点直接相连的知识节点,还包括在高维空间内距离小于打捞距离M的知识节点,从而实现了非直连知识节点的补充(因为相离较近的知识节点,虽然可能没有直连关系,但其相关程度仍是足够高的,是二次谣言的可能性也足够高)。其中,所述均衡参数a的取值为0.8-0.99,优选0.85。
本申请的基于谣言预测模型的信息发送方法,检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则对所述网络文章进行向量化处理,从而得到文本向量矩阵;将所述文本向量矩阵输入预设的谣言预测模型中,从而得到预测几率值;若所述预测几率值大于预设的几率阈值,则得到所述网络文章的关键词;调取指定知识图谱;生成知识节点打捞工具;利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。从而实现了对初始网络谣言的识别,以及对二次谣言的预防。
参照图2,本申请实施例提供一种基于谣言预测模型的信息发送装置,包括:
点击量检测单元10,用于检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
文本向量矩阵获取单元20,用于若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
预测几率值获取单元30,用于将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
预测几率值判断单元40,用于判断所述预测几率值是否大于预设的几率阈值;
关键词提取单元50,用于若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
指定知识图谱调取单元60,用于从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
打捞工具生成单元70,用于采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
节点集合获取单元80,用于利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
预警信息发送单元90,用于向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
其中上述单元或子单元分别用于执行的操作与前述实施方式的基于谣言预测模型的信息发送方法的步骤一一对应,在此不再赘述。
在一个实施方式中,所述装置,包括:
模型调用单元,用于调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
谣言数据扩充模型获取单元,用于利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述
Figure PCTCN2021096236-appb-000009
Figure PCTCN2021096236-appb-000010
其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
模拟谣言文本获取单元,用于将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
训练数据标记单元,用于将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
其中上述单元或子单元分别用于执行的操作与前述实施方式的基于谣言预测模型的信息发送方法的步骤一一对应,在此不再赘述。
在一个实施方式中,所述关键词提取单元,包括:
总字符数计算子单元,用于计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
初始向量计算子单元,用于若所述总字符数不大于预设的字符数量阈值,则根据公式:
Figure PCTCN2021096236-appb-000011
计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
关键词获取子单元,用于从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
其中上述单元或子单元分别用于执行的操作与前述实施方式的基于谣言预测模型的信息发送方法的步骤一一对应,在此不再赘述。
在一个实施方式中,所述装置,包括:
暂时实体获取单元,用于采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
同义消歧单元,用于对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
最终实体映射单元,用于构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
知识图谱构建单元,用于从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
其中上述单元或子单元分别用于执行的操作与前述实施方式的基于谣言预测模型的信息发送方法的步骤一一对应,在此不再赘述。
在一个实施方式中,所述打捞工具生成单元,包括:
第一坐标获取子单元,用于在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
第一关联节点获取子单元,用于在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
打捞距离计算子单元,用于获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
Figure PCTCN2021096236-appb-000012
计算出打捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
有界多维空间生成子单元,用于生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
知识节点打捞工具生成子单元,用于生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
其中上述单元或子单元分别用于执行的操作与前述实施方式的基于谣言预测模型的信息发送方法的步骤一一对应,在此不再赘述。
参照图3,本发明实施例中还提供一种计算机设备,该计算机设备可以是服务器,其内部结构可以如图所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中,该计算 机设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储基于谣言预测模型的信息发送方法所用数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种基于谣言预测模型的信息发送方法。
上述处理器执行上述基于谣言预测模型的信息发送方法的步骤:
检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
判断所述预测几率值是否大于预设的几率阈值;
若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
本领域技术人员可以理解,图中示出的结构,仅仅是与本申请方案相关的部分结构的框图,并不构成对本申请方案所应用于其上的计算机设备的限定。
本申请一实施例还提供一种计算机可读存储介质,所述计算机可读存储介质可以是非易失性,也可以是易失性,其上存储有计算机程序,计算机程序被处理器执行时实现上述任一个示例性实施例所示出的基于谣言预测模型的信息发送方法,所述基于谣言预测模型的信息发送方法包括以下步骤:
检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
判断所述预测几率值是否大于预设的几率阈值;
若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
进一步地,所述计算机可读存储介质可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需的应用程序等;存储数据区可存储根据区块链节点的使用所创建的数据等。
本发明所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
区块链底层平台可以包括用户管理、基础服务、智能合约以及运营监控等处理模块。其中,用户管理模块负责所有区块链参与者的身份信息管理,包括维护公私钥生成(账户管理)、密钥管理以及用户真实身份和区块链地址对应关系维护(权限管理)等,并且在授权的情况下,监管和审计某些真实身份的交易情况,提供风险控制的规则配置(风控审计);基础服务模块部署在所有区块链节点设备上,用来验证业务请求的有效性,并对有效请求完成共识后记录到存储上,对于一个新的业务请求,基础服务先对接 口适配解析和鉴权处理(接口适配),然后通过共识算法将业务信息加密(共识管理),在加密之后完整一致的传输至共享账本上(网络通信),并进行记录存储;智能合约模块负责合约的注册发行以及合约触发和合约执行,开发人员可以通过某种编程语言定义合约逻辑,发布到区块链上(合约注册),根据合约条款的逻辑,调用密钥或者其它的事件触发执行,完成合约逻辑,同时还提供对合约升级注销的功能;运营监控模块主要负责产品发布过程中的部署、配置的修改、合约设置、云适配以及产品运行中的实时状态的可视化输出,例如:告警、监控网络情况、监控节点设备健康状态等。
本申请可应用于智慧城市领域中,从而推动智慧城市的建设。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一非易失性计算机可读取存储介质中,该计算机程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的和实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可以包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双速据率SDRAM(SSRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。

Claims (20)

  1. 一种基于谣言预测模型的信息发送方法,其中,包括:
    检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
    若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
    将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
    判断所述预测几率值是否大于预设的几率阈值;
    若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
    从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
    采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
    利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
    向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
  2. 根据权利要求1所述的基于谣言预测模型的信息发送方法,其中,所述将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值的步骤之前,包括:
    调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
    利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述对抗网络模型在训练时的
    Figure PCTCN2021096236-appb-100001
    Figure PCTCN2021096236-appb-100002
    其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
    将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
    将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
  3. 根据权利要求1所述的基于谣言预测模型的信息发送方法,其中,所述对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词的步骤,包括:
    计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
    若所述总字符数不大于预设的字符数量阈值,则根据公式:
    Figure PCTCN2021096236-appb-100003
    计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
    从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
  4. 根据权利要求1所述的基于谣言预测模型的信息发送方法,其中,所述从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词的步骤之前,包括:
    采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
    对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
    构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
    从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
  5. 根据权利要求4所述的基于谣言预测模型的信息发送方法,其中,所述采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具的步骤,包括:
    在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
    在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
    获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
    Figure PCTCN2021096236-appb-100004
    计算出打 捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
    生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
    生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
  6. 一种基于谣言预测模型的信息发送装置,其中,包括:
    点击量检测单元,用于检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
    文本向量矩阵获取单元,用于若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
    预测几率值获取单元,用于将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
    预测几率值判断单元,用于判断所述预测几率值是否大于预设的几率阈值;
    关键词提取单元,用于若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
    指定知识图谱调取单元,用于从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
    打捞工具生成单元,用于采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
    节点集合获取单元,用于利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
    预警信息发送单元,用于向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
  7. 根据权利要求6所述的基于谣言预测模型的信息发送装置,其中,所述装置,包括:
    模型调用单元,用于调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
    谣言数据扩充模型获取单元,用于利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述对抗网络模型在训练时的
    Figure PCTCN2021096236-appb-100005
    其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
    模拟谣言文本获取单元,用于将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
    训练数据标记单元,用于将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
  8. 根据权利要求6所述的基于谣言预测模型的信息发送装置,其中,所述关键词提取单元,包括:
    总字符数计算子单元,用于计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
    初始向量计算子单元,用于若所述总字符数不大于预设的字符数量阈值,则根据公式:
    Figure PCTCN2021096236-appb-100006
    计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
    关键词获取子单元,用于从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
  9. 根据权利要求6所述的基于谣言预测模型的信息发送装置,其中,所述装置,包括:
    暂时实体获取单元,用于采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
    同义消歧单元,用于对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
    最终实体映射单元,用于构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
    知识图谱构建单元,用于从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
  10. 根据权利要求9所述的基于谣言预测模型的信息发送装置,其中,所述打捞工具生成单元,包括:
    第一坐标获取子单元,用于在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
    第一关联节点获取子单元,用于在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
    打捞距离计算子单元,用于获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
    Figure PCTCN2021096236-appb-100007
    计算出打捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
    有界多维空间生成子单元,用于生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
    知识节点打捞工具生成子单元,用于生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
  11. 一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,其中,所述处理器执行所述计算机程序时实现一种基于谣言预测模型的信息发送方法:
    其中,所述基于谣言预测模型的信息发送方法包括:
    检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
    若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
    将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
    判断所述预测几率值是否大于预设的几率阈值;
    若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词;
    从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
    采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
    利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指 定知识节点;
    向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
  12. 根据权利要求11所述的计算机设备,其中,所述将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值的步骤之前,包括:
    调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
    利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述对抗网络模型在训练时的
    Figure PCTCN2021096236-appb-100008
    Figure PCTCN2021096236-appb-100009
    其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
    将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
    将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
  13. 根据权利要求11所述的计算机设备,其中,所述对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词的步骤,包括:
    计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
    若所述总字符数不大于预设的字符数量阈值,则根据公式:
    Figure PCTCN2021096236-appb-100010
    计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
    从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
  14. 根据权利要求11所述的计算机设备,其中,所述从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词的步骤之前,包括:
    采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
    对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
    构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
    从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
  15. 根据权利要求14所述的计算机设备,其中,所述采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具的步骤,包括:
    在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
    在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
    获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
    Figure PCTCN2021096236-appb-100011
    计算出打捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
    生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
    生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
  16. 一种计算机可读存储介质,其上存储有计算机程序,其中,所述计算机程序被处理器执行时实现一种基于谣言预测模型的信息发送方法,其中,所述基于谣言预测模型的信息发送方法包括以下步骤:
    检测预设网站中的网络文章的点击量,并判断所述网络文章的点击量在预设时间内的增幅是否大于预设的增长阈值;
    若所述网络文章的点击量在预设时间内的增幅大于预设的增长阈值,则根据预设的文本向量化方法,对所述网络文章进行向量化处理,从而得到文本向量矩阵;
    将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值;
    判断所述预测几率值是否大于预设的几率阈值;
    若所述预测几率值大于预设的几率阈值,则对所述网络文章进行关键词提 取处理,从而得到所述网络文章的关键词;
    从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词;
    采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具;
    利用所述知识节点打捞工具,以所述指定知识节点为打捞基点进行知识节点进行打捞处理,从而对应得到节点集合;其中所述节点集合至少包括所述指定知识节点;
    向预设的服务器发送预警信息,所述预警信息上附带有所述网络文章和所述节点集合。
  17. 根据权利要求16所述的计算机可读存储介质,其中,所述将所述文本向量矩阵输入预设的谣言预测模型中,从而得到所述谣言预测模型输出的预测几率值,其中所述预测几率值指所述网络文章为网络谣言的几率数值的步骤之前,包括:
    调用预设的谣言文本集,以及调用预设的对抗网络模型;其中所述对抗网络模型包括生成模型和判别模型;
    利用所述谣言文本集对所述对抗网络模型进行训练,以得到谣言数据扩充模型;其中,所述对抗网络模型在训练时的
    Figure PCTCN2021096236-appb-100012
    Figure PCTCN2021096236-appb-100013
    其中,E表示期望值,X为所述谣言文本集中的真实谣言文本,pdata(x)为所述真实谣文文本的分布,p(z)为噪音分布,D为所述生成模型,G为所述判别模型;
    将所述谣言文本集中的真实谣言文本输入所述谣言数据扩充模型中,以得到所述谣言数据扩充模型输出的模拟谣言文本;
    将所述真实谣言文本和所述模拟谣言文本共同作为所述谣言预测模型的训练数据。
  18. 根据权利要求16所述的计算机可读存储介质,其中,所述对所述网络文章进行关键词提取处理,从而得到所述网络文章的关键词的步骤,包括:
    计算所述网络文章的总字符数,并判断所述总字符数是否大于预设的字符数量阈值;
    若所述总字符数不大于预设的字符数量阈值,则根据公式:
    Figure PCTCN2021096236-appb-100014
    计算出初始向量(Y1,Y2,...,Yi,...,Yn);其中,Q为所述网络文章中的所有单词出现的总次数,Qi为所述网络文章中的第i个单词出现次数,W为预设的文本库中包含的文本总数量,Wi为所述文本库中存在所述第i个单词的文本的数量,所述网络文章中共包含n个不同的单词;
    从所述初始向量的所有分向量中,选出数值大于预设的筛选阈值的指定分向量,并将所述指定分向量对应的单词记为所述网络文章的关键词。
  19. 根据权利要求16所述的计算机可读存储介质,其中,所述从预设的知识图谱库中调取指定知识图谱;其中所述指定知识图谱中的指定知识节点为所述网络文章的关键词的步骤之前,包括:
    采用预设的知识图谱构建工具对预先收集的数据进行实体识别处理,从而得到多个暂时实体;
    对所述多个暂时实体进行同义消歧处理,以得到语义不同的多个最终实体;
    构建高维空间坐标系,并根据预设的坐标点映射方法,将所有的最终实体映射为所述高维空间坐标系中的坐标点;
    从所述预先收集的数据中,提出所述多个最终实体之间的实体关系,并在所述高维空间坐标系中,以最终实体-实体关系-最终实体的结构形成三元组,从而以具有高维空间坐标点的所述三元组为基本结构构建为知识图谱,并将构建得到的知识图谱存入所述知识图谱库中。
  20. 根据权利要求19所述的计算机可读存储介质,其中,所述采用预设的打捞工具生成方法,根据所述指定知识节点,生成知识节点打捞工具的步骤,包括:
    在所述高维空间坐标系中,获取所述指定知识节点对应的第一坐标(A1,A2,...,An),其中所述高维空间坐标系为n维坐标系;
    在所述指定知识图谱中,获取与所述指定知识节点直接连接的全部第一关联节点,并从全部第一关联节点中选出与所述指定知识节点最近的近处节点和与所述指定知识节点最远的远处节点;
    获取所述近处节点的第二坐标(B1,B2,...,Bn)和所述远处节点的第三坐标(C1,C2,...,Cn),并根据公式:
    Figure PCTCN2021096236-appb-100015
    计算出打捞距离M,其中Ai为第一坐标(A1,A2,...,An)中的第i维坐标的数值,Bi为第二坐标(B1,B2,...,Bn)中的第i维坐标的数值,Ci为第三坐标(C1,C2,...,Cn)中的第i维坐标的数值,a为预设的均衡参数,a小于1且大于0;
    生成有界多维空间,所述有界多维空间为对称空间,所述有界多维空间的中心为打捞基点,并所述有界多维空间的中心离所述有界多维空间的任一边界点的距离均等于所述打捞距离M;
    生成知识节点打捞工具;其中所述知识节点打捞工具打捞得到的知识节点,等于在所述有界多维空间中的知识节点和与打捞基点直接连接的知识节点的并集。
PCT/CN2021/096236 2020-07-01 2021-05-27 基于谣言预测模型的信息发送方法、装置和计算机设备 WO2022001517A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010618415.0 2020-07-01
CN202010618415.0A CN111506710B (zh) 2020-07-01 2020-07-01 基于谣言预测模型的信息发送方法、装置和计算机设备

Publications (1)

Publication Number Publication Date
WO2022001517A1 true WO2022001517A1 (zh) 2022-01-06

Family

ID=71864564

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/096236 WO2022001517A1 (zh) 2020-07-01 2021-05-27 基于谣言预测模型的信息发送方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN111506710B (zh)
WO (1) WO2022001517A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116341521A (zh) * 2023-05-22 2023-06-27 环球数科集团有限公司 一种基于文本特征的aigc文章辨识系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111506710B (zh) * 2020-07-01 2020-11-06 平安国际智慧城市科技股份有限公司 基于谣言预测模型的信息发送方法、装置和计算机设备
CN113434691B (zh) * 2021-08-26 2021-11-19 南京山猫齐动信息技术有限公司 基于分类模拟模型提升短信触达率的方法及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103258039A (zh) * 2013-05-20 2013-08-21 中国地质大学(武汉) 一种微博伪造信息的检测方法
US20170357724A1 (en) * 2016-06-14 2017-12-14 International Business Machines Corporation Model independent and network structure driven ranking of nodes for limiting the spread of misinformation through location based social networks
CN108038240A (zh) * 2017-12-26 2018-05-15 武汉大学 基于内容、用户多因素分析的社交网络谣言检测方法
IN202021002623A (zh) * 2020-01-21 2020-03-13
CN110990683A (zh) * 2019-11-29 2020-04-10 重庆邮电大学 一种基于地域与情感特征的微博谣言集成识别方法及装置
CN111506710A (zh) * 2020-07-01 2020-08-07 平安国际智慧城市科技股份有限公司 基于谣言预测模型的信息发送方法、装置和计算机设备

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002215785A (ja) * 2001-01-13 2002-08-02 Toshio Shimoyama マインドセキュリティサービス
CN110633373B (zh) * 2018-06-20 2023-06-09 上海财经大学 一种基于知识图谱和深度学习的汽车舆情分析方法
CN109271425A (zh) * 2018-09-30 2019-01-25 北京字节跳动网络技术有限公司 构建谣言数据库的方法、分析谣言数据的方法及电子设备
CN109783629A (zh) * 2019-01-16 2019-05-21 福州大学 一种融合全局事件关系信息的微博客事件谣言检测方法
CN109918648B (zh) * 2019-01-31 2020-04-21 内蒙古工业大学 一种基于动态滑动窗口特征评分的谣言深度检测方法
CN110032733A (zh) * 2019-03-12 2019-07-19 中国科学院计算技术研究所 一种针对新闻长文本的谣言检测方法及系统
CN110909125B (zh) * 2019-10-30 2022-11-15 中山大学 推文级社会媒体谣言检测方法
CN111159395B (zh) * 2019-11-22 2023-02-17 国家计算机网络与信息安全管理中心 基于图神经网络的谣言立场检测方法、装置和电子设备
CN111144131B (zh) * 2019-12-25 2021-04-30 北京中科研究院 一种基于预训练语言模型的网络谣言检测方法
CN111221941B (zh) * 2020-01-10 2023-12-26 复旦大学 基于文本内容和行文风格的社交媒体谣言鉴别算法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103258039A (zh) * 2013-05-20 2013-08-21 中国地质大学(武汉) 一种微博伪造信息的检测方法
US20170357724A1 (en) * 2016-06-14 2017-12-14 International Business Machines Corporation Model independent and network structure driven ranking of nodes for limiting the spread of misinformation through location based social networks
CN108038240A (zh) * 2017-12-26 2018-05-15 武汉大学 基于内容、用户多因素分析的社交网络谣言检测方法
CN110990683A (zh) * 2019-11-29 2020-04-10 重庆邮电大学 一种基于地域与情感特征的微博谣言集成识别方法及装置
IN202021002623A (zh) * 2020-01-21 2020-03-13
CN111506710A (zh) * 2020-07-01 2020-08-07 平安国际智慧城市科技股份有限公司 基于谣言预测模型的信息发送方法、装置和计算机设备

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116341521A (zh) * 2023-05-22 2023-06-27 环球数科集团有限公司 一种基于文本特征的aigc文章辨识系统
CN116341521B (zh) * 2023-05-22 2023-07-28 环球数科集团有限公司 一种基于文本特征的aigc文章辨识系统

Also Published As

Publication number Publication date
CN111506710A (zh) 2020-08-07
CN111506710B (zh) 2020-11-06

Similar Documents

Publication Publication Date Title
WO2020253358A1 (zh) 业务数据的风控分析处理方法、装置和计算机设备
CN111506722B (zh) 基于深度学习技术的知识图谱问答方法、装置及设备
WO2022001517A1 (zh) 基于谣言预测模型的信息发送方法、装置和计算机设备
US11475143B2 (en) Sensitive data classification
US11645528B2 (en) Continuous learning neural network system using rolling window
CN110443274B (zh) 异常检测方法、装置、计算机设备及存储介质
WO2020000688A1 (zh) 财务风险验证处理方法、装置、计算机设备及存储介质
CN112347310B (zh) 事件处理信息的查询方法、装置、计算机设备和存储介质
EP3713191B1 (en) Identifying legitimate websites to remove false positives from domain discovery analysis
CN111737493A (zh) 基于决策树的异常数据源输出方法、装置和计算机设备
CN113011973B (zh) 基于智能合约数据湖的金融交易监管模型的方法及设备
CN112464117A (zh) 请求处理方法、装置、计算机设备和存储介质
WO2022105123A1 (zh) 文本分类的方法、话题生成的方法、装置、设备及介质
WO2021004318A1 (zh) 资源数据的处理方法、装置、计算机设备和存储介质
CN113139876B (zh) 风险模型训练方法、装置、计算机设备及可读存储介质
CN110855648A (zh) 一种网络攻击的预警控制方法及装置
WO2022179046A1 (zh) 人脸识别方法、装置、计算机设备和存储介质
Zhuo et al. Network intrusion detection using word embeddings
CN112836061A (zh) 智能推荐的方法、装置以及计算机设备
CN113672654B (zh) 数据查询方法、装置、计算机设备和存储介质
CN116186759A (zh) 一种面向隐私计算的敏感数据识别与脱敏方法
CN113051372A (zh) 材料数据的处理方法、装置、计算机设备和存储介质
CN113177109A (zh) 文本的弱标注方法、装置、设备以及存储介质
CN116633653A (zh) 金融系统防火墙管理控制方法、装置、系统及存储介质
CN116186629A (zh) 基于个性化联邦学习的金融客户分类及预测方法、装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21832737

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 22/02/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21832737

Country of ref document: EP

Kind code of ref document: A1