CN113919320A - Method, system and equipment for detecting early rumors of heteromorphic neural network - Google Patents
Method, system and equipment for detecting early rumors of heteromorphic neural network Download PDFInfo
- Publication number
- CN113919320A CN113919320A CN202111284676.4A CN202111284676A CN113919320A CN 113919320 A CN113919320 A CN 113919320A CN 202111284676 A CN202111284676 A CN 202111284676A CN 113919320 A CN113919320 A CN 113919320A
- Authority
- CN
- China
- Prior art keywords
- matrix
- neural network
- node
- text
- heterogeneous
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 39
- 239000011159 matrix material Substances 0.000 claims abstract description 131
- 238000001514 detection method Methods 0.000 claims abstract description 59
- 230000007246 mechanism Effects 0.000 claims abstract description 35
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 16
- 239000002356 single layer Substances 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 abstract description 5
- 238000004590 computer program Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 13
- 230000004044 response Effects 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 4
- 230000000644 propagated effect Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 235000009508 confectionery Nutrition 0.000 description 3
- 238000010606 normalization Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000000192 social effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the disclosure provides a method, a system and equipment for detecting early rumors of a heterogeneous graph neural network, which belong to the technical field of data processing, and specifically comprise the following steps: acquiring an initial information flow, wherein the initial information flow comprises text data and a propagation structure; establishing a first matrix, and establishing a second matrix; merging the first matrix and the second matrix into a heterogeneous matrix; inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics; acquiring a text matrix according to the text data; inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics; carrying out cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested; and inputting the matrix to be detected into the single-layer feedforward neural network to obtain a detection result. According to the scheme, the heterogeneous graph attention network is introduced to carry out cascade fusion after the heterogeneous graphs and the text matrix are processed respectively, and the efficiency, the accuracy and the adaptability of early rumor detection are improved.
Description
Technical Field
The embodiment of the disclosure relates to the technical field of data processing, in particular to a method, a system and equipment for detecting early rumors of a heterogeneous graph neural network.
Background
At present, early rumor detection is to analyze the content and propagation characteristics of propagation texts on the social network, judge the rumor texts as early and accurately as possible before the rumor spreads, and prevent the rumor texts from further propagation in time, thereby reducing the probability of the users contacting false information and avoiding adverse social effects. Therefore, the research of early rumor detection is crucial to maintaining the network space environment and maintaining the stability of society. The mainstream rumor detection is mainly classified into a rumor text content-based method and a rumor propagation structure-based method. The method based on the propagation structure needs a certain propagation structure and is more suitable for rumor propagation detection in middle and later periods. However, the current mainstream detection method has some technical problems which cannot be ignored:
first, rumors are propagated according to various purposes, and the propagation process involves cyclic processes of propagators, propagation media, audiences, propagation effects and feedback. Currently, the propagation of tweets is represented by a tree structure, which is generated by capturing the interaction relationships between users triggered by source tweets, such as: twitter, where a user has a fan/friend relationship with a user, allows the user to forward or comment on another user's post. Although the propagation tree is a modeling method suitable for the global topological relation, different nodes exist in the propagation relation, such as: user, tweet and word; and there are different edges, such as: the user-tweet relation, the tweet-word relation and the common propagation tree structure are difficult to consider the overall influence factors, and the problem of poor modeling effect of the traditional model exists.
Secondly, the text of the broadcast content is unstructured data information, and vectorization conversion is needed to be performed for utilization. Although the mining of text content can to some extent be combined with contextual information to mine relationships to rumors, there is no focus on rumor text content and on propagation relationships to further explore internal relationships. The two conventional methods have a lower accuracy for rumor detection,
therefore, the existing rumor detection method has the problems of poor detection efficiency, poor detection accuracy and poor adaptability.
Disclosure of Invention
In view of the above, embodiments of the present disclosure provide a method, a system, and an apparatus for detecting an early rumor in a neural network of a heterogeneous graph, which at least partially solve the problems of poor detection efficiency, detection accuracy, and adaptability in the prior art.
In a first aspect, an embodiment of the present disclosure provides an early rumor detection method for a neural network with heterogeneous graphs, including:
acquiring an initial information flow, wherein the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node;
establishing a first matrix according to the association between the tweet nodes and the user nodes, and establishing a second matrix according to the association between the tweet nodes and the word nodes;
merging the first matrix and the second matrix into a heterogeneous matrix;
inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics;
acquiring the text matrix according to the text data;
inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics;
performing cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested;
and inputting the matrix to be detected into a single-layer feedforward neural network to obtain a detection result.
According to a specific implementation manner of the embodiment of the present disclosure, the step of establishing the first matrix according to the association between the tweet node and the user node includes:
calculating the weight relationship between the text pushing node and the user node according to the corresponding relationship of the user to the text pushing;
and forming the first matrix according to the weight relation between the text pushing node and the user node.
According to a specific implementation manner of the embodiment of the present disclosure, the step of establishing the second matrix according to the association between the tweet node and the word node includes:
calculating the weight relation between the tweet nodes and the word nodes according to the word frequency-inverse document frequency;
and forming the second matrix according to the weight relation between the tweet nodes and the word nodes.
According to a specific implementation manner of the embodiment of the present disclosure, the step of inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain a global topological feature includes:
calculating attention coefficients among different nodes in the heterogeneous matrix;
normalizing all the attention coefficients by using a preset function;
and calculating a vector corresponding to each node according to the normalized attention coefficient and forming the global topological feature.
According to a specific implementation manner of the embodiment of the present disclosure, the step of inputting the text matrix into a convolutional neural network with an attention mechanism to obtain a local information feature includes:
inputting the text matrix into the convolutional neural network with the attention mechanism to obtain embedded vectors corresponding to different words in the text data;
forming the local information features from all of the embedded vectors.
In a second aspect, an embodiment of the present disclosure provides an early rumor detection system for a neural network with heterogeneous graphs, including:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring an initial information flow, the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node;
the establishing module is used for establishing a first matrix according to the association between the tweet node and the user node and establishing a second matrix according to the association between the tweet node and the word node;
a merging module, configured to merge the first matrix and the second matrix into a heterogeneous matrix;
the first input module is used for inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics;
the extraction module is used for acquiring the text matrix according to the text data;
the second input module is used for inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics;
the cascade module is used for carrying out cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested;
and the detection module is used for inputting the matrix to be detected into the single-layer feedforward neural network to obtain a detection result.
In a third aspect, an embodiment of the present disclosure further provides an electronic device, where the electronic device includes:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method for early rumor detection in a heterogeneous graph neural network of the first aspect or any implementation of the first aspect.
In a fourth aspect, the disclosed embodiments also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method for early rumor detection of a heterogeneous graph neural network in any of the implementations of the first aspect or the first aspect.
In a fifth aspect, the embodiments of the present disclosure also provide a computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the early rumor detection method for a heterogeneous graph neural network in the first aspect or any implementation manner of the first aspect.
The early rumor detection scheme of the neural network of the heterogeneous map in the disclosed embodiments includes: acquiring an initial information flow, wherein the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node; establishing a first matrix according to the association between the tweet nodes and the user nodes, and establishing a second matrix according to the association between the tweet nodes and the word nodes; merging the first matrix and the second matrix into a heterogeneous matrix; inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics; acquiring the text matrix according to the text data; inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics; performing cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested; and inputting the matrix to be detected into a single-layer feedforward neural network to obtain a detection result.
The beneficial effects of the embodiment of the disclosure are: according to the scheme, two subgraphs are extracted after various nodes are determined from the propagation structure and are mixed into the heteromorphic graph, then the attention mechanism is introduced, the heteromorphic graph and the text matrix are respectively processed and then are subjected to cascade fusion, and the newly generated matrix to be detected is input into the single-layer feedforward neural network for detection, so that a detection result is obtained, and the efficiency, the accuracy and the adaptability of early rumor detection are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings needed to be used in the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present disclosure, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart illustrating an early rumor detection method for a neural network with heterogeneous graphs according to an embodiment of the present disclosure;
FIG. 2 is a heterogeneous diagram of a sweet-User and sweet-Word relationship involved in an early rumor detection method for a heterogeneous diagram neural network according to an embodiment of the present disclosure;
fig. 3 is a graph of an attention network involved in an early rumor detection method for a neural network of a heterogeneous map according to an embodiment of the present disclosure;
fig. 4 is a diagram illustrating a fusion manner of text and a heterogeneous graph according to an early rumor detection method of a heterogeneous graph neural network provided by the present disclosure;
fig. 5 is a diagram illustrating the detection effect of the early rumor detection method of the neural network of the heterogeneous diagram according to the embodiment of the present disclosure;
fig. 6 is a flowchart illustrating an embodiment of an early rumor detection method for a neural network with a heterogeneous graph according to the present disclosure;
fig. 7 is a schematic structural diagram of an early rumor detection system of a neural network with heterogeneous graphs according to an embodiment of the present disclosure;
fig. 8 is a schematic view of an electronic device provided in an embodiment of the present disclosure.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure in the specification. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be embodied or carried out in various other specific embodiments, and various modifications and changes may be made in the details within the description without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
It is noted that various aspects of the embodiments are described below within the scope of the appended claims. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is merely illustrative. Based on the disclosure, one skilled in the art should appreciate that one aspect described herein may be implemented independently of any other aspects and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method practiced using any number of the aspects set forth herein. Additionally, such an apparatus may be implemented and/or such a method may be practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present disclosure, and the drawings only show the components related to the present disclosure rather than the number, shape and size of the components in actual implementation, and the type, amount and ratio of the components in actual implementation may be changed arbitrarily, and the layout of the components may be more complicated.
In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects may be practiced without these specific details.
The embodiment of the disclosure provides an early rumor detection method for a heterogeneous graph neural network, which can be applied to a social network rumor detection process in a network security scene.
Referring to fig. 1, a flow chart of an early rumor detection method for a neural network with heterogeneous graphs according to an embodiment of the present disclosure is schematically shown. As shown in fig. 1, the method mainly comprises the following steps:
s101, acquiring an initial information flow, wherein the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node;
in specific implementation, when a statement published by a user on a social network needs to be detected, text data and a propagation structure which need to be detected can be used as the initial information flow, wherein the propagation structure comprises a tweet node, a user node and a word node, so that the follow-up analysis is performed on the attribute of each node and the connection between each node.
S102, establishing a first matrix according to the association between the tweet nodes and the user nodes, and establishing a second matrix according to the association between the tweet nodes and the word nodes;
in specific implementation, as shown in fig. 2, the propagation structure includes three nodes, namely the User node User, the Tweet node sweet and the Word node Word, and two relations, namely a dotted line and a solid line, between the nodes, where a topological relation between the User and the Tweet is used as a sweet-User relation, and a topological relation between the Tweet and the Word information flow is used as a sweet-Word relation. And nodes and edges with different physical meanings exist in the graph, the nodes are required to be allowed to have characteristics or attributes with different dimensions, the global topological characteristics of the initial information flow are captured by using a heteromorphic graph, and a sweet-User relation and a sweet-Word relation are modeled to obtain the first matrix and the second matrix.
S103, merging the first matrix and the second matrix into a heterogeneous matrix;
in specific implementation, considering that the Tweet, the User and the Word are nodes with different attributes, and two relations of two-Word and two-User exist in the nodes, the common graph structure is difficult to process the complex structures, and different characteristic information is reserved. The modeling mode based on the heteromorphic graph is more comprehensive in contained information and richer in semantics, and the propagation relation can be better utilized. Two adjacent matrices are merged into one matrix as shown in fig. 2, and a representation of the heteromorphic image is obtained by the adjacent matrix.
S104, inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics;
in specific implementation, considering that the heterogeneous matrix is fused with more texts, structures and other information, and much noise is fused, so that the judgment performance of the model is reduced, the heterogeneous matrix can be input into a graph neural network with an attention mechanism as shown in fig. 3, so that the global topological characteristic is obtained, and the detection performance is improved.
S105, acquiring the text matrix according to the text data;
meanwhile, after the text data is acquired, the text matrix can be directly extracted and acquired.
S106, inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics;
after the text matrix is obtained, the text matrix can also be input into a convolutional neural network with an attention mechanism, and after interference information is screened out through the attention mechanism, local information characteristics are obtained.
S107, carrying out cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested;
as shown in fig. 4, after the global topology feature and the local information feature are obtained, the rumor propagation structure information and the rumor text information may be fused in a data fusion manner, that is, the global topology feature and the local information feature are cascaded, and the characteristics of the propagation structure and the text information are combined to obtain the matrix to be measured.
And S108, inputting the matrix to be detected into the single-layer feedforward neural network to obtain a detection result.
In specific implementation, after the matrix to be detected is obtained, the matrix to be detected may be input to the single-layer feedforward neural network to obtain a detection result, and of course, before the matrix to be detected is input to the single-layer feedforward neural network to obtain the detection result, the single-layer feedforward neural network may be trained by using a plurality of sample data sets, for example, a sample data set may include four kinds of tags: non-rumors, fake, true, and unverified rumors, the output test result may be any of the four labels, or may be set according to actual requirements. Meanwhile, under different detection delays, the detection effect corresponding to the method provided by the embodiment of the disclosure is shown in fig. 5, and the specific flow of the embodiment of the disclosure is shown in fig. 6.
In the early rumor detection method for the neural network with the heterogeneous graphs, two subgraphs are extracted after various nodes are determined from a propagation structure and are mixed into a heterogeneous graph, then an attention mechanism is introduced, the heterogeneous graph and a text matrix are respectively processed and then are subjected to cascade fusion, and a newly generated matrix to be detected is input into a single-layer feedforward neural network for detection, so that a detection result is obtained, and the efficiency, the accuracy and the adaptability of early rumor detection are improved.
On the basis of the foregoing embodiment, the step S102 of establishing a first matrix according to the association between the tweet node and the user node includes:
calculating the weight relationship between the text pushing node and the user node according to the corresponding relationship of the user to the text pushing;
and forming the first matrix according to the weight relation between the text pushing node and the user node.
For example, a heterogeneous graph includes three nodes, each tweet node being defined as TiThe set of tweets is defined as D, D ═ T1,T2,…,Tn}. User node uiMeaning, U represents a set of users, U ═ U1,u2,…,up}。wjRepresenting word nodes. Propagation structure of Tweet population GiIs represented by Gi={ri,u1i,u2i,…,umi},riThe rumor of rumors, ukiIs concerned with the user's response to this rumor. The label of Tweet is denoted by Y, { Y ═ Y1,y2,…ynIn which y isiIs for the tweet TiThe label of (2) comprises four kinds of labels: "non-rumors", "fake", "true", "unproven rumors". To better exploit propagation relationships and rumor text, we will extract the relationships of sweet-User and sweet-Word from the dataset based on the above definitions.
The calculation of the sweet-User weight relation is based on the response relation of the User to the Tweet and the response time. Generally, the response relation of a user to a Tweet comprises three types, namely love, forwarding and comment, and the time when the response occurs is recorded, such as: the tweets are at 00: 00: 00, user is released at 02: 00: when 00 hours (after 120 minutes), forwarding the message, recording the response relation of the user to the message as forwarding and the relative response time as 120 minutes, and utilizing the response relation to the message in the weight calculation of the edge, wherein the specific calculation formula is as follows:
where t represents the relative time of forwarding, like or commenting of a user on Tweet, AijRepresenting the relationship weight between the tweet and the user. Thus, we can use the tuple pi=(Ti,uji,Aij) To represent the sweet-User relationship. The two-User graph may be represented by an adjacency matrix p ═ p1,p2,…pnDenotes the first matrix.
Further, in step S102, establishing a second matrix according to the association between the tweet node and the word node includes:
calculating the weight relation between the tweet nodes and the word nodes according to the word frequency-inverse document frequency;
and forming the second matrix according to the weight relation between the tweet nodes and the word nodes.
For example, the computation of the sweet-Word weight relationship is based on the Word frequency-inverse document frequency (TF-IDF), where the importance of a Word increases in proportion to the number of times it appears in the entire training set, to evaluate how important a Word is for a piece of tweed. The word frequency is the number of times the word appears in the training set, the inverse document frequency is the total number of documents divided by the number of documents containing the word, and the complete calculation method is as follows:
TF-IDFij=TFij×IDFj (2)
wherein, TFijRepresents wjWord frequency, IDF, of wordsjRepresents wjThe inverse document frequency of the word.
Wherein n isijRepresents wjWording file TiNumber of occurrences in, denominator ∑knikThe sum of the occurrences of all the words in the document T is shown.
The IDF inverse document frequency is defined as:
wherein, | τ | represents the total number of documents in the corpus, | { k: wj∈tkDenotes the inclusion of the word wjNumber of files (i.e., n)ijNumber of files not equal to 0). If the word never appears, it results in a denominator of zero, so 1+ | { k: w ] is typically usedj∈tk}|。
Then we can go through the triplet (T)i,wj,TF-IDFij) To represent the relationship between two-Word, wherein TiRepresents Tweet i, wjRepresenting words, TF-IDFijRepresents wjMerging all the two-Word subgraphs into one large two-Word graph, i.e. the second matrix, represented by the adjacency matrix R, R ═ R { (R)1,R2,…,Rm}。
On the basis of the foregoing embodiment, step S104 includes inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain a global topological feature, where the method includes:
calculating attention coefficients among different nodes in the heterogeneous matrix;
normalizing all the attention coefficients by using a preset function;
and calculating a vector corresponding to each node according to the normalized attention coefficient and forming the global topological feature.
For example, the above steps will get Twitter-Word sub-graph R ═ { R }1,R2,…,RmAnd Twitter-User graph P ═ P { (P)1,P2,…,PnCombine them intoThe heterogeneous matrix is used for obtaining the representation of the overall propagation diagram. Then, in the heterogeneous graph neural network, the weight of the adjacent node features completely depends on the node features, an attention mechanism is applied to learn the importance of the adjacent nodes and to perform weighted summation on the adjacent node features, and vector representations of the sweet including sweet-User relation and sweet-Word relation are obtained.
Inputting an adjacency matrix of the overall propagation diagram, wherein the adjacency matrix is defined as g, and g is { g }1,g2,…,gn},giFor the nodes of the graph, i.e., the embedded vectors, N represents the number of nodes and F is the vector dimension for each node. The specific calculation steps of the attention machine mechanism are as follows:
first, the attention coefficient may be calculated first, and the importance between nodes is determined:
eij=α(Wgi,Wgj) (5)
wherein e isijRepresents gjNode pair giThe importance of the node is such that,w is a weight matrix shared by all graph nodes for converting the input node features to high-level features, F is the vector dimension of each node, and α is a function of the shared attention mechanism. We need to compute giWith all neighboring nodes gjAttention coefficient therebetween, i.e.Represents giAll the neighbor nodes.
Second, the attention coefficient e may be expressed using a preset function, such as a softmax function, in consideration of the need to make the attention coefficient easy to compare between different nodesijNormalization, the concrete formula is as follows:
wherein alpha isijThe attention coefficient after normalization is applied to the calculation of a new feature vector in the next step, and the specific calculation formula is as follows:
wherein, g'iRepresentative node gjThe vector representation after the map attention mechanism update finally outputs a new adjacency matrix g ', g ' ═ g '1,g'2,…,g'NAnd the obtained data is taken as the global topological characteristic.
On the basis of the foregoing embodiment, step S106 is to input the text matrix into a convolutional neural network with attention mechanism to obtain a local information feature, and includes:
inputting the text matrix into the convolutional neural network with the attention mechanism to obtain embedded vectors corresponding to different words in the text data;
forming the local information features from all of the embedded vectors.
For example, a rumor text is passed through a CNN network with an attention mechanism to obtain a vector with context semantic features to represent the sweet, and a propagated local information feature e is obtainedi,ei={vec1,vec2,…,vecm},vecjAn embedded vector representing a word j in the text, and then forming the local information features from all of the embedded vectors.
In correspondence with the above method embodiment, referring to fig. 7, the disclosed embodiment further provides an early rumor detection system 70 of a heterogeneous graph neural network, comprising:
an obtaining module 701, configured to obtain an initial information stream, where the initial information stream includes text data and a propagation structure, and the propagation structure includes a tweet node, a user node, and a word node;
an establishing module 702, configured to establish a first matrix according to the association between the tweet node and the user node, and establish a second matrix according to the association between the tweet node and the word node;
a merging module 703, configured to merge the first matrix and the second matrix into a heterogeneous matrix;
a first input module 704, configured to input the heterogeneous matrix into a graph neural network with an attention mechanism, so as to obtain a global topological feature;
an extracting module 705, configured to obtain the text matrix according to the text data;
a second input module 706, configured to input the text matrix into a convolutional neural network with an attention mechanism, so as to obtain a local information feature;
a cascading module 707, configured to perform a cascading operation on the global topology feature and the local information feature to obtain a matrix to be tested;
and the detection module 708 is configured to input the matrix to be detected into the single-layer feedforward neural network to obtain a detection result.
The system shown in fig. 7 may correspondingly execute the content in the above method embodiment, and details of the part not described in detail in this embodiment refer to the content described in the above method embodiment, which is not described again here.
Referring to fig. 8, an embodiment of the present disclosure also provides an electronic device 80, which includes: at least one processor and a memory communicatively coupled to the at least one processor. Wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method for early rumor detection in a neural network of a heterogeneous graph according to the above method embodiments.
The disclosed embodiments also provide a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the early rumor detection method of the heterogeneous graph neural network in the aforementioned method embodiments.
The disclosed embodiments also provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the early rumor detection method of a heterogeneous graph neural network in the aforementioned method embodiments.
Referring now to FIG. 8, a block diagram of an electronic device 80 suitable for use in implementing embodiments of the present disclosure is shown. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., car navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 8, the electronic device 80 may include a processing means (e.g., a central processing unit, a graphics processor, etc.) 801 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)802 or a program loaded from a storage means 808 into a Random Access Memory (RAM) 803. In the RAM 803, various programs and data necessary for the operation of the electronic apparatus 80 are also stored. The processing apparatus 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to bus 804.
Generally, the following devices may be connected to the I/O interface 805: input devices 806 including, for example, a touch screen, touch pad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, or the like; output devices 807 including, for example, a Liquid Crystal Display (LCD), speakers, vibrators, and the like; storage 808 including, for example, magnetic tape, hard disk, etc.; and a communication device 809. The communication means 809 may allow the electronic device 80 to communicate wirelessly or by wire with other devices to exchange data. While the figures illustrate an electronic device 80 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication means 809, or installed from the storage means 808, or installed from the ROM 802. The computer program, when executed by the processing apparatus 801, performs the above-described functions defined in the methods of the embodiments of the present disclosure.
It should be noted that the computer readable medium in the present disclosure can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the present disclosure, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, either in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device.
The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the steps associated with the method embodiments.
Alternatively, the computer readable medium carries one or more programs which, when executed by the electronic device, enable the electronic device to perform the steps associated with the method embodiments.
Computer program code for carrying out operations for aspects of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware.
It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.
The above description is only for the specific embodiments of the present disclosure, but the scope of the present disclosure is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present disclosure should be covered within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.
Claims (7)
1. An early rumor detection method for a heterogeneous graph neural network, comprising:
acquiring an initial information flow, wherein the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node;
establishing a first matrix according to the association between the tweet nodes and the user nodes, and establishing a second matrix according to the association between the tweet nodes and the word nodes;
merging the first matrix and the second matrix into a heterogeneous matrix;
inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics;
acquiring the text matrix according to the text data;
inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics;
performing cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested;
and inputting the matrix to be detected into a single-layer feedforward neural network to obtain a detection result.
2. The method of claim 1, wherein the step of establishing a first matrix based on the association of the tweet node with the user node comprises:
calculating the weight relationship between the text pushing node and the user node according to the corresponding relationship of the user to the text pushing;
and forming the first matrix according to the weight relation between the text pushing node and the user node.
3. The method of claim 1, wherein the step of building a second matrix based on the association of the tweet node with the word node comprises:
calculating the weight relation between the tweet nodes and the word nodes according to the word frequency-inverse document frequency;
and forming the second matrix according to the weight relation between the tweet nodes and the word nodes.
4. The method of claim 1, wherein the step of inputting the heterogeneous matrix into a graph neural network with attention mechanism to obtain global topological features comprises:
calculating attention coefficients among different nodes in the heterogeneous matrix;
normalizing all the attention coefficients by using a preset function;
and calculating a vector corresponding to each node according to the normalized attention coefficient and forming the global topological feature.
5. The method of claim 1, wherein the step of inputting the text matrix into a convolutional neural network with attention mechanism to obtain local information features comprises:
inputting the text matrix into the convolutional neural network with the attention mechanism to obtain embedded vectors corresponding to different words in the text data;
forming the local information features from all of the embedded vectors.
6. An early rumor detection system for a heterogeneous neural network, comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring an initial information flow, the initial information flow comprises text data and a propagation structure, and the propagation structure comprises a tweet node, a user node and a word node;
the establishing module is used for establishing a first matrix according to the association between the tweet node and the user node and establishing a second matrix according to the association between the tweet node and the word node;
a merging module, configured to merge the first matrix and the second matrix into a heterogeneous matrix;
the first input module is used for inputting the heterogeneous matrix into a graph neural network with an attention mechanism to obtain global topological characteristics;
the extraction module is used for acquiring the text matrix according to the text data;
the second input module is used for inputting the text matrix into a convolutional neural network with an attention mechanism to obtain local information characteristics;
the cascade module is used for carrying out cascade operation on the global topological characteristic and the local information characteristic to obtain a matrix to be tested;
and the detection module is used for inputting the matrix to be detected into the single-layer feedforward neural network to obtain a detection result.
7. An electronic device, characterized in that the electronic device comprises:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method for early rumor detection in a heterogeneous neural network of any of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111284676.4A CN113919320A (en) | 2021-11-01 | 2021-11-01 | Method, system and equipment for detecting early rumors of heteromorphic neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111284676.4A CN113919320A (en) | 2021-11-01 | 2021-11-01 | Method, system and equipment for detecting early rumors of heteromorphic neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113919320A true CN113919320A (en) | 2022-01-11 |
Family
ID=79244838
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111284676.4A Pending CN113919320A (en) | 2021-11-01 | 2021-11-01 | Method, system and equipment for detecting early rumors of heteromorphic neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113919320A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115114500A (en) * | 2022-07-06 | 2022-09-27 | 上海交通大学 | Rumor detection method and system based on reported information and propagation heteromorphic graph |
CN115118491A (en) * | 2022-06-24 | 2022-09-27 | 北京天融信网络安全技术有限公司 | Botnet detection method and device, electronic device and readable storage medium |
CN116386895A (en) * | 2023-04-06 | 2023-07-04 | 之江实验室 | Epidemic public opinion entity identification method and device based on heterogeneous graph neural network |
-
2021
- 2021-11-01 CN CN202111284676.4A patent/CN113919320A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115118491A (en) * | 2022-06-24 | 2022-09-27 | 北京天融信网络安全技术有限公司 | Botnet detection method and device, electronic device and readable storage medium |
CN115118491B (en) * | 2022-06-24 | 2024-02-09 | 北京天融信网络安全技术有限公司 | Botnet detection method, device, electronic equipment and readable storage medium |
CN115114500A (en) * | 2022-07-06 | 2022-09-27 | 上海交通大学 | Rumor detection method and system based on reported information and propagation heteromorphic graph |
CN116386895A (en) * | 2023-04-06 | 2023-07-04 | 之江实验室 | Epidemic public opinion entity identification method and device based on heterogeneous graph neural network |
CN116386895B (en) * | 2023-04-06 | 2023-11-28 | 之江实验室 | Epidemic public opinion entity identification method and device based on heterogeneous graph neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111274815B (en) | Method and device for mining entity focus point in text | |
CN110633423B (en) | Target account identification method, device, equipment and storage medium | |
WO2022121801A1 (en) | Information processing method and apparatus, and electronic device | |
CN110275962B (en) | Method and apparatus for outputting information | |
CN113919320A (en) | Method, system and equipment for detecting early rumors of heteromorphic neural network | |
CN114090779B (en) | Method, system, device and medium for classifying chapter-level texts by hierarchical multi-labels | |
CN111857720B (en) | User interface state information generation method and device, electronic equipment and medium | |
CN111460288B (en) | Method and device for detecting news event | |
CN111738316B (en) | Zero sample learning image classification method and device and electronic equipment | |
US20230315990A1 (en) | Text detection method and apparatus, electronic device, and storage medium | |
CN110457325B (en) | Method and apparatus for outputting information | |
CN116894188A (en) | Service tag set updating method and device, medium and electronic equipment | |
CN112182255A (en) | Method and apparatus for storing media files and for retrieving media files | |
US11437038B2 (en) | Recognition and restructuring of previously presented materials | |
CN114995691B (en) | Document processing method, device, equipment and medium | |
CN111382365B (en) | Method and device for outputting information | |
CN110737820B (en) | Method and apparatus for generating event information | |
CN110659208A (en) | Test data set updating method and device | |
CN113807056B (en) | Document name sequence error correction method, device and equipment | |
WO2021258972A1 (en) | Video retrieval method and apparatus, and electronic device and computer readable medium | |
CN111581455B (en) | Text generation model generation method and device and electronic equipment | |
CN114490946A (en) | Xlnet model-based class case retrieval method, system and equipment | |
CN111027281B (en) | Word segmentation method, device, equipment and storage medium | |
CN113806556A (en) | Method, device, equipment and medium for constructing knowledge graph based on power grid data | |
US11042808B2 (en) | Predicting activity consequences based on cognitive modeling |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |