CN112711951A - Induction consciousness-based false news interpretability detection system and method - Google Patents
Induction consciousness-based false news interpretability detection system and method Download PDFInfo
- Publication number
- CN112711951A CN112711951A CN202110008758.XA CN202110008758A CN112711951A CN 112711951 A CN112711951 A CN 112711951A CN 202110008758 A CN202110008758 A CN 202110008758A CN 112711951 A CN112711951 A CN 112711951A
- Authority
- CN
- China
- Prior art keywords
- comment
- node
- news
- nodes
- tree network
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 22
- 238000000034 method Methods 0.000 title claims abstract description 21
- 230000006698 induction Effects 0.000 title claims abstract description 14
- 238000010521 absorption reaction Methods 0.000 claims abstract description 6
- 238000012795 verification Methods 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 33
- 238000004590 computer program Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 11
- 238000012552 review Methods 0.000 claims description 10
- 239000002131 composite material Substances 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000007670 refining Methods 0.000 claims description 2
- 238000005065 mining Methods 0.000 abstract description 6
- 230000007547 defect Effects 0.000 abstract description 2
- 230000007812 deficiency Effects 0.000 abstract description 2
- 238000003062 neural network model Methods 0.000 abstract description 2
- 238000011160 research Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241000700605 Viruses Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000010429 evolutionary process Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- 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
- G06F40/216—Parsing using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Probability & Statistics with Applications (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a false news interpretability detection system and method based on induction consciousness, which are used for discovering a valuable comment subchain of questioning news by exploring the evolution rule and the hierarchical structure characteristics of a place in a comment thread and overcoming the defects of relevance mining and mutual influence deficiency among comments in the prior art. According to the method, interdisciplinary knowledge is combined with a neural network model, and semantic association mining and mutual influence mining among comment nodes in a false news comment tree are explored; the bottom-up evolutionary tree network considers two social psychology theories to research the evolution rule of the ground in the comment clues so as to strengthen the valuable comments of the questioned news; the top-down coordination tree network provided by the invention coordinates the information absorption between the father node and the brother node, enhances the hierarchical structure of the comment, fuses the comment with the characteristics of the evolutionary tree network to obtain a valuable comment, highlights the child chain of the comment, and enhances the interpretability of the verification result.
Description
Technical Field
The invention relates to a system and a method for detecting false news interpretability based on perceptual evolution of inductive consciousness and a coordination tree network.
Background
The development of social media opens a convenient door to the production and dissemination of false news. The curated and packaged fake news is more likely to attract the attention and consumption of the audience than is true and reliable news content. The world health organization has always defined this problem as an information epidemic, which calls fake news to be more quickly and easily disseminated than the virus itself. Therefore, how to detect and analyze the fake news and take intervention measures to reduce the spread of the fake news becomes one of the urgent problems in the current social media field.
A typical false news detection method usually extracts text features of news content, such as semantic, emotion, and genre, and performs binary classification by supervised learning (e.g., CNN and RNN). Due to the abundance of information content in social media, multi-modal semantic information such as images and videos and various context information appear. Researchers have also studied user portrayal features and forward propagation features. Furthermore, user reviews have proven useful as supplementary semantics to improve the performance of false news detection, especially where the standpoint of reviews is widely utilized. Currently, a common approach is to construct a multi-task learning model between false news detection and review standpoint detection to capture common features between the two tasks, thereby improving model performance. However, although effective, the multitask learning method for detecting false news based on the comment standpoint often treats comments in the comment tree as independent individuals, and lacks effective exploration of the correlation between comments because the subsequent comment semantics and posture may be affected or interfered by current or previous comments. In particular, they ignore the relationships between comments in the comment conversation thread, including the hierarchy of comments and the evolutionary process of the standpoint of comments.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a false news interpretability detection system and method based on induction consciousness. The invention not only improves the false news detection performance, but also provides the transparency of the detection process and the interpretability of the detection result.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a false news interpretability detection method based on induction consciousness comprises the following steps:
step 1, learning each tweet t by utilizing Bi-GRUiAnd using the last step to conceal the layer vector
Representing the tweet;
step 2, establishing a bottom-up evolution tree recursive network, gradually combining child nodes to father nodes, and finally forming a vector representing the whole comment thread; traversing the whole comment conversation thread from bottom to top in a reverse time sequence to obtain a comment sequence containing a plurality of comment subchains;
step 3, considering the relevance between a single comment position and the comment subchain where the single comment position is located, and evaluating the relation between the current node and the previous node by means of social psychology knowledge, namely, a utilization superposition theory and a conservative deviation theory;
step 4, constructing a significance scoring function s (-) and a composite vector p (-) to calculate the merging degree between the ith node and the previous (i-1) nodes in one comment subchain; the scoring function s (-) is used for measuring the degree of the ith node in the comment subchain needing to be strengthened, and the synthetic vector p (-) represents the fusion degree of the two nodes;
and 5, refining the significance score based on two social psychology theories:
1) probability τ of two adjacent nodes merging:
τ(ai，ai-1)＝ai TSai-1 (1)
2) semantic differences between nodes and their children chains
Wherein the content of the first and second substances,
representing the evaluation of semantic difference of the ith node and the previous (i-1) node;
mean semantics representing the preceding (i-1) node;
3) review the differences between the sites φ:
where φ (-) represents the difference between the review standpoint of the ith node and the previous (i-1) node; diA position vector representing the ith node,
an average vector representing the previous (i-1) nodes;
4) timeliness of the nodes ψ:
ψ(ti)＝gti -α-1 (4)
wherein, tiIs the interval between the ith node and the first node in the sub-chain of comments; g and alpha are both learnable parameters;
step 7, calculating a synthetic vector p (·):
p(i，i-1)＝s(i，i-1)m(i，i-1) (7)
step 8, merging all nodes in the whole comment subchain as follows:
step 9, utilizing TreeLSTM to realize a composite function m (·); TreeLSTM generates a hidden state vector h' and, given two input vectors aiAnd ajThe following cell state vector c' yields:
c′＝c′i⊙fi+c′j⊙fj+x⊙u (10)
h′＝o′+tanh(c′) (11)
wherein U and b are trainable parameters, which indicate a dot product operation;
step 10, traversing the whole comment sequence containing a plurality of comment subchains, and after evolution and combination, highlighting valuable comments of questioning news in the comment threads through the evolution tree network, wherein the final learnable expression is marked as BE.
Step 11, constructing a top-down coordination tree network to coordinate information absorption between a father node and a brother node;
step 12, updating the coordination tree network:
subscripts in, fo, hf, c and o are respectively used for an input gate, a temporary forgetting gate, a hierarchical forgetting gate, a cell unit and an output; vector int，fotAnd hftRespectively representing the weight of the newly captured information, memorizing old information from brother nodes, and memorizing old information from father nodes;
representing a hadamard product; if it is not
Then hp(t)And cp(t)Is set to an initial state value;
step 13, after traversing, obtaining the representation of the whole conversation thread, and recording the representation as TC;
step 14, integrating the comment of the questioning news acquired by the evolvement tree network and the hierarchical structure information of the comment acquired by the coordination tree network to obtain a valuable comment subchain of the questioning news:
F＝[BE；TC] (15)。
the invention further improves the following steps:
the specific method of the step 11 is as follows:
traversing the entire annotation tree from top to bottom in time order, assuming that π (t) represents the parent of t, k (t) represents its children, p (t) is the previous sibling at the same level in time, and s (t) represents its successor at the same level.
A induction awareness based false news interpretability detection system, comprising:
an input embedding module to learn a contextual semantic hidden layer representation of a conversation;
an evolutionary tree network for capturing valuable reviews that challenge news;
the coordination tree network is used for amplifying the comment subchains and explaining verification results;
and the task learning module integrates the position evolution semantics obtained by the evolution tree and the comment hierarchical structure semantics captured by the coordination tree network and inputs the position evolution semantics and the comment hierarchical structure semantics into Softmax to predict probability distribution.
A induction awareness based false news interpretability detection terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor when executing the computer program implementing the steps of the method as described above.
A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method as described above.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses a false news interpretability detection method based on induction consciousness perception evolution and coordination tree network, which discovers a valuable comment subchain of questioning news by exploring the evolution rule and the hierarchical structure characteristics of a place in a comment thread, and overcomes the defects of relevance mining and mutual influence deficiency among comments in the prior art.
The invention provides a mode of combining interdisciplinary knowledge and a neural network model for the first time, semantic association mining and mutual influence mining among comment nodes in a false news comment tree are explored, and theoretical support of interpretable false news detection is effectively provided; the bottom-up evolutionary tree network considers two social psychology theories to research the evolution rule of the ground in the comment clues so as to strengthen the valuable comments of the questioned news; the top-down coordination tree network provided by the invention coordinates the information absorption between the father node and the brother node, enhances the hierarchical structure of the comment, fuses the comment with the characteristics of the evolutionary tree network to obtain a valuable comment, highlights the child chain of the comment, and enhances the interpretability of the verification result; extensive experiments on both real data sets confirmed the better performance of the present invention compared to the most advanced models.
Drawings
In order to more clearly explain the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is an architectural diagram of the present invention;
FIG. 2 is a graph of experimental performance of the present invention under two data sets, RumourEval and PHEME;
FIG. 3 is a graph comparing the separation performance of different modules of the present invention under both RumourEval and PHEME data sets.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. are used for indicating the orientation or positional relationship based on the orientation or positional relationship shown in the drawings or the orientation or positional relationship which is usually arranged when the product of the present invention is used, the description is merely for convenience and simplicity, and the indication or suggestion that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and thus, cannot be understood as limiting the present invention. Furthermore, the terms "first," "second," and the like are used merely to distinguish one description from another, and are not to be construed as indicating or implying relative importance.
Furthermore, the term "horizontal", if present, does not mean that the component is required to be absolutely horizontal, but may be slightly inclined. For example, "horizontal" merely means that the direction is more horizontal than "vertical" and does not mean that the structure must be perfectly horizontal, but may be slightly inclined.
In the description of the embodiments of the present invention, it should be further noted that unless otherwise explicitly stated or limited, the terms "disposed," "mounted," "connected," and "connected" should be interpreted broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
The invention is described in further detail below with reference to the accompanying drawings:
referring to fig. 1, an embodiment of the present invention discloses a false news interpretability detection system based on induction awareness, including:
an input embedding module: considering an embedded representation of a real Twitter dialog tree process, the sequence is expanded in time sequence and Bi-GRU is adopted to learn the semantic hidden layer representation of the context.
Evolving the tree network from bottom to top: in order to explore the evolution law of the comment position, a low-upward evolutionary tree network is established, and two social psychology theories including availability superposition and conservative bias are integrated to capture valuable comments which question news.
Top-down coordination tree network: in order to strengthen the hierarchical relationship of the comment threads, a top-down coordination tree network is provided to adjust the absorption of information semantics between a father node and a brother node, so as to enlarge the comment child chain and explain the verification result.
A task learning module: integrating the vertical evolution semantics obtained by the evolution tree with the comment hierarchical structure semantics captured by the coordination tree network, and inputting the integrated vertical evolution semantics into Softmax to predict probability distribution.
The embodiment of the invention discloses a false news interpretability detection method based on induction consciousness, which comprises the following steps:
stage 0: data initialization
Step 0: consider a Twitter conversation process C that includes a source tweet1(original rumor news) and a series of responsive commentary tweets in chronological order C ═ t2，t3，...，t|C|H, each of which pushes text tiExpressing the position of source tweet, including four positions of support, negation, inquiry and comment. Indicates the tweet tiHas a stand characteristic of di。
Step 1: learning each tweet t using Bi-GRUiAnd using the last step to conceal the layer vector
To represent the tweet.
Stage 1: bottom-up evolution tree network
Step 2: in order to highlight valuable questioning news comments and better express the semantics of the news in the comment thread, a bottom-up evolutionary tree recursive network (BETN) is established to gradually merge child nodes to parent nodes, and finally a vector is formed to represent the whole comment thread. First, the whole comment conversation thread is traversed from bottom to top in a reverse chronological order, so that a comment sequence containing many comment subchains is obtained.
And step 3: note that the creation of the ith comment position may be affected by the previous i-1 comment position in the comment subchain. The invention deeply considers the relevance between a single comment position and the comment subchain where the single comment position is located, and evaluates the relationship between the current node (the ith comment) and the previous node by means of social and psychological knowledge, namely the availability superposition theory and the conservative deviation theory.
And 4, step 4: the invention constructs a significance scoring function s (-) and a composite vector p (-) to calculate the merging degree between the ith node and the previous (i-1) nodes in a comment subchain. s (-) is responsible for measuring the degree to which the ith node in the comment subchain needs reinforcement, and p (-) represents the fusion degree of the two nodes.
And 5: based on these two social psychology theories, the significance score is mainly refined from the following aspects:
1) probability of two adjacent nodes merging. By utilizing a learnable parameter matrix S, the invention realizes the merging probability of two adjacent nodes by means of a bilinear function, thereby evaluating the enhancement degree of the node semantics.
τ(ai，ai-1)＝ai TSai-1 (1)
2) Semantic differences of nodes and their children chains.
For evaluating the semantic difference between the ith node and the previous (i-1) node. The larger the difference, the more prominent the semantic information of the ith node.
Wherein the content of the first and second substances,
representing the average semantics of the preceding (i-1) node.
3) Comment on the differences between the stands. φ (-) is used to investigate the difference between the review standpoint of the ith node and the previous (i-1) node. The larger the difference, the more obvious the i-th comment from the standpoint.
Wherein d isiA position vector representing the ith node,
represents the average vector of the previous (i-1) nodes.
4) And (4) timeliness of the nodes. The invention applies power law to slightly reinforce the latest comments to highlight the latest possible rumor splitting sound.
ψ(ti)＝gti -α-1 (4)
5) Wherein, tiIs the interval between the ith node and the first node in the child chain of comments. Both g and α are learnable parameters.
Step 6: finally, the saliency score may be described as:
and 7: the resultant vector p (-) can be computed as:
p(i，i-1)＝s(i，i-1)m(i，i-1) (7)
and 8: the merged results of all nodes in the whole sub-chain of comments are represented as:
and step 9: the invention utilizes TreeLSTM to realize the composite function m (·). TreeLSTM generates a hidden state vector h' and, given two input vectors aiAnd ajCell state vector c' of (c).
c′＝c′i⊙fi+c′j⊙fj+x⊙u (10)
h′＝o′+tanh(c′) (11)
Where U and b are trainable parameters, which indicate a dot product operation. In particular, h here denotes in the framework diagram of the invention
Step 10: traversing the whole comment sequence containing a plurality of comment subchains in the manner, after evolution and combination, the evolution tree network highlights valuable comments of questioning news in the comment threads, and finally the learnable representation is marked as BE (namely the valuable comments are marked as BE)
)。
And (2) stage: top-down coordination tree network
Step 11: considering that the hierarchy of the comment structure is not effectively considered in the bottom-up evolution tree network, the invention constructs a top-down coordination tree network (TCTN) to coordinate the information absorption between the father node and the brother node and strengthen the hierarchical relationship. Specifically, the entire annotation tree is first traversed from top to bottom in chronological order, assuming that π (t) represents the parent of t, k (t) represents its child, p (t) is the previous sibling at the same level in time, and s (t) representsIts successor at the same level. In the top-down coordination tree network of the present architecture diagram, with node t2For example, π (t) can be obtained2)＝t1，k(t2)＝t5，
And s (t)2)＝t3. If with the node t3For example, π (t) can be obtained2)＝t1，
p(t3)＝t2And s (t)3)＝t4。
Step 12: the present invention provides for a top-down coordinated tree network update process using subscripts in, fo, hf, c, and o for input gate, temporal forgetting gate, hierarchical forgetting gate, cell unit, and output, respectively. Vector int，fotAnd hftWeights representing newly captured information, old information from siblings, and old information from parents, respectively, are memorized.
Representing a hadamard product. If it is not
Then hp(t)And cp(t)Is set to an initial state value. In particular, htIn the architecture diagram corresponding to the invention
Step 13: finally, after traversing, we get a representation of the entire conversation thread as TC (i.e., h)|C|)。
Step 14: according to the invention, the comments of the questioning news acquired from the bottom-up evolution tree network and the hierarchical structure information of the comments acquired from the top-down coordination tree network are integrated, so that a valuable comment subchain of the questioning news is obtained.
F＝[BE；TC] (15)
Stage 3: task learning
Step 15: at the top of the TreeLSTM units of both nets, the bottom-up and top-down state vectors are concatenated and passed to the softmax layer to predict the probability distribution and cross entropy error minimization for the training samples with the true label y:
p＝softmax(WpF+bp) (16)
loss＝-∑ylogp (17)
the method is suitable for social network environments and can be used in social media network environments with rich hierarchical comment features.
The device provided by the embodiment of the invention. The embodiment comprises the following steps: a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor realizes the steps of the above-mentioned method embodiments when executing the computer program. Alternatively, the processor implements the functions of the modules/units in the above device embodiments when executing the computer program.
The computer program may be partitioned into one or more modules/units that are stored in the memory and executed by the processor to implement the invention.
The terminal device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The terminal device may include, but is not limited to, a processor, a memory.
The processor may be a Central Processing Unit (CPU), other general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc.
The memory may be used for storing the computer programs and/or modules, and the processor may implement various functions of the terminal device by executing or executing the computer programs and/or modules stored in the memory and calling data stored in the memory.
The terminal device integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by a computer program, which may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method embodiments may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer memory, Read-only memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, etc. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (5)
1. A false news interpretability detection method based on induction consciousness is characterized by comprising the following steps:
step 1, learning each tweet t by utilizing Bi-GRUiAnd using the last step to conceal the layer vector
Representing the tweet;
step 2, establishing a bottom-up evolution tree recursive network, gradually combining child nodes to father nodes, and finally forming a vector representing the whole comment thread; traversing the whole comment conversation thread from bottom to top in a reverse time sequence to obtain a comment sequence containing a plurality of comment subchains;
step 3, considering the relevance between a single comment position and the comment subchain where the single comment position is located, and evaluating the relation between the current node and the previous node by means of social psychology knowledge, namely, a utilization superposition theory and a conservative deviation theory;
step 4, constructing a significance scoring function s (-) and a composite vector p (-) to calculate the merging degree between the ith node and the previous (i-1) nodes in one comment subchain; the scoring function s (-) is used for measuring the degree of the ith node in the comment subchain needing to be strengthened, and the synthetic vector p (-) represents the fusion degree of the two nodes;
and 5, refining the significance score based on two social psychology theories:
1) probability τ of two adjacent nodes merging:
τ(ai，ai-1)＝ai TSai-1 (1)
2) semantic differences between nodes and their children chains
Wherein the content of the first and second substances,
representing the evaluation of semantic difference of the ith node and the previous (i-1) node;
mean semantics representing the preceding (i-1) node;
3) review the differences between the sites φ:
where φ (-) represents the difference between the review standpoint of the ith node and the previous (i-1) node; diA position vector representing the ith node,
an average vector representing the previous (i-1) nodes;
4) timeliness of the nodes ψ:
ψ(ti)＝gti -α-1 (4)
wherein, tiIs the interval between the ith node and the first node in the sub-chain of comments; g and alpha are both learnable parameters;
step 6, significance scoring:
step 7, calculating a synthetic vector p (·):
p(i，i-1)＝s(i，i-1)m(i，i-1) (7)
step 8, merging all nodes in the whole comment subchain as follows:
step 9, utilizing TreeLSTM to realize a composite function m (·); TreeLSTM generates a hidden state vector h' and, given two input vectors aiAnd ajThe following cell state vector c' yields:
c′＝c′i⊙fi+c′j⊙fj+x⊙u (10)
h′＝o′+tanh(c′) (11)
wherein U and b are trainable parameters, which indicate a dot product operation;
step 10, traversing the whole comment sequence containing a plurality of comment subchains, and after evolution and combination, highlighting valuable comments of questioning news in the comment threads through the evolution tree network, wherein the final learnable expression is marked as BE.
Step 11, constructing a top-down coordination tree network to coordinate information absorption between a father node and a brother node;
step 12, updating the coordination tree network:
subscripts in, fo, hf, c and o are respectively used for an input gate, a temporary forgetting gate, a hierarchical forgetting gate, a cell unit and an output; vector int，fotAnd hftRespectively representing the weight of the newly captured information, memorizing old information from brother nodes, and memorizing old information from father nodes;
representing a hadamard product; if it is not
Then hp(t)And cp(t)Is set to an initial state value;
step 13, after traversing, obtaining the representation of the whole conversation thread, and recording the representation as TC;
step 14, integrating the comment of the questioning news acquired by the evolvement tree network and the hierarchical structure information of the comment acquired by the coordination tree network to obtain a valuable comment subchain of the questioning news:
F＝[BE；TC] (15)。
2. the induction consciousness-based false news interpretability detection method of claim 1, wherein the specific method of the step 11 is as follows:
traversing the entire annotation tree from top to bottom in time order, assuming that π (t) represents the parent of t, k (t) represents its children, p (t) is the previous sibling at the same level in time, and s (t) represents its successor at the same level.
3. A induction awareness based false news interpretability detection system, comprising:
an input embedding module to learn a contextual semantic hidden layer representation of a conversation;
an evolutionary tree network for capturing valuable reviews that challenge news;
the coordination tree network is used for amplifying the comment subchains and explaining verification results;
and the task learning module integrates the position evolution semantics obtained by the evolution tree and the comment hierarchical structure semantics captured by the coordination tree network and inputs the position evolution semantics and the comment hierarchical structure semantics into Softmax to predict probability distribution.
4. A induction awareness based false news interpretability detection terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the method according to claim 1 or 2.
5. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to claim 1 or 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110008758.XA CN112711951A (en) | 2021-01-05 | 2021-01-05 | Induction consciousness-based false news interpretability detection system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110008758.XA CN112711951A (en) | 2021-01-05 | 2021-01-05 | Induction consciousness-based false news interpretability detection system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112711951A true CN112711951A (en) | 2021-04-27 |
Family
ID=75548282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110008758.XA Pending CN112711951A (en) | 2021-01-05 | 2021-01-05 | Induction consciousness-based false news interpretability detection system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112711951A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113177164A (en) * | 2021-05-13 | 2021-07-27 | 聂佼颖 | Multi-platform collaborative new media content monitoring and management system based on big data |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180349355A1 (en) * | 2017-05-31 | 2018-12-06 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Artificial Intelligence Based Method and Apparatus for Constructing Comment Graph |
| CN109005055A (en) * | 2018-07-16 | 2018-12-14 | 西安交通大学 | Complex network information node different degree evaluation method based on multiple dimensioned manifold |
| US20190079995A1 (en) * | 2016-09-09 | 2019-03-14 | Guangzhou Shenma Mobile Information Technology Co., Ltd. | Method, System, Server and User Terminal for Displaying User Comment Data |
| CN110210016A (en) * | 2019-04-25 | 2019-09-06 | 中国科学院计算技术研究所 | Bilinearity neural network Deceptive news detection method and system based on style guidance |
| CN111177554A (en) * | 2019-12-27 | 2020-05-19 | 西安交通大学 | False news identification system and method capable of explaining exploration based on generation of confrontation learning |
| CN111581980A (en) * | 2020-05-06 | 2020-08-25 | 西安交通大学 | False news detection system and method based on decision tree and common attention cooperation |
| CN111831790A (en) * | 2020-06-23 | 2020-10-27 | 广东工业大学 | False news identification method based on low threshold integration and text content matching |
-
2021
- 2021-01-05 CN CN202110008758.XA patent/CN112711951A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190079995A1 (en) * | 2016-09-09 | 2019-03-14 | Guangzhou Shenma Mobile Information Technology Co., Ltd. | Method, System, Server and User Terminal for Displaying User Comment Data |
| US20180349355A1 (en) * | 2017-05-31 | 2018-12-06 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Artificial Intelligence Based Method and Apparatus for Constructing Comment Graph |
| CN109005055A (en) * | 2018-07-16 | 2018-12-14 | 西安交通大学 | Complex network information node different degree evaluation method based on multiple dimensioned manifold |
| CN110210016A (en) * | 2019-04-25 | 2019-09-06 | 中国科学院计算技术研究所 | Bilinearity neural network Deceptive news detection method and system based on style guidance |
| CN111177554A (en) * | 2019-12-27 | 2020-05-19 | 西安交通大学 | False news identification system and method capable of explaining exploration based on generation of confrontation learning |
| CN111581980A (en) * | 2020-05-06 | 2020-08-25 | 西安交通大学 | False news detection system and method based on decision tree and common attention cooperation |
| CN111831790A (en) * | 2020-06-23 | 2020-10-27 | 广东工业大学 | False news identification method based on low threshold integration and text content matching |
Non-Patent Citations (1)
| Title |
|---|
| 王绍卿: "个性化新闻推荐技术研究综述", 《计算机科学与探索》, vol. 14, no. 1, 26 September 2019 (2019-09-26), pages 18 - 30 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113177164A (en) * | 2021-05-13 | 2021-07-27 | 聂佼颖 | Multi-platform collaborative new media content monitoring and management system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hagras | Toward human-understandable, explainable AI | |
| Suma | Computer vision for human-machine interaction-review | |
| Zhao et al. | An improved association rule mining algorithm for large data | |
| França et al. | An overview of deep learning in big data, image, and signal processing in the modern digital age | |
| Borodo et al. | Big data platforms and techniques | |
| Hsu et al. | Integrating machine learning and open data into social Chatbot for filtering information rumor | |
| Dang et al. | Increasing text filtering accuracy with improved LSTM | |
| CN112199961A (en) | Knowledge graph acquisition method based on deep learning | |
| KR20220047228A (en) | Method and apparatus for generating image classification model, electronic device, storage medium, computer program, roadside device and cloud control platform | |
| CN114330966A (en) | Risk prediction method, device, equipment and readable storage medium | |
| Gheisari et al. | Data mining techniques for web mining: a survey | |
| Bennetot et al. | Greybox XAI: A Neural-Symbolic learning framework to produce interpretable predictions for image classification | |
| CN111831826A (en) | Training method, classification method and device of cross-domain text classification model | |
| US20200019603A1 (en) | Systems, methods, and computer-readable media for improved table identification using a neural network | |
| Aftab et al. | Sentiment analysis of customer for ecommerce by applying AI | |
| CN112711951A (en) | Induction consciousness-based false news interpretability detection system and method | |
| CN115544212A (en) | Document-level event element extraction method, apparatus and medium | |
| CN112650851B (en) | False news identification system and method based on multilevel interactive evidence generation | |
| CN115168609A (en) | Text matching method and device, computer equipment and storage medium | |
| Devi et al. | Novel Trio-Neural Network towards Detecting Fake News on Social Media | |
| Umar et al. | A survey on state-of-the-art knowledge-based system development and issues | |
| CN114547310A (en) | False news early detection method, system, equipment and medium | |
| Szela̧g et al. | Rule‐based approach to multicriteria ranking | |
| Kaushik et al. | Sentiment analysis based on movie reviews using various classification techniques: A review | |
| Rautela et al. | Unveiling Sentiment Analysis: Exploring Techniques and Navigating Challenges |
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 |