Rumor detection method based on multivariate feature fusion and emotion difference analysis

Abstract

The invention discloses a rumor detection method based on multivariate feature fusion and emotion difference analysis, and particularly relates to the technical field of rumor detection, and the technical key points are as follows: respectively inputting network text data and a plurality of pieces of network comment data into a BERT model for feature extraction; text semantic features and text emotion features corresponding to the network text data are obtained, and comment semantic features and comment emotion features corresponding to each piece of network comment data are obtained; performing feature fusion on the text semantic feature and the plurality of comment semantic features to obtain a fused semantic feature, and calculating the difference between the text emotion feature and each comment emotion feature to obtain a difference vector feature; and performing feature fusion on the fusion semantic features and the difference vector features to generate fusion features, and identifying the fusion features by using a full-connection network to obtain a rumor identification result corresponding to the news event.

Description

Technical Field

[0001] The present invention relates to the technical field of rumor detection, and in particular to a rumor detection method based on multi-feature fusion and sentiment difference analysis. Background Art

[0002] Rumors are generally considered to be unverified information that is not supported by evidence, authoritative information or official statements. Unlike real events, rumors often use extreme or strong words to express emotions, or use exaggerated titles and pictures that are not strongly related to the content to attract people's attention, attempting to manipulate readers' emotions and positions, trigger strong emotional reactions, and thus be more easily accepted and spread.

[0003] In recent years, rumor detection technology has gradually received attention. However, existing methods still have some significant shortcomings. Many traditional rumor detection studies mainly rely on the extraction of semantic features, often ignoring the impact of emotional features on rumor propagation. Secondly, existing methods rely on manual annotation to obtain emotional labels, and usually adopt a single emotional classification system, which limits the in-depth understanding of the complex emotions in the comments and lacks effective mining of emotional differences, which may lead to bias in rumor judgment and affect the overall performance of the model. Some studies have attempted to use deep learning technology to improve the effect of feature extraction, but there are still deficiencies in feature fusion and sentiment analysis, and the effective combination of semantic features and emotional features has not been achieved, thus affecting the accuracy and efficiency of rumor detection.

[0004] Therefore, the present invention aims to provide a rumor detection method based on multi-feature fusion and sentiment difference analysis to solve the above-mentioned related problems. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the existing technology fails to realize the effective combination of semantic features and sentiment features, thereby affecting the accuracy and efficiency of rumor detection. The purpose is to provide a rumor detection method based on multi-feature fusion and sentiment difference analysis, by inputting the acquired network text data of the news event and the multiple network comment data corresponding to the network text data into the BERT model, respectively obtaining the sentiment features and semantic features of the network text data and the multiple network comment data; then the text semantic features are respectively fused with the multiple comment semantic features in a fine-grained manner to obtain a fused semantic feature, and the difference vector feature is obtained by calculating the difference between the text sentiment feature and the sentiment feature of each comment; finally, the fused semantic feature and the difference vector feature are fused to obtain the final fused feature, and a fully connected network is used as a discriminator to perform rumor identification on the fused feature to obtain a rumor identification result, thereby realizing the combination of semantic features and sentiment features to improve the rumor identification effect, and solving the problem that the existing technology fails to realize the effective combination of semantic features and sentiment features, thereby affecting the accuracy and efficiency of rumor detection.

[0006] The present invention is achieved through the following technical solutions:

[0007] A rumor detection method based on multi-feature fusion and sentiment difference analysis, the method comprising:

[0008] Acquire network text data of a news event and a plurality of network comment data corresponding to the network text data;

[0009] Input the network text data and multiple network comment data into the BERT model for feature extraction, obtain the text semantic features and text sentiment features corresponding to the network text data, and obtain the comment semantic features and comment sentiment features corresponding to each network comment data;

[0010] The main text semantic features are fused with the semantic features of multiple comments to obtain fused semantic features, and the difference between the main text sentiment features and the sentiment features of each comment is calculated to obtain the difference vector features;

[0011] The fused semantic features and the difference vector features are fused to generate fused features, and the fused features are identified using a pre-trained fully connected network to obtain the rumor identification results corresponding to the news events.

[0012] Furthermore, the method further comprises:

[0013] The cross entropy loss function is used to calculate the error value between the rumor identification result and the true result. The parameters of the fully connected network are iteratively optimized according to the error value until the loss function converges to obtain the fully connected network after optimized training.

[0014] Furthermore, the network text data and multiple network comment data are respectively input into the BERT model for feature extraction, and the text semantic features and text sentiment features corresponding to the network text data are obtained, and the comment semantic features and comment sentiment features corresponding to each network comment data are respectively obtained, specifically:

[0015] Input the network text data and multiple network comment data into the BERT model respectively to obtain the text feature representation corresponding to the network text data and the comment feature representation corresponding to each network comment data;

[0016] The semantic features in the feature representation of the main text and the feature representation of multiple comments are extracted by a multi-layer perceptron to obtain the semantic features of the main text and the semantic features of multiple comments respectively;

[0017] Then, the sentiment features in the text feature representation and the multiple comment feature representations are extracted through a multi-layer perceptron to obtain the text sentiment features and the multiple comment sentiment features respectively.

[0018] Furthermore, the main text semantic features are fused with the semantic features of multiple comments to obtain fused semantic features, and the difference between the main text sentiment features and the sentiment features of each comment is calculated to obtain the difference vector features, which are as follows:

[0019] A comment semantic feature matrix is ​​constructed using multiple comment semantic features. The comment semantic feature matrix is ​​transposed and then matrix multiplied with the text semantic features to obtain a semantic score. The semantic score is then matrix multiplied with the comment semantic feature matrix to obtain a fused semantic feature.

[0020] By calculating the Euclidean distances between the sentiment features of the text and multiple comment sentiment features, the difference vector features are constructed using multiple Euclidean distances.

[0021] Furthermore, the fusion semantic features and the difference vector features are fused to generate fusion features, and the fusion features are identified using a fully connected network to obtain the rumor identification results corresponding to the network text data, specifically:

[0022] Perform feature fusion and splicing on the fused semantic features and the difference vector features to obtain fused features;

[0023] The fused features and the corresponding true and false labels are input into the fully connected network, and rumor identification is performed through the softmax function in the fully connected network to obtain the rumor identification results corresponding to the news events.

[0024] The present invention also provides a rumor detection system based on multi-feature fusion and sentiment difference analysis, which is used in any one of the rumor detection methods based on multi-feature fusion and sentiment difference analysis described above, and the system includes:

[0025] A data acquisition module, used to acquire network text data of a news event and a plurality of network comment data corresponding to the network text data;

[0026] A feature extraction module is used to input the network text data and multiple network comment data into the BERT model for feature extraction, thereby obtaining the text semantic features and text sentiment features corresponding to the network text data, and obtaining the comment semantic features and comment sentiment features corresponding to each network comment data;

[0027] The feature processing module is used to fuse the text semantic features with the semantic features of multiple comments to obtain fused semantic features, and to calculate the difference between the text sentiment feature and the sentiment feature of each comment to obtain a difference vector feature;

[0028] The rumor identification module is used to fuse the fused semantic features and the differential vector features to generate fused features, and use the pre-trained fully connected network to identify the fused features to obtain the rumor identification results corresponding to the news events.

[0029] Furthermore, the system also includes:

[0030] The network optimization module is used to calculate the error value between the rumor identification result and the true result using the cross entropy loss function, and iteratively optimize the parameters of the fully connected network according to the error value until the loss function converges to obtain the fully connected network after optimization training.

[0031] The present invention also provides a computer device, comprising a system memory and a processor, wherein the system memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0032] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of any one of the methods described above when executed by a processor.

[0033] The present invention also provides a computer program product comprising instructions, wherein when the instructions are executed by a computer device cluster, the computer device cluster executes any of the above methods.

[0034] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0035] In the present invention, the network text data of the acquired news event and multiple network comment data corresponding to the network text data are input into the BERT model to obtain the sentiment features and semantic features of the network text data and the multiple network comment data respectively; the text semantic features are then fused with the multiple comment semantic features in a fine-grained manner to obtain fused semantic features, and the difference vector features are obtained by calculating the difference between the text sentiment features and the sentiment features of each comment; finally, the fused semantic features and the difference vector features are fused to obtain the final fused features, and a fully connected network is used as a discriminator to perform rumor identification on the fused features to obtain a rumor identification result, thereby realizing the combination of semantic features and sentiment features to improve the rumor identification effect, solving the problem that the existing technology fails to achieve an effective combination of semantic features and sentiment features, thereby affecting the accuracy and efficiency of rumor detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0037] Figure 1 : is a method flow chart of a rumor detection method based on multi-feature fusion and sentiment difference analysis in this embodiment;

[0038] Figure 2 Schematic diagram of module connection of a rumor detection system based on multi-feature fusion and sentiment difference analysis in this embodiment;

[0039] Figure 3 A schematic diagram of the structure of a computer device. DETAILED DESCRIPTION

[0040] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0041] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0042] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0043] Example 1

[0044] See also Figure 1 As shown, this embodiment provides a rumor detection method based on multi-feature fusion and sentiment difference analysis, the method comprising:

[0045] S1: Obtain the network text data of the news event and multiple network comment data corresponding to the network text data, specifically:

[0046] After obtaining the network text data of the news event and multiple network comment data corresponding to the network text data, the network text data and the multiple network comment data are first processed to clean up the redundant information not involved in the news event, and the number of network comment data is uniformly set to N; then the emotional labels of the network text data and the multiple network comment data are obtained, and the network text data and the multiple network comment data are formatted.

[0047] S2: Input the network text data and multiple network comment data into the BERT model for feature extraction, obtain the text semantic features and text sentiment features corresponding to the network text data, and obtain the comment semantic features and comment sentiment features corresponding to each network comment data, specifically:

[0048] The network text data and multiple network comment data are input into the BERT model respectively to obtain the text feature representation T corresponding to the network text data. t , and the comment feature representation T corresponding to each network comment data cN , specifically:

[0049]

[0050] Among them, X t For network text data; X c ={Xc1 ,X c2 ,…,X cN} is the online comment data; T cN is the comment feature representation corresponding to each online comment data, N is the number of comments, N is a positive integer; T c A review feature representation matrix constructed for multiple review feature representations;

[0051] The text feature representation T is represented by a multi-layer perceptron t and multiple review feature representations T cN The semantic features in are extracted to obtain the text semantic features F t and multiple review semantic features F cN , specifically:

[0052]

[0053] Among them, W t is the weight matrix of the text semantic features; W c is the weight matrix of the semantic features of the comment group; b t , b c are all bias vectors; d fc is the feature dimension of the fully connected layer output; F c A review semantic feature matrix constructed for multiple review semantic features;

[0054] Then, the text feature representation T is represented by a multi-layer perceptron t and multiple review feature representations T cN The emotional features in the text are extracted to obtain the emotional features E t and multiple comment sentiment features E cN , specifically:

[0055]

[0056] Among them, W et is the weight matrix of the text sentiment features; W ec is the weight matrix of comment sentiment features; b et , b ec are all bias vectors; d fc is the feature dimension of the fully connected layer output; E c A review sentiment feature matrix constructed for multiple review sentiment features.

[0057] S3: The main text semantic features are fused with the semantic features of multiple comments to obtain fused semantic features, and the difference between the main text sentiment features and the sentiment features of each comment is calculated to obtain the difference vector features, which are as follows:

[0058] Using multiple review semantic features FcN Construct the comment semantic feature matrix F c , for the comment semantic feature matrix F c After transposition, they are respectively combined with the text semantic features F t Perform matrix multiplication to obtain the semantic score R, and then add the semantic score R to the comment semantic feature matrix F c Perform matrix multiplication to obtain the fused semantic feature F fusion , specifically:

[0059]

[0060] F fusion =R·F c

[0061] in, is the transposed comment semantic feature matrix;

[0062] By calculating the text sentiment feature E t Respectively with multiple comment sentiment features E cN The Euclidean distance D between N , using multiple Euclidean distances D N The difference vector feature D is constructed, specifically:

[0063]

[0064] S4: The fused semantic features and the difference vector features are fused to generate fused features, and the fused features are identified using the pre-trained fully connected network to obtain the rumor identification results corresponding to the news events, specifically:

[0065] The fusion semantic feature F fusion The feature fusion and splicing are performed with the difference vector feature D to obtain the fusion feature F final =[F fusion ; D];

[0066] The fusion feature F final and the corresponding true or false label y are input into the fully connected network, and the rumor identification is performed through the softmax function in the fully connected network. final Map to the true and false classification target space to get the classification result That is, the rumor identification results corresponding to the news event are obtained, specifically:

[0067]

[0068] Among them, W represents the weight matrix in the classifier; b represents the bias term.

[0069] In another embodiment, the method further includes: calculating the error value between the rumor identification result and the true result using a cross entropy loss function, and iteratively optimizing the parameters of the fully connected network according to the error value until the loss function converges to obtain the fully connected network after optimization training, specifically:

[0070] Using cross entropy loss function Calculate the error between the rumor identification result and the true result, and optimize the parameters of the rumor detection task. The true result of the event is y i ∈{0,1} (0 indicates real news, 1 indicates rumor), the model identification result is The calculation results are as follows:

[0071]

[0072] Among them, y i Represents the true result of the i-th event; It is represented as the identification result of the i-th event;

[0073] At the same time, the true sentiment label of each comment is y c ∈{0,1,2} (representing negative, neutral and positive emotions respectively), the emotion probability predicted by the model is Use the cross entropy loss function to calculate the error of the sentiment classification of the comments:

[0074]

[0075] Among them, k represents the comment index; j represents the sentiment category index; Represented as the true sentiment label of the kth comment; It is represented as the true sentiment label of the kth comment; C is the number of sentiment categories;

[0076] The final overall loss function is defined as the weighted addition of the classification loss for rumor detection and the sentiment classification loss.

[0077]

[0078] Among them, λ is the weight coefficient.

[0079] Specifically, in this embodiment, the acquired network text data of the news event and multiple network comment data corresponding to the network text data are input into the BERT model to obtain the sentiment features and semantic features of the network text data and the multiple network comment data respectively; the text semantic features are then fused with the multiple comment semantic features in a fine-grained manner to obtain fused semantic features, and the difference vector features are obtained by calculating the difference between the text sentiment features and the sentiment features of each comment; finally, the fused semantic features and the difference vector features are fused to obtain the final fused features, and a fully connected network is used as a discriminator to perform rumor identification on the fused features to obtain a rumor identification result, thereby realizing the combination of semantic features and sentiment features to improve the rumor identification effect, solving the problem that the existing technology fails to achieve an effective combination of semantic features and sentiment features, thereby affecting the accuracy and efficiency of rumor detection.

[0080] Example 2

[0081] See also Figure 2 As shown, this embodiment also provides a rumor detection system based on multi-feature fusion and sentiment difference analysis, which is used in any one of the rumor detection methods based on multi-feature fusion and sentiment difference analysis described above, and the system includes:

[0082] The data acquisition module 100 is used to acquire network text data of a news event and a plurality of network comment data corresponding to the network text data;

[0083] The feature extraction module 200 is used to input the network text data and the plurality of network comment data into the BERT model to extract semantic features, obtain the text semantic features and text sentiment features corresponding to the network text data, and obtain the comment semantic features and comment sentiment features corresponding to each network comment data;

[0084] The feature processing module 300 is used to fuse the text semantic features with the multiple comment semantic features to obtain fused semantic features, and to calculate the difference between the text sentiment feature and each comment sentiment feature to obtain a difference vector feature;

[0085] The rumor identification module 400 is used to generate a fused feature by fusing the fused semantic feature and the differential vector feature, and to identify the fused feature using a pre-trained fully connected network to obtain a rumor identification result corresponding to the news event.

[0086] Furthermore, the system also includes: using a cross entropy loss function to calculate the error value between the rumor identification result and the true result, and iteratively optimizing the parameters of the fully connected network according to the error value until the loss function converges to obtain the fully connected network after optimized training.

[0087] It should be noted that the modules in the system of Example 2 correspond to the steps in the method of Example 1. The steps in the method of Example 1 have been described in detail in Example 1. In this Example 2, the contents of the modules in the system will not be described in detail.

[0088] Example 3

[0089] See also Figure 3 As shown, this embodiment further provides a computer device, including a system memory 1005 and a processor 1001, wherein the system memory 1005 stores a computer program, and the processor 1001 implements the steps of any of the above methods when executing the computer program.

[0090] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, the processor 1001 implements the functions of each module / unit in the above system / device embodiments when executing the computer program.

[0091] Specifically, in this embodiment, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the system memory 1005, and are executed by the processor 1001 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0092] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will appreciate that this does not constitute a limitation on the terminal device, and may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0093] The processor 1001 may be a central processing unit (CPU), or other general-purpose processors 1001, digital signal processors 1001 (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor 1001 may be a microprocessor or any conventional processor, etc.

[0094] The system memory 1005 may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. The system memory 1005 may also be a storage device 1004 of the terminal device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device. Further, the system memory 1005 may also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 may also be used to temporarily store data that has been output or is to be output.

[0095] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0096] Example 4

[0097] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0098] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0099] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In an embodiment of the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device.

[0100] Example 5

[0101] This embodiment also provides a computer program product including instructions. When the instructions are executed by a computer device cluster, the computer device cluster executes the method described in Embodiment 1.

[0102] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A rumor detection method based on multi-feature fusion and sentiment difference analysis, characterized in that: Methods include: Acquire network text data of a news event and a plurality of network comment data corresponding to the network text data; Input the network text data and multiple network comment data into the BERT model for feature extraction, obtain the text semantic features and text sentiment features corresponding to the network text data, and obtain the comment semantic features and comment sentiment features corresponding to each network comment data; The main text semantic features are fused with the semantic features of multiple comments to obtain fused semantic features, and the difference between the main text sentiment features and the sentiment features of each comment is calculated to obtain the difference vector features; The fused semantic features and the difference vector features are fused to generate fused features, and the fused features are identified using a pre-trained fully connected network to obtain the rumor identification results corresponding to the news events.

2. According to claim 1, a rumor detection method based on multi-feature fusion and sentiment difference analysis is characterized in that: The method also includes: The cross entropy loss function is used to calculate the error value between the rumor identification result and the true result. The parameters of the fully connected network are iteratively optimized according to the error value until the loss function converges to obtain the trained fully connected network.

3. The rumor detection method based on multi-feature fusion and sentiment difference analysis according to claim 1 is characterized in that: The network text data and multiple network comment data are respectively input into the BERT model for feature extraction, and the text semantic features and text sentiment features corresponding to the network text data are obtained, and the comment semantic features and comment sentiment features corresponding to each network comment data are obtained, specifically: Input the network text data and multiple network comment data into the BERT model respectively to obtain the text feature representation corresponding to the network text data and the comment feature representation corresponding to each network comment data; The semantic features in the feature representation of the main text and the feature representation of multiple comments are extracted by a multi-layer perceptron to obtain the semantic features of the main text and the semantic features of multiple comments respectively; Then, the sentiment features in the text feature representation and the multiple comment feature representations are extracted through a multi-layer perceptron to obtain the text sentiment features and the multiple comment sentiment features respectively.

4. The rumor detection method based on multi-feature fusion and sentiment difference analysis according to claim 3 is characterized in that: The main text semantic features are fused with the semantic features of multiple comments to obtain the fused semantic features, and the difference between the main text sentiment features and the sentiment features of each comment is calculated to obtain the difference vector features, which are as follows: A comment semantic feature matrix is ​​constructed using multiple comment semantic features. The comment semantic feature matrix is ​​transposed and then matrix multiplied with the text semantic features to obtain a semantic score. The semantic score is then matrix multiplied with the comment semantic feature matrix to obtain a fused semantic feature. By calculating the Euclidean distances between the sentiment features of the text and multiple comment sentiment features, the difference vector features are constructed using multiple Euclidean distances.

5. The rumor detection method based on multi-feature fusion and sentiment difference analysis according to claim 3 is characterized in that: The fusion semantic features and the difference vector features are fused to generate fusion features, and the fusion features are identified using a fully connected network to obtain the rumor identification results corresponding to the network text data, specifically: Perform feature fusion and splicing on the fused semantic features and the difference vector features to obtain fused features; The fused features and the corresponding true and false labels are input into the fully connected network, and rumor identification is performed through the softmax function in the fully connected network to obtain the rumor identification results corresponding to the news events.

6. A rumor detection system based on multi-feature fusion and sentiment difference analysis, characterized in that: The system is used in a rumor detection method based on multi-feature fusion and sentiment difference analysis as described in any one of claims 1 to 5, and the system comprises: A data acquisition module, used to acquire network text data of a news event and a plurality of network comment data corresponding to the network text data; A feature extraction module is used to input the network text data and multiple network comment data into the BERT model for feature extraction, thereby obtaining the text semantic features and text sentiment features corresponding to the network text data, and obtaining the comment semantic features and comment sentiment features corresponding to each network comment data; The feature processing module is used to fuse the text semantic features with the semantic features of multiple comments to obtain fused semantic features, and to calculate the difference between the text sentiment feature and the sentiment feature of each comment to obtain a difference vector feature; The rumor identification module is used to fuse the fused semantic features and the differential vector features to generate fused features, and use the pre-trained fully connected network to identify the fused features to obtain the rumor identification results corresponding to the news events.

7. The rumor detection system based on multi-feature fusion and sentiment difference analysis according to claim 6 is characterized in that: The system also includes: The network optimization module is used to calculate the error value between the rumor identification result and the true result using the cross entropy loss function, and iteratively optimize the parameters of the fully connected network according to the error value until the loss function converges to obtain the fully connected network after optimization training.

8. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product comprising instructions, characterized in that When the instructions are executed by a computer device cluster, the computer device cluster executes the method according to any one of claims 1 to 5.

Images