Method and system for false media identification based on single layer separation

Abstract

The present invention provides a method and system for false media identification based on single-layer separation, which converts the dimension of the acquired network multimedia data stream, converges the data stream using a bidirectional gated recurrent unit model with an attention mechanism and a hierarchical attention network model, performs single-layer separation according to tensor calculation results, mines out positive and negative attitudes, obtains word components through semantic analysis, performs secondary sampling according to changes in the values ​​of word component breakpoints, and separates features in which dimensions jump, thereby more quickly detecting whether multimedia data is compliant and achieving the purpose of identifying false information.

Description

Technical Field

[0001] The present application relates to the field of network multimedia, and in particular to a method and system for false media identification based on single-layer separation. Background Art

[0002] There are a large number of words in the existing online media. In order to keep up with the development of online words, the system needs to be constantly trained. In order to facilitate the system to call the corresponding word vector, each word must be matched with the trained word vector one by one. This process consumes a lot of resources. Even the same comment in the comment text may express the opposite meaning.

[0003] Therefore, there is an urgent need for a targeted method and system for false media identification based on single-layer separation. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for false media identification based on single-layer separation, by converting the dimension of the acquired network multimedia data stream, and using a bidirectional gated recurrent unit model with an attention mechanism and a hierarchical attention network model to converge the data stream, performing single-layer separation according to the tensor calculation results, mining the positive and negative attitudes, and then obtaining word components through semantic analysis, and performing secondary sampling according to the change in the value of the word component breakpoint, which can separate the features in which the dimension jumps, thereby faster detecting whether the multimedia data is compliant and achieving the purpose of identifying false information.

[0005] In a first aspect, the present application provides a method for false media identification based on single-layer separation, the method comprising:

[0006] The server obtains a network multimedia data stream, samples the network multimedia data stream, vectorizes the sampling result, and performs dimensionality conversion to convert the received P*Q dimensional multimedia signal into an M*N dimensional multimedia signal, where P*Q is the dimension of the signal transmission channel, M*N is the dimension of the server load processing, and P, Q, M, and N are all non-zero positive integers;

[0007] The multimedia signal after the dimension conversion is reorganized into a first data stream according to the user identifier, and input into a bidirectional gated recurrent unit model and a hierarchical attention network model with an attention mechanism, wherein the bidirectional gated recurrent unit model and the hierarchical attention network model converge the first data stream respectively, complete their respective tensor calculations after convergence, extract the output vectors of the attention layer for splicing, and obtain a first intermediate result;

[0008] Using the hierarchical attention network model again to identify the first intermediate result, mining the positive and negative attitudes in the first data stream, semantically analyzing the word components corresponding to the positive and negative attitudes, and obtaining the corresponding word meanings;

[0009] Detecting the value of each breakpoint in the word component, wherein the value of the breakpoint is calculated by weighted average of the features of each breakpoint and the features of surrounding adjacent breakpoints, and setting an anchor point where the value of the breakpoint jumps;

[0010] The first data stream is sampled again at the anchor point, a second feature is obtained, the second feature is encoded, and then input into an N-layer convolution unit, and a second intermediate result is obtained according to an output result of the N-layer convolution unit;

[0011] The second intermediate result is smoothed to obtain a high-dimensional signal set carrying boundary and regional local features, the high-dimensional signal set is analyzed to separate the features in which the dimensions jump, the multimedia data corresponding to the features in which the dimensions jump are queried, and whether the multimedia data is compliant is detected. If it is not compliant with the rules, it is determined that the network multimedia data stream contains false information and an alarm is generated.

[0012] In combination with the first aspect, in a first possible implementation of the first aspect, the N-layer convolution unit is composed of N convolution operation modules connected in sequence, and the value of N reflects the load processing capability of the server.

[0013] In combination with the first aspect, in a second possible implementation manner of the first aspect, the user identifier is carried by a network multimedia data stream.

[0014] In combination with the first aspect, in a third possible implementation of the first aspect, a neural network model is used in the process of separating features in which dimension jumps occur.

[0015] In a second aspect, the present application provides a system for false media identification based on single-layer separation, the system comprising a processor and a memory:

[0016] The memory is used to store program code and transmit the program code to the processor;

[0017] The processor is used to execute any one of the four possible methods of the first aspect according to the instructions in the program code.

[0018] In a third aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute any one of the four possible methods in the first aspect.

[0019] The present invention provides a method and system for false media identification based on single-layer separation, which converts the dimension of the acquired network multimedia data stream, converges the data stream using a bidirectional gated recurrent unit model with an attention mechanism and a hierarchical attention network model, performs single-layer separation according to tensor calculation results, mines out positive and negative attitudes, obtains word components through semantic analysis, performs secondary sampling according to changes in the values ​​of word component breakpoints, and separates features in which dimensions jump, thereby more quickly detecting whether multimedia data is compliant and achieving the purpose of identifying false information. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0022] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0023] Figure 1 The flowchart of the method for false media identification based on single-layer separation provided in this application includes:

[0024] The server obtains a network multimedia data stream, samples the network multimedia data stream, vectorizes the sampling result, and performs dimensionality conversion to convert the received P*Q dimensional multimedia signal into an M*N dimensional multimedia signal, where P*Q is the dimension of the signal transmission channel, M*N is the dimension of the server load processing, and P, Q, M, and N are all non-zero positive integers;

[0025] The multimedia signal after the dimension conversion is reorganized into a first data stream according to the user identifier, and input into a bidirectional gated recurrent unit model and a hierarchical attention network model with an attention mechanism, wherein the bidirectional gated recurrent unit model and the hierarchical attention network model converge the first data stream respectively, complete their respective tensor calculations after convergence, extract the output vectors of the attention layer for splicing, and obtain a first intermediate result;

[0026] Using the hierarchical attention network model again to identify the first intermediate result, mining the positive and negative attitudes in the first data stream, semantically analyzing the word components corresponding to the positive and negative attitudes, and obtaining the corresponding word meanings;

[0027] Detecting the value of each breakpoint in the word component, wherein the value of the breakpoint is calculated by weighted average of the features of each breakpoint and the features of surrounding adjacent breakpoints, and setting an anchor point where the value of the breakpoint jumps;

[0028] The first data stream is sampled again at the anchor point, a second feature is obtained, the second feature is encoded, and then input into an N-layer convolution unit, and a second intermediate result is obtained according to an output result of the N-layer convolution unit;

[0029] The second intermediate result is smoothed to obtain a high-dimensional signal set carrying boundary and regional local features, the high-dimensional signal set is analyzed to separate the features in which the dimensions jump, the multimedia data corresponding to the features in which the dimensions jump are queried, and whether the multimedia data is compliant is detected. If it is not compliant with the rules, it is determined that the network multimedia data stream contains false information and an alarm is generated.

[0030] In some preferred embodiments, the N-layer convolution unit is composed of N convolution operation modules connected in sequence, and the value of N reflects the load processing capability of the server.

[0031] In some preferred embodiments, the user identification is carried by a network multimedia data stream.

[0032] In some preferred embodiments, a neural network model is used in the process of separating features in which dimension jumps occur.

[0033] The present application provides a system for false media identification based on single-layer separation, the system comprising: the system comprising a processor and a memory:

[0034] The memory is used to store program code and transmit the program code to the processor;

[0035] The processor is used to execute the method described in any one of all embodiments of the first aspect according to the instructions in the program code.

[0036] The present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method described in any one of the embodiments of the first aspect.

[0037] In a specific implementation, the present invention further provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment of the present invention. The storage medium may be a disk, an optical disk, a read-only storage memory (abbreviated as: ROM) or a random access memory (abbreviated as: RAM), etc.

[0038] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention or some parts of the embodiments.

[0039] The same and similar parts between the various embodiments of this specification can be referred to each other. In particular, for the embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0040] The above-described embodiments of the present invention do not limit the protection scope of the present invention.

Claims

1. A method for false media identification based on single-layer separation, It is characterized in that The method comprises: The server obtains a network multimedia data stream, samples the network multimedia data stream, vectorizes the sampling result, and performs dimensionality conversion to convert the received P*Q dimensional multimedia signal into an M*N dimensional multimedia signal, where P*Q is the dimension of the signal transmission channel, M*N is the dimension of the server load processing, and P, Q, M, and N are all non-zero positive integers; The multimedia signal after the dimension conversion is reorganized into a first data stream according to the user identifier, and input into a bidirectional gated recurrent unit model and a hierarchical attention network model with an attention mechanism, wherein the bidirectional gated recurrent unit model and the hierarchical attention network model converge the first data stream respectively, complete their respective tensor calculations after convergence, extract the output vectors of the attention layer for splicing, and obtain a first intermediate result; Using the hierarchical attention network model again to identify the first intermediate result, mining the positive and negative attitudes in the first data stream, semantically analyzing the word components corresponding to the positive and negative attitudes, and obtaining the corresponding word meanings; Detecting the value of each breakpoint in the word component, wherein the value of the breakpoint is calculated by weighted average of the features of each breakpoint and the features of surrounding adjacent breakpoints, and setting an anchor point where the value of the breakpoint jumps; The first data stream is sampled again at the anchor point, a second feature is obtained, the second feature is encoded, and then input into an N-layer convolution unit, and a second intermediate result is obtained according to an output result of the N-layer convolution unit; The second intermediate result is smoothed to obtain a high-dimensional signal set carrying boundary and regional local features, the high-dimensional signal set is analyzed to separate the features in which the dimensions jump, the multimedia data corresponding to the features in which the dimensions jump are queried, and whether the multimedia data is compliant is detected. If it is not compliant with the rules, it is determined that the network multimedia data stream contains false information and an alarm is generated.

2. The method according to claim 1, Features: The N-layer convolution unit is composed of N convolution operation modules connected in sequence, and the value of N reflects the load processing capability of the server.

3. The method according to any one of claims 1 to 2, Features: The user identification is carried by the network multimedia data stream.

4. The method according to any one of claims 1 to 3, Features: A neural network model is used in the process of separating features in which dimension jumps occur.

5. A system for false media identification based on single-layer separation, It is characterized in that The system comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the method according to any one of claims 1 to 4 according to the instructions in the program code.

6. A computer-readable storage medium, It is characterized in that The computer-readable storage medium is used to store program codes, and the program codes are used to execute the method according to any one of claims 1 to 4.

Images