Fusion detection method and system based on content semantics

Abstract

The application provides a content semantic-based fusion detection method and system, which performs syntactic analysis and semantic analysis through a syntactic model and a semantic analysis model to perform fusion detection, the syntactic model completes positioning of an object according to a subject and a predicate to obtain a triple topic vector, and then outputs word meanings through the semantic analysis model to reorganize a new sentence, and whether the sentence contains non-compliant content is detected, thereby overcoming the problem that the prior art cannot detect incomplete sentence structures.

Description

Technical Field

[0001] This application relates to the field of cybersecurity technology, and in particular to a content semantics-based fusion detection method and system. Background Technology

[0002] Existing topic identification and detection methods tend to miss trending topics, especially in online news, where concise language, incomplete sentence structures, or highly time-sensitive scenarios are common. How to quickly and efficiently identify trending news topics and verify the compliance of their content is a crucial issue in the field of cybersecurity.

[0003] Therefore, there is an urgent need for a targeted content semantic-based fusion detection method and system. Summary of the Invention

[0004] The purpose of this invention is to provide a content-semantic fusion detection method and system, which performs syntactic and semantic analysis through syntactic and semantic analysis to perform fusion detection, and reconstructs triplet topic vectors into new sentences, thereby improving sentence detection.

[0005] Firstly, this application provides a content semantic-based fusion detection method, the method comprising:

[0006] Build a cloud computing platform on the server to construct syntactic and semantic analysis models;

[0007] According to the acquisition strategy, the data stream of the Internet platform is acquired, the text features of the data stream are extracted, vectorized, and the text clustering operation is completed to obtain hotspot clusters generated by category;

[0008] For each hotspot cluster, input syntactic model, perform fusion syntactic analysis, extract subject and predicate from several sentences of the hotspot cluster, and discover the dependency relationship between subject and action, as well as the correlation relationship between subject and object based on the predicate, analyze the relational structure of the several sentences, the relational structure is a parallel structure or a verb-object structure, and complete the location of object;

[0009] The corresponding role label is derived from the subject. Multiple subjects are then linked together in a network structure based on the role label and vectorized to obtain the subject vector. The subject vector, predicate, and located object are then input into a preset triplet template to obtain the triplet topic vector.

[0010] The triplet topic vector is input into the semantic analysis model, the word meaning is output, and the word meaning is reassembled into a new sentence.

[0011] Determine whether the meaning of the word includes the specified keyword. If so, continue to determine whether the new sentence after the word meaning is recombined forms the specified meaning. If the new sentence forms the specified meaning, the triplet topic vector is considered to be in a situation that requires an alarm, and an alarm message is issued. If the new sentence does not form the specified meaning, the triplet topic vector is considered to be compliant.

[0012] Secondly, this application provides a content semantic-based fusion detection system, the system comprising:

[0013] The building unit is used to build a cloud computing platform on a server and construct syntactic and semantic analysis models;

[0014] The extraction unit is used to acquire the data stream of the Internet platform according to the acquisition strategy, extract the text features of the data stream, vectorize them, and complete the text clustering operation to obtain hotspot clusters generated by category;

[0015] The syntactic analysis unit is used to input each hotspot cluster into the syntactic model, perform fusion syntactic analysis, extract the subject and predicate from several sentences of the hotspot cluster, and discover the dependency relationship between the subject and the action, as well as the correlation relationship between the subject and the object based on the predicate, and analyze the relational structure of the several sentences. The relational structure is a parallel structure or a verb-object structure, and completes the location of the object.

[0016] The template unit is used to derive the corresponding role label based on the subject, and to associate multiple subjects into a network structure based on the role label to complete vectorization and obtain the subject vector; the subject vector, predicate, and located object are input into a preset triplet template to obtain the triplet topic vector;

[0017] The semantic analysis unit is used to input the triplet topic vector into the semantic analysis model, output word meanings, and reorganize the word meanings into new sentences.

[0018] The judgment unit is used to determine whether the meaning of the word includes the specified keyword. If so, it continues to determine whether the new sentence after the word meaning is recombined forms the specified meaning. If the new sentence forms the specified meaning, the triplet topic vector is determined to be in a situation that requires an alarm, and an alarm message is issued. If the new sentence does not form the specified meaning, the triplet topic vector is determined to be compliant.

[0019] Thirdly, this application provides a content semantic-based fusion detection system, the system comprising a processor and a memory:

[0020] The memory is used to store program code and transmit the program code to the processor;

[0021] The processor is configured to execute any one of the four possible methods of the first aspect according to the instructions in the program code.

[0022] Fourthly, this application provides a computer-readable storage medium for storing program code for performing any one of the four possible methods of the first aspect.

[0023] Beneficial effects

[0024] This invention provides a content-semantic fusion detection method and system. It performs fusion detection by using a syntactic model and a semantic analysis model for syntactic and semantic analysis. The syntactic model locates the object based on the subject and predicate to obtain a triplet topic vector. Then, the semantic analysis model outputs word meanings, reassembles them into a new sentence, and detects whether the sentence contains non-compliant content. This overcomes the problem that existing technologies cannot detect incomplete sentence structures. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0026] Figure 1 This is a general flowchart of the content semantic-based fusion detection method of the present invention;

[0027] Figure 2 This is an architecture diagram of the content semantics-based fusion detection system of the present invention. Detailed Implementation

[0028] The preferred embodiments of the present invention will now be described in detail with reference to 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 providing a clearer and more explicit definition of the scope of protection of the present invention.

[0029] Figure 1 A general flowchart of the content semantic-based fusion detection method provided in this application is shown, the method comprising:

[0030] Build a cloud computing platform on the server to construct syntactic and semantic analysis models;

[0031] According to the acquisition strategy, the data stream of the Internet platform is acquired, the text features of the data stream are extracted, vectorized, and the text clustering operation is completed to obtain hotspot clusters generated by category;

[0032] For each hotspot cluster, input syntactic model, perform fusion syntactic analysis, extract subject and predicate from several sentences of the hotspot cluster, and discover the dependency relationship between subject and action, as well as the correlation relationship between subject and object based on the predicate, analyze the relational structure of the several sentences, the relational structure is a parallel structure or a verb-object structure, and complete the location of object;

[0033] The corresponding role label is derived from the subject. Multiple subjects are then linked together in a network structure based on the role label and vectorized to obtain the subject vector. The subject vector, predicate, and located object are then input into a preset triplet template to obtain the triplet topic vector.

[0034] The triplet topic vector is input into the semantic analysis model, the word meaning is output, and the word meaning is reassembled into a new sentence.

[0035] Determine whether the meaning of the word includes the specified keyword. If so, continue to determine whether the new sentence after the word meaning is recombined forms the specified meaning. If the new sentence forms the specified meaning, the triplet topic vector is considered to be in a situation that requires an alarm, and an alarm message is issued. If the new sentence does not form the specified meaning, the triplet topic vector is considered to be compliant.

[0036] After generating forwarding information and sending the corresponding forwarding status and path to the relevant monitoring router, the monitoring router will perform its attack interception function and execute the interception policy issued by the control center. If the control center has not yet issued the interception policy, the monitoring router will block the attack according to the default interception policy.

[0037] Among them, the corresponding role label is derived based on the subject. The subject can be input into the semantic analysis model to obtain the corresponding role label, or the corresponding mapping relationship can be queried through the subject to obtain the corresponding role label.

[0038] In some preferred embodiments, if the sentence has a parallel structure, then it is necessary to find the verb-object clauses in the single column, and locate and extract the object for each verb-object clause; if the sentence has a verb-object structure, then the object is located and extracted directly.

[0039] In some preferred embodiments, associating multiple subjects into a network structure based on the role labels includes classifying the subjects according to the role labels and establishing the network structure based on the classification.

[0040] In some preferred embodiments, the kernels of both the semantic analysis model and the syntactic model use neural network models.

[0041] Figure 2This is an architecture diagram of the content semantic-based fusion detection system provided in this application. The system includes:

[0042] The building unit is used to build a cloud computing platform on a server and construct syntactic and semantic analysis models;

[0043] The extraction unit is used to acquire the data stream of the Internet platform according to the acquisition strategy, extract the text features of the data stream, vectorize them, and complete the text clustering operation to obtain hotspot clusters generated by category;

[0044] The syntactic analysis unit is used to input each hotspot cluster into the syntactic model, perform fusion syntactic analysis, extract the subject and predicate from several sentences of the hotspot cluster, and discover the dependency relationship between the subject and the action, as well as the correlation relationship between the subject and the object based on the predicate, and analyze the relational structure of the several sentences. The relational structure is a parallel structure or a verb-object structure, and completes the location of the object.

[0045] The template unit is used to derive the corresponding role label based on the subject, and to associate multiple subjects into a network structure based on the role label to complete vectorization and obtain the subject vector; the subject vector, predicate, and located object are input into a preset triplet template to obtain the triplet topic vector;

[0046] The semantic analysis unit is used to input the triplet topic vector into the semantic analysis model, output word meanings, and reorganize the word meanings into new sentences.

[0047] The judgment unit is used to determine whether the meaning of the word includes the specified keyword. If so, it continues to determine whether the new sentence after the word meaning is recombined forms the specified meaning. If the new sentence forms the specified meaning, the triplet topic vector is determined to be in a situation that requires an alarm, and an alarm message is issued. If the new sentence does not form the specified meaning, the triplet topic vector is determined to be compliant.

[0048] This application provides a content semantic-based fusion detection system, the system comprising: a processor and a memory.

[0049] The memory is used to store program code and transmit the program code to the processor;

[0050] The processor is configured to execute the method described in any one of the embodiments of the first aspect according to the instructions in the program code.

[0051] This application provides a computer-readable storage medium for storing program code for performing the method described in any one of the embodiments of the first aspect.

[0052] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and the program, when executed, may include some or all of the steps in the various embodiments of the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0053] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0054] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0055] The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention.

Claims

1. A content semantic-based fusion detection method, characterized in that, The method comprises: building a cloud computing platform on a server, constructing a syntax model and a semantic analysis model; according to an acquisition strategy, acquiring a data stream of an Internet platform, extracting text features of the data stream, vectorizing, and completing text clustering operations to obtain hot spot clusters generated according to categories; inputting each hot spot cluster into a syntax model to perform fused syntax analysis, extracting subjects and predicates from a plurality of sentences of the hot spot cluster, discovering a dependency relationship between the subject and an action and a correlation relationship between the subject and an object according to the predicate, analyzing a relationship structure of the plurality of sentences, the relationship structure being a parallel structure or a verb-object structure, and completing positioning of the object; according to the subject, obtaining corresponding role annotations, associating a plurality of subjects into a mesh structure according to the role annotations, completing vectorization to obtain a subject vector, inputting the subject vector, the predicate, and the positioned object into a preset triple template to obtain a triple topic vector; inputting the triple topic vector into a semantic analysis model to output word meanings, and reorganizing the word meanings into a new sentence; determining whether the word meanings include a specified keyword, if so, determining whether the new sentence formed by reorganizing the word meanings forms a specified meaning, if the new sentence forms the specified meaning, determining that the triple topic vector belongs to a situation requiring an alarm, and issuing an alarm message, if the new sentence does not form the specified meaning, determining that the triple topic vector is compliant.

2. The method of claim 1, wherein: If the relationship structure of the sentence is a parallel structure, it is necessary to find a verb-object structure in a single column, and for the verb-object structure, positioning and extracting the object are completed respectively; if the relationship structure of the sentence is a verb-object structure, the object is directly positioned and extracted.

3. The method of claim 1, wherein: According to the role annotations, the plurality of subjects are classified, and the mesh structure is established according to the classification.

4. The method of claim 1, wherein: According to the role annotations, the plurality of subjects are classified, and the mesh structure is established according to the classification.

5. A content semantic based fusion detection system, characterized in that, The system comprises: a building unit configured to build a cloud computing platform on a server, construct a syntax model and a semantic analysis model; an extraction unit configured to acquire a data stream of an Internet platform according to an acquisition strategy, extract text features of the data stream, vectorize, and complete text clustering operations to obtain hot spot clusters generated according to categories; a syntax analysis unit configured to input each hot spot cluster into a syntax model to perform fused syntax analysis, extract subjects and predicates from a plurality of sentences of the hot spot cluster, discover a dependency relationship between the subject and an action and a correlation relationship between the subject and an object according to the predicate, analyze a relationship structure of the plurality of sentences, the relationship structure being a parallel structure or a verb-object structure, and complete positioning of the object; a template unit configured to obtain corresponding role annotations according to the subject, associate a plurality of subjects into a mesh structure according to the role annotations, complete vectorization to obtain a subject vector, and input the subject vector, the predicate, and the positioned object into a preset triple template to obtain a triple topic vector; a semantic analysis unit configured to input the triple topic vector into a semantic analysis model to output word meanings, and reorganize the word meanings into a new sentence. The semantic analysis unit is configured to input the triadic topic vector into a semantic analysis model, output word meanings, and recombine the word meanings into a new sentence. The judgment unit is configured to determine whether the word meanings include a specified keyword, and if so, determine whether the new sentence formed after recombination of the word meanings forms a specified meaning. If the new sentence forms the specified meaning, it is determined that the triadic topic vector belongs to a situation requiring an alarm, and an alarm message is sent. If the new sentence does not form the specified meaning, it is determined that the triadic topic vector is compliant.

6. A content semantic based fusion detection system, characterized in that, The system comprises: The building unit is configured to build a cloud computing platform on a server, and construct a syntax model and a semantic analysis model. The extraction unit is configured to acquire data streams of an Internet platform according to an acquisition strategy, extract text features of the data streams, vectorize, and complete text clustering operations to obtain hot spot clusters generated by categories. The syntax analysis unit is configured to input each hot spot cluster into a syntax model, perform fused syntax analysis, extract subjects and predicates from several sentences of the hot spot cluster, discover a dependency relationship between the subject and an action and a correlation relationship between the subject and an object according to the predicate, analyze a relationship structure of the several sentences, the relationship structure being a parallel structure or a verb-object structure, and complete positioning of the object. The template unit is configured to derive corresponding role labels according to the subjects, associate multiple subjects into a mesh structure according to the role labels, complete vectorization, obtain subject vectors, input the subject vectors, the predicate, and the positioned object into a preset triadic template, and obtain a triadic topic vector. The semantic analysis unit is configured to input the triadic topic vector into a semantic analysis model, output word meanings, and recombine the word meanings into a new sentence. The judgment unit is configured to determine whether the word meanings include a specified keyword, and if so, determine whether the new sentence formed after recombination of the word meanings forms a specified meaning. If the new sentence forms the specified meaning, it is determined that the triadic topic vector belongs to a situation requiring an alarm, and an alarm message is sent. If the new sentence does not form the specified meaning, it is determined that the triadic topic vector is compliant.

7. A computer readable storage medium characterized by The computer readable storage medium is configured to store program code for implementing the method of any one of claims 1-4.

Images