Digital media content processing method, device and system

Abstract

The invention relates to a digital media content processing method, device and system. The method comprises the following steps: performing logic semantic analysis on digital media data to be detected to obtain digital media logic analysis data; carrying out illegal content identification on identification in the digital media logic analysis data to obtain media word identification data and media logic identification data; under the condition that the media word identification data or the media logic identification data represents that violation information is recognized in the digital media data to be detected, abnormal word filtering is conducted on the digital media data to be detected according to the media word identification data, and initial digital media content filtering data is obtained; performing context exception filtering on the initial digital media content filtering data according to the media logic identification data to obtain target digital media content filtering data; and performing asymmetric content encryption on the target digital media content filtering data to obtain detected digital media data. The efficiency of monitoring and identifying the abnormal content of the digital media can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device and system for processing digital media content. Background Art

[0002] With the development of computer technology, with the support of computer and network technology, technical means are being developed to collect, edit, store, transmit, analyze and display various digital media contents (such as text, images, audio, video, etc.). These methods usually include data compression, format conversion, content enhancement, information retrieval, content recognition, data encryption and other operations, aiming to improve the quality, availability and transmission efficiency of the content. For example, videos can be edited and processed with special effects through video editing software, objects in pictures can be automatically annotated through image recognition technology, and noise can be reduced or sound quality can be improved through audio processing technology. These technologies are widely used in entertainment, education, medical care, advertising and other fields.

[0003] Since digital media often contain content that does not comply with regulations, traditional technologies use manual review to monitor and identify illegal content in the form of text, images, videos, etc. in real time, mark the illegal content and take corresponding measures, such as deleting, blocking or restricting its dissemination. At the same time, user reports are often used as a supplementary means. However, the traditional method of monitoring and identifying abnormal content in digital media requires manual review of the huge media database one by one, and it is also difficult to determine whether the content is abnormal due to personal factors. Therefore, the traditional method leads to low efficiency in monitoring and identifying abnormal content in digital media. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device and system for processing digital media content that can effectively improve the efficiency of monitoring and identifying abnormal content in digital media in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for processing digital media content, comprising: Performing logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data; Identify illegal contents on the digital media logic analysis data to obtain media word identification data and media logic identification data; In the case where the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected, filtering the digital media data to be detected for abnormal words according to the media word identification data to obtain initial digital media content filtering data; According to the media logic identification data, the initial digital media content filtering data is subjected to context anomaly filtering to obtain target digital media content filtering data; Asymmetric content encryption is performed on the target digital media content filtering data to obtain detected digital media data.

[0006] In a second aspect, the present application further provides a digital media content processing device, comprising: A data analysis module is used to perform logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data; A content identification module, used to identify illegal content on the digital media logic analysis data to obtain media word identification data and media logic identification data; A content filtering module, configured to filter abnormal words on the digital media data to be detected according to the media word identification data, and obtain initial digital media content filtering data, when the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected; The content filtering module is further used to perform context anomaly filtering on the initial digital media content filtering data according to the media logic identification data to obtain target digital media content filtering data; The content encryption module is used to perform asymmetric content encryption on the target digital media content filtering data to obtain detected digital media data.

[0007] In a third aspect, the present application further provides a digital media content processing system, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a digital media content processing method when executing the computer program.

[0008] The above-mentioned digital media content processing method, device and system can fully understand the internal structure and semantic relationship of the data by performing logical semantic analysis on the digital media data to be detected, thereby providing a solid foundation for the subsequent identification of illegal content. Then, the identified illegal content is carefully classified to generate media word identification data and media logic identification data to effectively distinguish potential illegal information. Based on these identification data, abnormal word filtering and context abnormality filtering are performed to gradually eliminate inappropriate content in the data to ensure that the filtering results are more accurate and comprehensive. Finally, the detected digital media data is encrypted using asymmetric content encryption technology to ensure the security and privacy of the data. Through multi-level and multi-dimensional digital media data processing methods, it can not only effectively improve the efficiency of monitoring and identifying abnormal content in digital media, but also protect data security through encryption to avoid data leakage or tampering, and has a powerful prevention function; further, because of its adaptation, it is widely used in multiple fields such as digital media platforms, social networks and online content supervision, thereby effectively improving the quality and efficiency of platform content review and ensuring the network security and information security of users and society. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1 is an application environment diagram of a method for processing digital media content in one embodiment; Figure 2 is a flow chart of a method for processing digital media content in one embodiment; Figure 3 A schematic diagram of a flow chart of a first method for obtaining digital media logic parsing data in an embodiment; Figure 4 It is a flowchart of a second method for obtaining digital media logic parsing data in one embodiment; Figure 5 It is a flowchart of a third method for obtaining digital media logic parsing data in one embodiment; Figure 6 A schematic diagram of a flow chart of a method for obtaining media word identification data and media logic identification data in one embodiment; Figure 7 A schematic diagram of a flow chart of a method for obtaining text segment detection data in one embodiment; Figure 8 A schematic diagram of a flow chart of a method for obtaining mathematical logic detection data in one embodiment; Fig. 9 A schematic diagram of a flow chart of a method for obtaining detected digital media data in one embodiment; Fig.10 is a structural block diagram of a digital media content processing device in one embodiment; Fig.11 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0012] The digital media content processing method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the server 104 can be implemented with an independent server or a server cluster composed of multiple servers.

[0013] In an exemplary embodiment, Figure 2 As shown, a method for processing digital media content is provided, and the method is applied to Figure 1 The server in the example is used to illustrate, including the following steps 202 to 210. Among them: Step 202: Perform logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data.

[0014] The digital media data to be detected may be original digital content that needs to be analyzed, processed and screened. These data may include various forms, such as text, images, videos, audio, etc., and may contain potential illegal information or inappropriate content.

[0015] Among them, logical semantic analysis can be an in-depth analysis of the digital media data to be detected through technical means to understand the grammatical structure, semantic information and internal logical relationship of text and mathematical text.

[0016] Among them, the digital media logical analysis data can be data obtained through the logical semantic analysis step, which contains detailed analysis results of the digital media data to be detected, including the grammatical structure, semantic relationship, contextual information, implementation logic, etc. of text and mathematical text, or element recognition and object positioning of images and videos.

[0017] Specifically, through natural language processing technology (such as word segmentation, syntactic analysis, entity recognition, etc.), image content recognition algorithm and video content recognition algorithm, the key information and semantic structure in the digital media data to be detected are extracted, and the text data and mathematical text data are further determined from the key information and semantic structure. For mathematical text data, mathematical structure conversion tools (such as MathJax or KaTeX, etc.) are used to convert the mathematical text data into a structured tree representation so that the model can better understand the logical relationship within the formula; at the same time, symbolic calculation tools (such as SymPy) are integrated to perform symbol-level analysis and verification on the mathematical text data to enhance the model's understanding of mathematical logic. For text data, grammatical analysis and semantic parsing techniques are used to convert the text into a structured logical representation, and key logical units such as premises, arguments and conclusions are identified; then relationship extraction and entity recognition techniques are applied to determine the associations and dependencies between these logical units and construct a logical relationship diagram; then a logical consistency check is performed to ensure that the reasoning steps in the discussion process are reasonable and coherent, and to avoid logical loopholes or contradictions. In addition, it is also possible to adaptively select and compare with a pre-built knowledge base to verify the accuracy and legitimacy of the facts and statements stated in the text. The above-mentioned logical analysis results on the text data and the mathematical text data are integrated to generate the logical analysis data of the digital media.

[0018] Step 204: Identify illegal contents of the digital media logic analysis data to obtain media word identification data and media logic identification data.

[0019] Among them, illegal content identification can be to identify any content contained in digital media logical analysis data that violates laws and regulations, platform rules or social ethics.

[0020] The media word identification data may be identification results indicating specific illegal words in the digital media data to be detected.

[0021] The media logic identification data may be data describing the semantics and logical relationship of illegal content in the digital media data to be detected.

[0022] Specifically, the digital media logic analysis data is used to identify illegal content. Since the digital media logic analysis data includes text content and mathematical text content, for text content, the malicious words, discriminatory language, inflammatory speech and other content that violates the platform regulations are identified by combining sentiment analysis, grammatical analysis, entity recognition and other methods. For example, through vocabulary matching or contextual reasoning, information such as violence, pornography, and inappropriate behavior can be automatically marked. For mathematical text content, although mathematical symbols and formulas themselves do not contain emotional colors, the system will analyze the variables, symbols and structures in the formulas, and combine the mathematical context to identify whether there are misleading or potentially inappropriate expressions or plagiarism (for example, some numbers or symbols may be inappropriate, copyright infringement, academic misconduct, etc.). Through this multi-dimensional illegal content identification, the generated media word identification data will mark all words or symbols that clearly violate the rules, and the media logic identification data will provide the semantics and logical context of the illegal information, such as whether there are misused symbols in a formula, whether the expression conforms to academic norms, etc.

[0023] Step 206 , when the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected, the digital media data to be detected is filtered for abnormal words according to the media word identification data to obtain initial digital media content filtering data.

[0024] Among them, abnormal word filtering can be to clean up inappropriate content in digital media data by replacing, deleting or blocking these illegal words.

[0025] The initial digital media content filtering data may be the result obtained after the abnormal word filtering step, indicating that the digital media data from which the obviously illegal words or content parts have been removed or modified.

[0026] Specifically, when illegal information is identified in the digital media data to be detected represented by the media word identification data or the media logical identification data, the digital media data to be detected will be filtered for abnormal words based on the digital media data to be detected represented by the media word identification data or the media logical identification data. For text content, the system will delete, block or replace the words, phrases or sentences that have been identified as illegal to prevent these inappropriate languages ​​from causing adverse effects on users. For example, the system may replace inappropriate words (such as insulting language, hate speech, etc.) with symbols (such as "*") or delete them directly. In some cases, the system may delete or modify the entire paragraph of text with obvious illegal emotions to ensure that the text content complies with the platform regulations. For mathematical text content, the system will not only identify and filter out illegal content in mathematical symbols or formulas (for example, sensitive numbers, misuse of formulas, etc.), but also adjust inappropriate variable naming or symbol use when necessary. If there is copyright infringement, academic misconduct, etc., the corresponding abnormal characters or words will be automatically deleted or reference marks will be added to ensure that the mathematical expression meets academic and ethical standards. After this process, the initial digital media content filtering data is obtained.

[0027] Step 208: Filter the initial digital media content filtering data for context anomalies according to the media logic identification data to obtain target digital media content filtering data.

[0028] Among them, context anomaly filtering can be to analyze the context and logical structure of the data, and identify those contents that are not illegal at the lexical level but may cause misleading or violate regulations in a specific context.

[0029] The target digital media content filtering data may be the final data after context anomaly filtering, including all compliant data after strict screening and processing.

[0030] Specifically, the key to contextual anomaly filtering of the initial digital media content filtering data is to further identify those contents that may have misleading or inappropriate meanings in a specific context, although there is no obvious violation in individual words or symbols, by analyzing the context and structure of the content. For text content, contextual anomaly filtering not only focuses on the violation of individual words, but also deeply analyzes the semantics of the entire sentence or paragraph. For example, some words may be misinterpreted or distorted in a specific context, such as some words may contain sexist, politically sensitive or inflammatory words in a specific context. For mathematical text content, contextual anomaly filtering focuses on whether the combination and expression of formulas, symbols, and variables conform to conventional mathematical norms, and whether there are misleading or incorrect derivation processes. For example, some mathematical formulas may be ambiguous in expression, leading to misleading understanding, especially when the formula undergoes some disguised changes in the derivation or use process, but in fact there are still acts such as plagiarism and fraud. At the same time, it also pays attention to the formula being used to express views with political or social tendencies. Therefore, the system will further review the formula and its related instructions to ensure its academic and reasonableness. Ultimately, after contextual anomaly filtering, the generated target digital media content filtering data will be more accurate, removing potential illegal or misleading content, thereby ensuring the legality and compliance of the content.

[0031] Step 210: Asymmetric content encryption is performed on the target digital media content filtering data to obtain detected digital media data.

[0032] Among them, asymmetric content encryption can be an encryption technology that uses a pair of public keys and private keys to ensure the security of data.

[0033] The detected digital media data may be a final compliant data version obtained after complete detection and processing.

[0034] Specifically, in order to protect user privacy and data security, the processed target digital media content filtering data is asymmetrically encrypted. Generally, sensitive information is encrypted by using public key and private key encryption algorithms to prevent data from being maliciously obtained or tampered with during transmission. This encryption step ensures data security, especially when processing personal information or data containing sensitive content, to ensure data integrity and confidentiality, ensure compliance and reduce the risk of data leakage.

[0035] In the above-mentioned method for processing digital media content, by performing logical semantic analysis on the digital media data to be detected, the internal structure and semantic relationship of the data can be fully understood, thereby providing a solid foundation for the subsequent identification of illegal content. Then, the identified illegal content is carefully classified to generate media word identification data and media logic identification data to effectively distinguish potential illegal information. Based on these identification data, abnormal word filtering and context abnormality filtering are performed to gradually eliminate inappropriate content in the data to ensure that the filtering results are more accurate and comprehensive. Finally, the detected digital media data is encrypted using asymmetric content encryption technology to ensure the security and privacy of the data. Through multi-level and multi-dimensional digital media data processing methods, not only can the efficiency of monitoring and identifying abnormal content in digital media be effectively improved, but also data security can be protected by encryption to avoid data leakage or tampering, and it has a powerful prevention function; further, because of its adaptation, it is widely used in multiple fields such as digital media platforms, social networks and online content supervision, thereby effectively improving the quality and efficiency of platform content review and ensuring the network security and information security of users and society.

[0036] In an exemplary embodiment, Figure 3 As shown, performing logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data includes steps 302 to 304. Among them: Step 302: Identify digital media text data and digital media formula data from the digital media data to be detected.

[0037] Among them, digital media text data can be content presented mainly in text in digital media, which can come from text data in the form of articles, comments, conversations, social media posts, news reports, etc.

[0038] The digital media formula data may be mainly mathematical formulas, symbols and expressions, which are usually used to describe formulas or theoretical expressions in the fields of mathematics, physics, chemistry, etc. The formula data includes but is not limited to numbers, operators, variables, functions, fractions, integral symbols, etc.

[0039] Specifically, the digital media data to be detected is analyzed through a neural network algorithm to identify digital media text data and digital media formula data. For digital media text data, the system uses text recognition technology (such as OCR, voice recognition, etc.) to extract text content from images, videos or audio; for digital media formula data, it uses a specific formula recognition algorithm to identify mathematical symbols, formulas, expressions and other contents. For example, through image processing technology and symbol recognition algorithms, the system can extract mathematical formulas from images and convert them into machine-understandable formats, such as LaTeX or MathML.

[0040] Step 304: Use a natural language processing model to perform semantic analysis on the logical relationship between the digital media text data and the digital media formula data to obtain digital media logical analysis data.

[0041] Specifically, for digital media text data, the system uses the general language processing sub-model of the natural language processing (NLP) model, such as lexical analysis, syntactic analysis, and semantic analysis, to identify the vocabulary, sentence structure, and grammatical relations in digital media text data, and further extracts information such as entities, relations, and sentiment, which helps the system understand the meaning of the text and determine potential illegal information or logical relations in the text. For digital media formula data, the system uses the specialized mathematical semantic parsing sub-model of the natural language processing (NLP) model to understand the mathematical relations between symbols, operators, variables, etc. in the formula by parsing them. For example, the system can identify and process complex mathematical expressions, deduce the dependencies between variables in the formula, or judge whether the formula has ambiguous or non-compliant use in a specific context based on its structure, and finally generate digital media logical parsing data, that is, a comprehensive representation of the logical relations, structures, and semantic information of text data and formula data, to provide support for subsequent illegal content detection and processing.

[0042] In this embodiment, by identifying digital media text data and digital media formula data from the digital media data to be detected, and using the natural language processing (NLP) model to perform semantic analysis on their logical relationships, the accuracy and efficiency of data processing can be significantly improved. The ability to automatically identify and distinguish text and formula content enables the system to better understand and parse the semantics and logical structure of these data, thereby providing a solid foundation for subsequent illegal content detection, information screening, and context analysis. Through deep semantic analysis, the system can not only more accurately capture the complex logical relationships in the data, but also identify potential abnormal patterns or errors, ensure the quality and credibility of digital media data, and avoid misunderstandings and processing errors.

[0043] In an exemplary embodiment, Figure 4 As shown, using a natural language processing model, semantic analysis is performed on the logical relationship between digital media text data and digital media formula data to obtain digital media logical analysis data, including steps 402 to 406. Among them: Step 402, using a text encoder of a natural language processing model, converts the digital media text data into a word vector form to obtain context-aware word vector data.

[0044] The word vector form may be a mathematical representation method for converting words or phrases into vectors, where each word corresponds to a numerical vector of a fixed length.

[0045] Among them, context-aware word vector data can be a word vector generated through a context environment, which can dynamically adjust its vector representation according to the position and meaning changes of the word in a specific sentence or context.

[0046] Specifically, the text encoder in the natural language processing (NLP) model is used to convert digital media text data (such as articles, comments, conversations, etc.) into a machine-processable word vector form. The text encoder usually uses a pre-trained language model, such as Word2Vec, GloVe, or BERT, to map each word or phrase into a high-dimensional vector in a context-aware manner. In this process, the model not only captures the grammatical information of the word, but also understands the meaning of the word in the context. These word vectors can represent the semantic information of each word in the text and can capture the relationship between different words (such as synonyms, near synonyms, etc.), so the system obtains context-aware word vector data, that is, a set of word vectors with context information.

[0047] Step 404: convert the digital media formula data into an abstract syntax tree format to obtain abstract syntax tree formula data.

[0048] Among them, the abstract syntax tree formula data can be a structured representation used to express the logical structure of a mathematical formula or expression. AST organizes mathematical symbols, operators, variables and other elements in the formula into a tree structure, where each node represents an operator or operand, and the hierarchical structure of the tree represents the priority and dependency of the operations.

[0049] Specifically, digital media formula data (such as mathematical formulas, equations, etc.) is converted into a structured expression form - the abstract syntax tree (AST). Among them, AST is a tree structure that represents the various symbols in the formula and their relationships. Each node represents a mathematical operator, variable or constant, and the hierarchy of the tree reflects the priority and dependency of mathematical operations. For example, in the formula "(a+b)*c", AST will first create a multiplication operator node, then an addition operator node, and finally the involved variable nodes. The conversion process converts the formula from the original symbolic representation to a tree model that can parse its internal structure to obtain the abstract syntax tree formula data, which provides an accurate representation for subsequent formula understanding and processing.

[0050] Step 406, using the mathematical formula encoder of the natural language processing model, converts the abstract syntax tree formula data into a dense matrix form to obtain a digital media formula matrix.

[0051] Among them, the digital media formula matrix can be a mathematical structure that represents a mathematical formula, encoding the symbols, operators and their relationships in the formula in matrix form.

[0052] Specifically, the mathematical formula encoder in the natural language processing model is used to further convert the previously constructed abstract syntax tree formula data into a dense matrix form. A dense matrix is ​​a way to represent mathematical formulas. It uses elements in the matrix (such as numbers, symbols, operators, etc.) to represent the mathematical structure and relationship in the formula. Each matrix element may correspond to an operator or variable in the formula, and the dimension and structure of the matrix reflect the complexity and computational dependency of the formula. In this way, the system converts complex mathematical formulas into a matrix representation, which is convenient for subsequent mathematical derivation, formula verification, or symbolic calculation tasks. The digital media formula matrix, as the final mathematical representation, can effectively support the calculation and reasoning of the formula content.

[0053] In a specific embodiment, consider the mathematical formula "(a+b)*c". First, the system parses the formula into an abstract syntax tree, where the root node represents the multiplication operator *, the left subtree represents the addition operator +, the right subtree represents the variable c, and the + node has a and b as child nodes. Next, the system uses a mathematical formula encoder to convert this AST into a dense matrix. For example, a position in the matrix may represent addition operations (+) and multiplication operations (*), while other positions store variables such as a, b, and c.

[0054] Step 408, based on the text formula dependency, the context-aware word vector data and the digital media formula matrix are integrated to obtain digital media logic analysis data.

[0055] Among them, the text formula dependency can be the mutual connection and dependency between the text content and the formula. In many cases, the language description in the text and the mathematical expression in the formula complement each other, the text explains the formula, and the formula provides mathematical support.

[0056] Specifically, the context-aware word vector data is fused with the digital media formula matrix by analyzing the text formula dependency between the context-aware word vector data and the digital media formula matrix. The text formula dependency refers to the interaction and dependency between the text and the formula, such as the variables in the formula may be mentioned or explained in the text. The system effectively integrates the word vector and the formula matrix by combining these dependencies. In this process, the model can not only connect the semantic information in the text with the structural information of the mathematical formula, but also identify the connection between the text and the formula, ensuring that the information of the two is semantically consistent and complementary, and generating a comprehensive digital media logic analysis data.

[0057] In this embodiment, by converting digital media text data into context-aware word vector data, and converting digital media formula data into an abstract syntax tree and further into a digital media formula matrix, the system can deeply understand the logical structure between text and formula. Effectively convert complex text and formula content into processable mathematical forms, which is convenient for subsequent semantic analysis and logical reasoning. By integrating context-aware word vector data and formula matrices, the system can not only maintain the semantic information of the text, but also combine the mathematical relationship of the formula to achieve more accurate logical parsing. It improves the processing ability of mixed-type data (such as digital media content mixed with text and formulas), enhances the accuracy and flexibility of the system in multimodal data understanding, and can better support complex content analysis, violation detection and other intelligent processing tasks.

[0058] In an exemplary embodiment, Figure 5 As shown, according to the text formula dependency, the context-aware word vector data and the digital media formula matrix are integrated to obtain the digital media logic parsing data, including steps 502 to 504. Among them: Step 502, using the cross-attention mechanism, fuses the context-aware word vector data and the digital media formula matrix to obtain text formula bidirectional dependency fusion data.

[0059] Among them, the cross-attention mechanism can be a technology that extends the self-attention mechanism (Self-Attention), which is designed to handle the interaction between multimodal data (such as text and images, text and formulas). In the cross-attention mechanism, the model uses data from two different sources (such as text vectors and formula matrices) and weights the fusion information by calculating the correlation between them. In the specific implementation, the model uses a query (Query) from the data of one modality and matches it with the data from another modality to generate a weighted representation. For example, in the process of fusing text and formulas, text data (such as word vectors) is used as a query, and formula data (such as formula matrices) is used as keys (Key) and values ​​(Value). The model adjusts the weights of the corresponding parts in the formula based on the context of certain specific words or symbols in the text.

[0060] Among them, the text-formula bidirectional dependency fusion data can be the result of fusing text data and formula data with each other through a cross-attention mechanism and capturing the bidirectional dependency between them. Specifically, there are usually two types of dependencies between text and formulas: the text can describe certain concepts in the formula, and the formula provides a specific mathematical expression for the text. For example, the text may mention "Formula 1" or "Variable X", and a specific mathematical expression or derivation steps may appear in the formula. Bidirectional dependency means that the text and formula not only explain the formula in the text, but the formula can also reversely support the description in the text. This fusion process is achieved through a cross-attention mechanism. The model considers the information in the text and formula at the same time and captures the interaction between them when fusing.

[0061] Specifically, the cross-attention mechanism is used to fuse context-aware word vector data and digital media formula matrix, where the cross-attention mechanism enables data of different modalities (such as text and formula) to influence and combine each other under the same model framework. Therefore, in this process, the text vector (represented by context-aware word vector data) and the formula matrix (a dense matrix generated by encoding mathematical formulas) are used as inputs respectively, and the relationship between text and formula is calculated through the cross-attention mechanism. Further, the cross-attention mechanism calculates the correlation between text and formula (such as the dependency between a variable mentioned in the text and the corresponding variable in the formula), weights and fuses them, and obtains text-formula bidirectional dependency fusion data, which not only contains the independent representation of text and formula, but also reflects the bidirectional dependency between them.

[0062] Step 504, encoding the relative positions of each text information and each formula information of the text formula bidirectional dependency fusion data to obtain digital media logic analysis data.

[0063] Specifically, the system encodes each piece of information in the text-formula bidirectional dependency fusion data, especially the relative position of the text information and the formula information. The "relative position" here refers to the positional relationship between the text and the formula in the entire content, for example, the position of the formula in the text, the order of the elements in the formula, or the logical order between the text and the formula. By encoding this position information, the system can more accurately capture the contextual connection between the text and the formula, as well as their structured relationship in the overall data. Usually, position encoding can be implemented using position embedding technology (such as position encoding in Transformer), so that the model can recognize and utilize the sequential relationship between text and formula. The data after position encoding constitutes the digital media logic parsing data, which contains a comprehensive representation of text and formula, including its semantic information and structural information, to ensure that the system can accurately understand and process the combination of text and formula.

[0064] In this embodiment, by using the cross-attention mechanism to fuse the context-aware word vector data with the digital media formula matrix, the system can effectively capture the bidirectional dependency between text and formula, thereby achieving more accurate semantic understanding and logical reasoning. The cross-attention mechanism can automatically focus on the interaction between text and formulas in different contexts, ensuring that the relationship between text content and mathematical formulas is accurately modeled. Then, by encoding the relative position of the text formula bidirectional dependency fusion data, the system can further optimize the organization and representation of information, making the key information in the text and formula easier to process and analyze. It significantly improves the logical parsing ability of digital media data, ensures the accurate fusion and deep understanding of text and formulas in complex data environments, and provides a more reliable data foundation for subsequent content analysis, anomaly identification and intelligent processing.

[0065] In an exemplary embodiment, Figure 6 As shown, the illegal content is identified on the digital media logic parsing data to obtain media word identification data and media logic identification data, including steps 602 to 608. Among them: Step 602: Perform text segment detection on the digital media logic parsing data to obtain text segment detection data.

[0066] The text segment detection data may be the result of processing and analyzing the text content in the digital media to identify potential illegal or abnormal text segments.

[0067] Specifically, text fragment detection is performed on the text part in the digital media logical parsing data, where the goal of text fragment detection is to identify potential abnormal text or suspicious content, especially those text fragments that may violate regulations or contain illegal information. Through natural language processing (NLP) technology, such as syntactic analysis, sentiment analysis, named entity recognition (NER) and other methods, each text fragment in the digital media logical parsing data is processed to identify possible sensitive words, malicious sentences, hate speech or other illegal content. This process not only focuses on individual words, but also analyzes information such as sentence structure and sentiment tendency to evaluate whether each text fragment has improper behavior or semantic deviation. Finally, text fragment detection data is generated, which contains all text fragments that are marked as potentially violating regulations.

[0068] Step 604: Perform mathematical logic detection on the digital media logic analysis data to obtain mathematical logic detection data.

[0069] Among them, mathematical logic detection can be the process of analyzing and verifying mathematical formulas in digital media to ensure that the formulas conform to mathematical rules in logic and structure.

[0070] The mathematical logic detection data may be detailed information including illegal or erroneous formulas obtained after performing mathematical logic detection on mathematical formulas in digital media.

[0071] Specifically, mathematical logic detection will be performed on the mathematical formula part of the digital media logic analysis data, where the purpose of mathematical logic detection is to ensure the correctness and rationality of the formula in terms of implementation logic, including detecting whether there are logical errors, improper reasoning, or non-conformity with mathematical principles in the formula. For example, the system will check whether the operators in the formula match (such as whether there are symbols that do not conform to mathematical rules or the wrong order of operations), and whether there are self-contradictory or incorrect reasoning steps in the derivation of the formula. Generally, mathematical reasoning algorithms or symbolic calculation tools (such as computer algebra system CAS) are used, and the system can detect logical anomalies in the formula to ensure that the structure and logic of the formula conform to mathematical specifications. Finally, mathematical logic detection data is generated, including all formulas or expressions that are identified to have mathematical logic problems.

[0072] Step 606: annotate the characters with semantic anomalies in the text segment detection data and the mathematical logic detection data to obtain media term identification data.

[0073] Specifically, based on the text segment detection data and mathematical logic detection data generated previously, the semantic anomalies are marked. Semantically abnormal characters usually refer to those words or symbols that do not conform to the semantic rules in the context, for example: incorrect expressions, ambiguous content or words with inappropriate meanings appear in the text; in mathematical formulas, there may be symbols that do not conform to semantics or meaningless operations. In the process of marking, the system usually marks abnormal characters in texts and formulas at the character level (with 1 byte of memory as the minimum search and marking unit), and identifies them as media term identification data, which includes all characters or parts that are considered to be semantically inconsistent, abnormal or potentially illegal.

[0074] Step 608: Mark the data segments with reasoning logic anomalies in the text segment detection data and the mathematical logic detection data to obtain media logic identification data.

[0075] Among them, reasoning logic anomalies can be inconsistencies, contradictions, or unreasonable situations in the reasoning process or logical deduction in the text or formula. Specifically, in the text, reasoning logic anomalies may be manifested as inconsistencies between arguments and conclusions, or logical loopholes in the argumentation process; while in mathematical formulas, reasoning logic anomalies refer to the derivation steps of the formula not conforming to mathematical principles, such as drawing wrong conclusions from known formulas, or the deduction of logical relationships is invalid.

[0076] Specifically, a secondary logical analysis is performed on the data fragments in the text fragment detection and mathematical logic detection. If anomalies are found again, the data fragments with reasoning logic anomalies are marked. Reasoning logic anomalies refer to the situation where there are logical errors or inconsistencies in the reasoning or argumentation process in the text or formula. For example, the arguments in the text do not match the conclusions, or there are unreasonable reasonings in the steps of the formula. Therefore, the system will perform a secondary logical analysis to re-identify these logically inaccurate or non-compliant data fragments to avoid errors caused by the primary analysis. If there are still logical anomalies in the secondary analysis and they are marked, media logic identification data will be generated to indicate the problematic parts of all reasoning processes, helping to make more precise interventions during subsequent processing or manual review.

[0077] In this embodiment, by performing text segment detection and mathematical logic detection on digital media logic analysis data, the system can automatically identify and distinguish semantic anomalies in the text and logical errors in mathematical formulas, thereby effectively improving the accuracy and comprehensiveness of data analysis. Text segment detection helps to discover grammatical, semantic or emotional anomalies in the text, while mathematical logic detection identifies problems such as imprecise reasoning and incorrect conclusions in the formula. These test results provide an accurate basis for labeling and classification. By labeling text and mathematical logic anomalies, the system can clearly identify media word identification data and media logic identification data, and provide effective support for subsequent content filtering, correction or violation identification. It helps to improve the quality of digital media content, ensure that data does not contain misleading information or erroneous reasoning during the dissemination process, and enhance the intelligent processing capabilities of the system.

[0078] In an exemplary embodiment, Figure 7 As shown, performing text segment detection on digital media logic parsing data to obtain text segment detection data includes steps 702 to 706. Among them: Step 702 , performing text similarity detection on the digital media logic parsing data to obtain text similarity abnormal segments.

[0079] Among them, the abnormal text similarity fragment can be a text fragment that is identified in the text similarity detection process and has a high similarity with known illegal, sensitive or inappropriate content.

[0080] Specifically, text similarity detection is performed on the text part in the digital media logical parsing data, where the goal of text similarity detection is to identify text fragments that are highly similar to known illegal or sensitive content. In the specific implementation process, the system calculates the similarity between the text to be detected and the known text library, and uses techniques such as cosine similarity, Jaccard similarity or TF-IDF to compare the similarity between texts. If the similarity of a fragment is higher than the preset threshold, it means that the fragment may have problems such as plagiarism, duplicate content, or borrowing inappropriate information, and the system marks it as an abnormal text similarity fragment.

[0081] Step 704 , in the case of marking abnormal text similarity segments, performing text sensitive information detection on the unmarked portion of the digital media logic parsing data to obtain abnormal text information segments.

[0082] Among them, the abnormal fragment of text information can be a text part containing sensitive, confidential or illegal information identified during the text sensitive information detection process.

[0083] Specifically, after marking the abnormal text similarity fragments, the system will perform text sensitive information detection on the remaining unmarked text of the digital media logic parsing data, where text sensitive information detection refers to identifying the content of personal privacy, confidential data, sensitive political information, financial data, identity information or other content prohibited by laws and regulations. In the specific implementation process, the system usually uses keyword matching, entity recognition (NER) technology or pre-trained models (such as BERT) to detect whether the remaining unmarked text of the digital media logic parsing data contains sensitive information such as ID number, bank card information, address, telephone number, etc. If such sensitive information appears in the text, it will be marked as an abnormal text information fragment.

[0084] Step 706, in the case of marking abnormal fragments of text information, performing text negative sentiment analysis on the unmarked portion of the digital media logic parsing data to obtain text fragment detection data.

[0085] Specifically, after marking the abnormal fragments of text information, the system further performs text negative sentiment analysis on the remaining unmarked text of the digital media logical parsing data, where the purpose of text negative sentiment analysis is to identify whether there are negative emotions or speeches such as negativity, hostility, hate speech, terrorist tendencies, etc. in the text. In the specific implementation process, the system usually uses sentiment analysis technology (such as sentiment dictionary, deep learning model, etc.) to judge the sentiment tendency of the remaining unmarked text of the digital media logical parsing data. If the text contains obvious negative emotions or emotional words, the system will mark these fragments as text fragment detection data containing negative emotions, which can effectively identify harmful emotional content and prevent the spread of negative emotions and negative impact on readers, especially in social media, news releases or public discussions.

[0086] In this embodiment, by performing multi-level text detection on digital media logical parsing data, including text similarity detection, text sensitive information detection, and text negative sentiment analysis, the system can fully identify and mark potential abnormal content. This process first identifies abnormal fragments that may be plagiarized, repeated, or improperly quoted through text similarity detection, and then further performs text sensitive information detection to help the system discover and mark parts involving personal privacy, sensitive information, or illegal content. Finally, through text negative sentiment analysis, text that may contain negative emotions or malicious expressions is identified. Multi-dimensional detection and labeling ensures in-depth analysis of digital media content, effectively avoids the spread of false information, improves the legitimacy and security of content, and provides strong support for subsequent data filtering, content correction, and risk control.

[0087] In an exemplary embodiment, Figure 8 As shown, performing mathematical logic detection on digital media logic analysis data to obtain mathematical logic detection data includes steps 802 to 806. Among them: Step 802, performing error conclusion detection on the digital media logic analysis data to obtain mathematical conclusion degree abnormal data.

[0088] The abnormal data of mathematical conclusions can be the wrong conclusions or parts that do not conform to mathematical rules identified when testing the conclusions drawn in the mathematical formula or reasoning process.

[0089] Specifically, the mathematical part of the digital media logic analysis data is detected for wrong conclusions, wherein the purpose of the wrong conclusion detection is to identify whether an erroneous conclusion is drawn in the formula or reasoning process. In the specific implementation process, the system checks whether there are logical reasoning errors, improper use of operators, or inconsistent derivations between formulas in the mathematical part of the digital media logic analysis data by analyzing each step of the formula derivation. For example, if the conclusion drawn in a mathematical derivation contradicts a known mathematical principle or theorem, the system will mark the conclusion as abnormal. In order to detect erroneous conclusions, the system may combine symbolic calculation tools (such as computer algebra system CAS) and predefined mathematical theorem libraries to verify whether each mathematical conclusion conforms to the rules of logical reasoning. Ultimately, all conclusions identified as erroneous will be marked as mathematical conclusion degree abnormal data, providing a basis for subsequent correction and analysis.

[0090] Step 804, in the case of marking the abnormal mathematical conclusion data, performing false proof detection on the unmarked portion of the digital media logic analysis data to obtain the abnormal mathematical proof data.

[0091] Among them, mathematical proof abnormal data can be the part of the mathematical reasoning or proof process that identifies false or unreasonable proof steps.

[0092] Specifically, after marking the mathematical conclusion abnormal data, the system will perform false proof detection on the mathematical part of the remaining unmarked digital media logic analysis data, where the goal of false proof detection is to identify whether there is an illogical reasoning process or an imprecise and untrue mathematical proof method. In the specific implementation process, the system analyzes the reasoning steps to check whether the mathematical part of the remaining unmarked digital media logic analysis data has a proof process with skipped steps, undefined assumptions, or logical loopholes. A false proof may appear to be correct in form, but in fact there are major flaws in the derivation process or the assumptions are not valid. In order to detect false proofs, the system can use automated reasoning tools, symbolic calculations, and verification mechanisms to check the rigor and legitimacy of each step in the proof. All parts marked as false proofs will be identified as mathematical proof abnormal data, which need to be further reviewed and corrected.

[0093] Step 806, in the case of marking the mathematical proof abnormal data, originality analysis is performed on the unmarked portion of the digital media logic analysis data to obtain mathematical logic detection data.

[0094] Specifically, after marking the abnormal data of mathematical proofs, the system performs originality analysis on the mathematical part of the remaining unmarked digital media logic analysis data, where the purpose of the originality analysis is to determine whether the mathematical reasoning and formulas are innovative, whether they are original content, or whether there is plagiarism and duplication. In the specific implementation process, the system compares the content of the mathematical part of the remaining unmarked digital media logic analysis data with existing literature, formula libraries or patent libraries to check whether there is excessive citation or unauthorized use of other people's formulas or derivation processes. At the same time, by matching existing mathematical literature and formula databases, the system uses machine learning technology to perform plagiarism analysis to determine whether the mathematical derivation is original. Finally, after the originality analysis, the system will obtain mathematical logic detection data, which contains all the identified originality issues, potential plagiarism or duplication content, to ensure the independence and innovation of the mathematical content.

[0095] In this embodiment, by screening the digital media logic analysis data layer by layer, including wrong conclusion detection, false proof detection and originality analysis, the system can accurately identify and mark potential mathematical logic errors to ensure the rigor and credibility of the mathematical content. First, wrong conclusion detection can identify unreasonable conclusions or wrong calculation results in mathematical reasoning to avoid the spread of false information; then, false proof detection further examines the improper deductions or lack of valid proofs in the mathematical proof process to ensure that the conclusions given have sufficient logical support; finally, originality analysis helps identify whether there is unverified or plagiarized content to ensure the originality and credibility of the mathematical content. This series of detections helps to improve the quality of digital media content, avoid the spread of misleading mathematical information or pseudo-scientific content, and enhance the system's accuracy, reliability and security of mathematical data and content.

[0096] In an exemplary embodiment, Fig. 9 As shown, asymmetric content encryption is performed on the target digital media content filtering data to obtain the detected digital media data, including steps 902 to 908. Among them: Step 902 , filtering data according to target digital media content, and determining a data display framework and a data display style.

[0097] Specifically, data is filtered according to the target digital media content, and a suitable data display framework and data display style are designed to display digital media content. The data display framework refers to determining how the data is organized and laid out on the screen. For example, elements such as article content, pictures, tables, and videos need to be arranged reasonably on the page to ensure that the information is conveyed clearly and the logic is smooth. The framework design also includes navigation, column layout, scrolling behavior, etc., so that users can quickly find and access the required content. The data display style focuses on the visual design of the page to ensure that the displayed content meets the visual aesthetics and usability specifications. Including the design of elements such as the font, color, size, paragraph spacing, and background color of the text. These styles and frameworks should be optimized according to the needs of the target user group, device type (such as mobile terminal, PC terminal), and platform design standards to improve the user experience and ensure the readability, accessibility, and aesthetics of the content.

[0098] In a specific embodiment, the design is ensured to meet the expectations and needs of users based on the characteristics of the target user group, including their device usage habits, operating preferences, cultural background, etc. For example, for mobile users, the design should take into account the characteristics of touch screen interaction, smaller screen size, and the need for fast loading, while on the PC side, it may rely more on mouse operation and large screen display, and more complex layouts and more visual elements can be used. Secondly, consider the design standards of different platforms, such as the design guidelines of iOS and Android, and ensure that UI elements (buttons, input boxes, icons, etc.) comply with platform specifications to provide a consistent user experience and smooth interaction. Finally, responsive design and adaptive layout can ensure that the same set of content can automatically adjust the layout and style on different devices to ensure the readability of the content and the convenience of operation. These optimization measures can provide the best visual effects and usage experience based on device characteristics and user needs.

[0099] Step 904, setting the human-computer interaction dialog box and human-computer interaction background data according to the data display framework and data display style.

[0100] Specifically, based on the data display framework and data display style, the human-computer interaction dialog box and background data are further designed. The human-computer interaction dialog box (UI component) refers to the interface elements that users see and operate when interacting with the system, such as text input boxes, buttons, drop-down menus, check boxes, etc. These components need to be clear and intuitive so that users can easily interact with the system, such as submitting feedback, selecting options, searching for content, etc. The human-computer interaction background data refers to the data processing and storage behind these interactive components, including user input information, operation records, user preferences, interaction feedback and other data.

[0101] The human-computer interaction dialog box (UI component) needs to be designed according to the information type and user operation process in the data display framework. For example, if the data display framework displays text and picture content, it may be necessary to design an interactive image box and an editable text input box that supports scrolling and zooming. In addition, the design of the interactive dialog box should also take into account user needs and operational convenience, such as providing clear instructions when users fill out forms, or adding interactive elements such as buttons and tabs when displaying content to facilitate users to perform operations or view more information. The human-computer interaction background data is the part that works closely with these interactive components, including user-entered data, operation records, interaction results, etc. The background data management system is responsible for receiving, processing and storing input from user interactions, ensuring synchronization with the front-end display, and dynamically updating the interactive data and display content through the database or API interface to maintain the real-time and personalized content.

[0102] Step 906 , according to the data display framework, data display style, human-computer interaction dialog box and human-computer interaction background data, set the media content running parameters corresponding to the target digital media content filtering data on different running hardware.

[0103] The media content operation parameters can be various technical configurations and optimization settings that ensure that digital media content can be displayed and run correctly and smoothly in different hardware or platform environments. These parameters usually include the hardware specifications of the device (such as processor performance, memory size, screen resolution, display ratio, etc.), operating system version, browser compatibility, network bandwidth, video or audio playback quality, etc.

[0104] Specifically, it is necessary to ensure that all content can be displayed and run smoothly on different hardware platforms. At this time, it is necessary to configure appropriate media content operation parameters for different hardware platforms (such as PC, smart phone, tablet computer, etc.) according to the data display framework, data display style, human-computer interaction dialog box and human-computer interaction background data. These parameters involve the optimization configuration of hardware performance, including the device's screen resolution, processor speed, memory capacity, battery consumption, etc. For different hardware platforms, the system may adjust the display size, resolution, interaction method and loading speed of the content. For example, on the mobile phone side, it may be necessary to optimize the page layout, font size, clickable area of ​​the button, etc.; on the PC side, richer interface elements and larger display space can be used. This step ensures that digital media content can run smoothly and be accurately presented on any device, thereby providing a seamless and personalized experience for different users.

[0105] Step 908 , asymmetric content encryption is performed on the data display framework, data display style, human-computer interaction dialog box, media content operation parameters, and target digital media content filtering data to obtain detected digital media data.

[0106] Specifically, in order to ensure the security of transmission or storage, it is necessary to perform asymmetric content encryption on the data display framework, data display style, human-computer interaction dialog box, media content operation parameters, and target digital media content filtering data. Asymmetric encryption is a data encryption technology that involves the use of a pair of public keys and private keys for encryption and decryption. The public key is used to encrypt data, while the private key is used to decrypt data. Through this encryption method, only authorized recipients can decrypt and view the data, while unauthorized third parties cannot access it. The encrypted data becomes detected digital media data, ensuring that the integrity and security of the data are fully guaranteed during the publishing and exchange process.

[0107] In this embodiment, by filtering data according to the target digital media content to design the data display framework, display style, and human-computer interaction interface, the system can achieve a more personalized and easy-to-use user interface while ensuring the security and integrity of the data content. First, based on the data display framework and style, the system can adjust the visual presentation according to different device types (such as mobile phones, PCs, etc.) so that the content can be clearly presented and easy to operate on different platforms. Then, by setting the human-computer interaction dialog box and background data, it ensures that the interaction between the user and the system is smooth and intuitive, and provides real-time data support. The system further optimizes the operating parameters of the content according to different hardware environments (such as processor performance, memory, display devices, etc.) to ensure the best performance of digital media content on various devices. Finally, all data is encrypted through asymmetric content encryption technology to ensure that the content will not be illegally tampered with or leaked during data transmission and processing, thereby improving the security and privacy protection of the data. Overall, it not only improves the user experience, but also strengthens the security of the system and the credibility of the content.

[0108] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0109] Based on the same inventive concept, the embodiment of the present application also provides a digital media content processing device for implementing the digital media content processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations of the one or more digital media content processing device embodiments provided below can refer to the limitations of a digital media content processing method above, and will not be repeated here.

[0110] In an exemplary embodiment, Fig.10 As shown, a digital media content processing device is provided, comprising: a data parsing module 1002, a content identification module 1004, a content filtering module 1006 and a content encryption module 1008, wherein: The data analysis module 1002 is used to perform logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data; The content identification module 1004 is used to identify illegal content on the digital media logic analysis data to obtain media word identification data and media logic identification data; The content filtering module 1006 is used to filter the digital media data to be detected for abnormal words according to the media word identification data, and obtain initial digital media content filtering data when the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected; The content filtering module 1006 is further used to perform context anomaly filtering on the initial digital media content filtering data according to the media logic identification data to obtain target digital media content filtering data; The content encryption module 1008 is used to perform asymmetric content encryption on the target digital media content filtering data to obtain the detected digital media data.

[0111] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig.11 The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface.

[0112] In one embodiment, a digital media content processing system is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements steps of a digital media content processing method when executing the computer program.

[0113] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0114] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0115] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. The processor involved in each embodiment provided in this application can be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0116] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing digital media content, characterized in that: The method comprises: Performing logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data; Identify illegal contents on the digital media logic analysis data to obtain media word identification data and media logic identification data; In the case where the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected, filtering the digital media data to be detected for abnormal words according to the media word identification data to obtain initial digital media content filtering data; According to the media logic identification data, the initial digital media content filtering data is subjected to context anomaly filtering to obtain target digital media content filtering data; Asymmetric content encryption is performed on the target digital media content filtering data to obtain detected digital media data.

2. The method according to claim 1, characterized in that The step of performing logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data includes: Identify digital media text data and digital media formula data from the digital media data to be detected; A natural language processing model is used to perform semantic analysis on the logical relationship between the digital media text data and the digital media formula data to obtain the digital media logical analysis data.

3. The method according to claim 2, characterized in that The using of the natural language processing model to semantically parse the logical relationship between the digital media text data and the digital media formula data to obtain the digital media logical parsing data includes: Using the text encoder of the natural language processing model, converting the digital media text data into word vector form to obtain context-aware word vector data; Converting the digital media formula data into an abstract syntax tree form to obtain abstract syntax tree formula data; Using the mathematical formula encoder of the natural language processing model, converting the abstract syntax tree formula data into a dense matrix form to obtain a digital media formula matrix; According to the text formula dependency, the context-aware word vector data and the digital media formula matrix are fused to obtain the digital media logic analysis data.

4. The method according to claim 3, characterized in that The step of fusing the context-aware word vector data and the digital media formula matrix according to the text formula dependency to obtain the digital media logic parsing data includes: Using a cross-attention mechanism, the context-aware word vector data and the digital media formula matrix are fused to obtain text formula bidirectional dependency fusion data; The relative positions of each text information and each formula information of the text formula bidirectional dependency fusion data are encoded to obtain the digital media logic analysis data.

5. The method according to any one of claims 1 to 4, characterized in that: The step of identifying illegal contents of the digital media logic parsing data to obtain media word identification data and media logic identification data includes: Performing text segment detection on the digital media logic parsing data to obtain text segment detection data; Performing mathematical logic detection on the digital media logic analysis data to obtain mathematical logic detection data; Marking the characters with semantic anomalies in the text segment detection data and the mathematical logic detection data to obtain the media term identification data; Furthermore, data segments with reasoning logic anomalies in the text segment detection data and the mathematical logic detection data are marked to obtain the media logic identification data.

6. The method according to claim 5, characterized in that The performing text segment detection on the digital media logic parsing data to obtain text segment detection data includes: Performing text similarity detection on the digital media logic parsing data to obtain text similarity abnormal segments; In the case of marking the abnormal text similarity segment, performing text sensitive information detection on the unmarked portion of the digital media logic parsing data to obtain the abnormal text information segment; In the case of marking the abnormal fragments of the text information, negative sentiment analysis of the text is performed on the unmarked portion of the digital media logic parsing data to obtain the text fragment detection data.

7. The method according to claim 5, characterized in that The performing mathematical logic detection on the digital media logic analysis data to obtain mathematical logic detection data includes: Performing error conclusion detection on the digital media logic analysis data to obtain mathematical conclusion degree abnormal data; In the case of marking the mathematical conclusion abnormal data, performing false proof detection on the unmarked portion of the digital media logic analysis data to obtain mathematical proof abnormal data; In the case where the mathematical proof abnormal data is marked, originality analysis is performed on the unmarked portion of the digital media logic parsing data to obtain the mathematical logic detection data.

8. The method according to claim 1, characterized in that The asymmetric content encryption of the target digital media content filtering data to obtain the detected digital media data includes: Filter data according to the target digital media content, and determine a data display framework and a data display style; According to the data display framework and data display style, the human-computer interaction dialog box and the human-computer interaction background data are set; According to the data display framework, the data display style, the human-computer interaction dialog box and the human-computer interaction background data, setting the media content operation parameters corresponding to different operation hardware of the target digital media content filtering data; Asymmetric content encryption is performed on the data display framework, the data display style, the human-computer interaction dialog box, the media content operation parameters, and the target digital media content filtering data to obtain the detected digital media data.

9. A digital media content processing device, characterized in that: The device comprises: A data analysis module is used to perform logical semantic analysis on the digital media data to be detected to obtain digital media logical analysis data; A content identification module, used to identify illegal content on the digital media logic analysis data to obtain media word identification data and media logic identification data; A content filtering module, configured to filter abnormal words on the digital media data to be detected according to the media word identification data, and obtain initial digital media content filtering data, when the media word identification data or the media logic identification data indicates that illegal information is identified in the digital media data to be detected; The content filtering module is further used to perform context anomaly filtering on the initial digital media content filtering data according to the media logic identification data to obtain target digital media content filtering data; The content encryption module is used to perform asymmetric content encryption on the target digital media content filtering data to obtain detected digital media data.

10. A digital media content processing system, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images