Intelligent copywriting analysis method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of copywriting analysis. The intelligent copywriting analysis method comprises the steps that based on user input information, an analysis task is generated, a task identifier is distributed to the analysis task, the analysis task is stored in an analysis storage database, meanwhile, the state of the analysis task is set to be an initial state, and the analysis task is stored in the analysis storage database; the initial state comprises to-be-executed or in-execution; when the state of the analysis task is to be executed, determining the execution opportunity of the analysis task based on a preset task scheduling strategy, and updating the state of the analysis task to be in execution; when the state of the analysis task is in execution, reading corresponding user input information from an analysis storage database according to the task identifier; semantic features and syntactic structures of information input by a user are extracted, key information in a text is recognized, and a corresponding analysis task type is matched according to the key information; and obtaining an analysis calculation result. The method has the effect of improving the copywriting analysis efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of document analysis, and in particular to an intelligent document analysis method, device, computer equipment, and storage medium. Background Art

[0002] Currently, copywriting analysis primarily relies on rule-based parsing methods, traditional machine learning, and deep learning technologies. Rule-based methods parse text content through keyword matching or grammatical templates and are suitable for standardized analysis in specific fields. Machine learning methods rely on feature engineering to perform text classification, clustering, or sentiment calculation. Deep learning models (such as BERT and GPT) further enhance the understanding of text context, making copywriting analysis more accurate. However, existing technologies still have limitations in task scheduling, process management, and adaptive optimization, which impacts task execution efficiency and system flexibility, making it difficult to meet diverse business needs. This is particularly problematic in high-concurrency environments. Summary of the Invention

[0003] In order to improve the efficiency of copy analysis, the present application provides an intelligent copy analysis method, apparatus, computer equipment and storage medium.

[0004] The above-mentioned invention objective of this application is achieved through the following technical solutions: An intelligent document analysis method, comprising: Obtaining user input information, generating an analysis task based on the user input information, assigning a task identifier to the analysis task, storing the analysis task in an analysis storage database, and setting the state of the analysis task to an initial state, wherein the initial state includes pending execution or executing; When the status of the analysis task is "to be executed", based on a preset task scheduling strategy, determining the execution timing of the analysis task, and updating the status of the analysis task to "in execution"; When the status of the analysis task is in execution, reading corresponding user input information from the analysis storage database according to the task identifier; Extracting semantic features and syntactic structures of the user input information, identifying key information in the text, and matching corresponding analysis task types according to the key information; Based on the analysis task type, the corresponding analysis process is called to obtain analysis and calculation results, which include optimized copywriting, sentiment analysis results, and translated text; The analysis calculation results are stored in the analysis storage database, and the status of the analysis task is updated, where the status includes being executed, completed, and failed.

[0005] By adopting the above technical solution, based on a preset task scheduling strategy, the execution timing of analysis tasks is determined and the status of the analysis tasks is updated to "in progress." This ensures that high-priority tasks can be processed quickly and avoids inefficient sequential execution. Furthermore, based on the usage of computing resources, analysis tasks are scheduled for execution when resources are available and delayed if resources are unavailable, enabling multi-task parallel processing and improving computing throughput. In terms of data processing, semantic features and syntactic structures are extracted based on user input information, key information in the text is identified, and the corresponding analysis task type is matched. This accurately identifies task requirements, reduces invalid calculations, and improves execution efficiency. Furthermore, the analysis and calculation results are stored in an analysis storage database and the status of the analysis task is updated. This not only makes task results traceable but also allows them to be directly reused when similar tasks arise, reducing duplicate calculations and speeding up document analysis. Task execution and querying are independent of each other. The front-end only needs to query the task status to understand the progress, avoiding waiting for execution completion to obtain results and improving user response speed.

[0006] In a preferred example, the present application may be further configured as follows: obtaining user input information, generating an analysis task based on the user input information, assigning a task identifier to the analysis task, and storing the analysis task in an analysis storage database, while setting the state of the analysis task to an initial state, wherein the initial state includes pending execution or executing, including: Use preset input validation rules to check the integrity and format of the user input information, and generate an error message if the user input information is missing parameters; If the user input information meets the preset input verification rule, analyzing the type of the user input information to obtain the analysis task; Assigning the task identifier to the analysis task through a unique identifier, and storing the analysis task and the task identifier in the analysis storage database; According to the execution mode of the analysis task, the initial state is set to the pending execution or the executing state. If the analysis task needs to queue for scheduling, the initial state is set to the pending execution. If the analysis task does not need to queue and can be executed directly, the initial state is set to the executing state.

[0007] By adopting the above technical solution and using the preset input validation rules to check the integrity and format of the user input information, it is possible to effectively avoid task execution failures due to input errors or missing parameters and improve data quality. At the same time, if missing information is detected, the system can generate an error prompt message to guide the user to complete the content, thereby reducing unnecessary task termination or repeated submission. In terms of task management, if the user input information meets the preset input validation rules, the type of user input information is analyzed to obtain the analysis task, which can ensure that the task classification accurately matches the analysis process and reduce the waste of computing resources caused by incorrect classification. In the task allocation link, a task identifier is assigned to the analysis task through a unique identifier, and the analysis task and task identifier are stored in the analysis storage database to ensure the traceability of the task and make subsequent queries and execution more efficient.

[0008] In a preferred example, the present application can be further configured as follows: the preset task scheduling strategy includes: Assigning execution priorities to tasks based on the urgency, business requirements, and user levels of the analysis tasks, and adjusting the execution order of tasks according to their priorities; Sort the analysis tasks according to their creation time and use a first-in-first-out control strategy to adjust the execution order of the tasks; Based on the current usage of available computing resources, the analysis tasks are scheduled for execution when resources are available, and delayed if resources are not available, and the execution order of tasks is dynamically adjusted according to the load balancing principle; When the analysis task has a dependency relationship with other tasks, the dependency relationship is parsed based on the task dependency tree, and the execution of the analysis task is arranged after the dependency conditions are met.

[0009] By adopting the above technical solution, the execution priority of tasks is assigned according to the urgency of the analysis tasks, business needs and user levels, and the execution order of tasks is adjusted according to the priority level, so that high-priority tasks can be processed first, avoiding the delay of critical tasks due to the occupation of resources by low-priority tasks, thereby improving the response speed of tasks. In addition, the analysis tasks are sorted according to the creation time, and the execution order of tasks is adjusted using a first-in-first-out control strategy to ensure that tasks are executed in the order of submission time, avoiding long-term backlog of tasks and improving the fairness and orderliness of task processing. At the same time, according to the current usage of available computing resources, the analysis tasks are scheduled for execution when resources are available, and delayed if resources are not available. The execution order of tasks is dynamically adjusted according to the load balancing principle, which can reasonably allocate computing resources under high load conditions and prevent resource overload.

[0010] In a preferred example, the present application may be further configured as follows: extracting the semantic features and syntactic structure of the user input information, identifying key information in the text, and matching the corresponding analysis task type according to the key information, including: Performing text preprocessing on the user input information to obtain the standardized text; Based on the standardized text, using a pre-trained language model to extract grammatical structure and semantic features, parse sentence components, identify syntactic relationships in the text, and identify the key information in the standardized text; According to the key information, the corresponding analysis task type is matched.

[0011] By adopting the above technical solution, user input information is preprocessed to obtain standardized text. This can standardize the input format, eliminate redundant characters, special symbols, and invalid information, ensure the consistency of text data, and thus reduce task recognition errors caused by formatting issues. In terms of text understanding and parsing, based on standardized text, a pre-trained language model is used to extract grammatical structure and semantic features, parse sentence components, identify syntactic relationships in the text, and identify key information in the standardized text. This can fully utilize natural language processing technology to deeply understand the structure and meaning of the text, ensure that the core information in the text is accurately extracted, and avoid task recognition deviations caused by grammatical complexity or ambiguity. At the same time, according to key information, matching the corresponding analysis task type can ensure that different types of copy analysis tasks (such as copy generation, sentiment analysis, or text translation) are accurately classified, making subsequent processing more efficient and avoiding waste of computing resources or task execution failures due to incorrect matching.

[0012] In a preferred example, the present application may be further configured as follows: based on the analysis task type, the corresponding analysis process is called to obtain analysis calculation results, wherein the analysis calculation results include optimized copywriting, sentiment analysis results, and translated text, including: According to the analysis task type, identifying whether the analysis process is a copywriting generation, comment analysis, or text translation process; When the analysis process corresponding to the analysis task type is the copywriting generation, based on the keywords, prompts and negation words in the user input information, according to the preset copywriting generation rules, a copywriting that conforms to grammatical specifications is constructed, and the syntactic structure and vocabulary selection of the copywriting are adjusted to obtain the optimized copywriting; When the analysis process corresponding to the analysis task type is the comment analysis, parsing the user input information, using a text sentiment analysis model to identify the sentiment tendency of the text, and calculating the proportion of different sentiment categories to obtain the sentiment analysis result; When the analysis process corresponding to the analysis task type is text translation, the user input information is translated through a translation model based on the target language set by the user, the terminology consistency is adjusted in combination with the context information, and the syntactic structure is optimized to obtain the translated text that conforms to the expression specifications of the target language.

[0013] By adopting the above technical solution, according to the analysis task type, the analysis process can be identified as belonging to copy generation, comment analysis or text translation, which can ensure that different types of tasks are correctly classified and avoid resource waste or calculation errors caused by incorrect matching. In terms of copy generation, when the analysis process corresponding to the analysis task type is copy generation, based on the keywords, prompts and negative words of the user input information, according to the preset copy generation rules, a copy that conforms to grammatical specifications is constructed, and the syntactic structure and vocabulary selection of the copy are adjusted to obtain an optimized copy, ensuring that the generated copy not only conforms to the language expression specifications but also meets user needs, thereby improving the readability and accuracy of the copy. In terms of comment analysis, when the analysis process corresponding to the analysis task type is comment analysis, the user input information is parsed, the text sentiment tendency is identified using the text sentiment analysis model, and the proportion of different sentiment categories is calculated to obtain the sentiment analysis results. This can quickly identify the text sentiment attitude, help users understand the sentiment distribution of the text, and improve the accuracy and execution efficiency of sentiment analysis. In terms of text translation, when the analysis process corresponding to the analysis task type is text translation, the user input information is translated through the translation model according to the target language set by the user, the terminology consistency is adjusted in combination with the context information, and the syntactic structure is optimized to obtain a translation text that conforms to the expression specifications of the target language, ensuring that the translation text meets the target language standards in terms of grammar, semantics and terminology consistency, thereby improving the accuracy and readability of the translation.

[0014] In a preferred example, the present application may be further configured as follows: based on the analysis task type, a corresponding analysis process is called to obtain analysis calculation results, wherein the analysis calculation results include optimized copywriting, sentiment analysis results, and translated text, and further includes: When the analysis process corresponding to the analysis task type is user feedback optimization, obtaining the user's interactive behavior information on the analysis calculation result; According to the interactive behavior information, the strategy parameters of the analysis process are adjusted to obtain adjusted strategy parameters; based on the adjusted strategy parameters, the preset copy generation rules, the parameters of the text sentiment analysis model or the parameters of the translation model are updated.

[0015] By adopting the above technical solution, when the analysis process corresponding to the analysis task type is optimized for user feedback, information on user interaction behavior on the analysis calculation results is obtained. This allows real-time capture of user actions such as modification, likes, selection, or evaluation of the analysis results, providing data support for subsequent optimization. During the optimization process, the strategy parameters of the analysis process are adjusted based on the interaction behavior information, and the adjusted strategy parameters are obtained. Based on the adjusted strategy parameters, the preset copy generation rules, the parameters of the text sentiment analysis model, or the parameters of the translation model are updated, allowing the system to continuously adjust the optimization strategy based on actual user usage. For example, in a copy generation task, if a user frequently modifies a certain sentence structure or vocabulary, the system can adjust the copy generation rules to make subsequent copies more in line with user preferences; in a sentiment analysis task, if a user objects to the sentiment classification of certain texts, the parameters of the classification model are adjusted to improve classification accuracy; in a translation task, if a user repeatedly modifies the translation of certain terms, the system can update the terminology library to ensure that the terms remain consistent in subsequent translation tasks.

[0016] In a preferred example, the present application may be further configured as follows: adjusting the strategy parameters of the analysis process according to the interaction behavior information to obtain the adjusted strategy parameters; and updating the preset copywriting generation rules, the parameters of the text sentiment analysis model, or the parameters of the translation model based on the adjusted strategy parameters, including: Analyzing the user's modification frequency, content adjustment method, and vocabulary replacement record of the generated copy based on the interaction behavior information to obtain adjusted copy generation strategy parameters, and updating the preset copy generation rules based on the adjusted strategy parameters; Optimizing the classification threshold, feature extraction method, and sentiment dictionary of the text sentiment analysis model according to the interactive behavior information to obtain adjusted sentiment analysis strategy parameters, and updating the parameters of the text sentiment analysis model based on the adjusted strategy parameters; According to the interactive behavior information, term matching, syntax adjustment and context relevance of the translation model are optimized, a term library is updated, adjusted translation strategy parameters are obtained, and parameters of the translation model are updated based on the adjusted strategy parameters.

[0017] By adopting the above technical solution, based on interaction behavior information, the user's modification frequency, content adjustment methods, and vocabulary replacement records of generated copy are analyzed to obtain adjusted copy generation strategy parameters. Based on the adjusted strategy parameters, the preset copy generation rules are updated, enabling the system to automatically learn the user's copy preferences. For example, if a user frequently modifies certain sentence structures or vocabulary, the system will adjust the copy generation strategy to make the subsequent generated content more consistent with the user's style and expression habits, thereby reducing the user's manual adjustment costs and improving the efficiency and quality of copy generation. In terms of sentiment analysis optimization, based on interaction behavior information, the classification threshold, feature extraction method, and sentiment dictionary of the text sentiment analysis model are optimized to obtain adjusted sentiment analysis strategy parameters. Based on the adjusted strategy parameters, the parameters of the text sentiment analysis model are updated to ensure more accurate sentiment classification. For example, if a user frequently modifies the sentiment classification of certain texts, the system can optimize the sentiment analysis results by adjusting the classification threshold to make them more consistent with the actual context. At the same time, it optimizes the feature extraction method to improve the system's ability to recognize implicit emotions and enhance the model's adaptability. In terms of translation optimization, based on interaction behavior information, the translation model's terminology matching, syntactic adjustments, and contextual relevance are optimized. The terminology database is updated to obtain adjusted translation strategy parameters. Based on these adjusted strategy parameters, the translation model parameters are updated, enabling the system to learn the user's terminology preferences and translation style. For example, if a user repeatedly modifies the translation of certain professional terms, the system can automatically update the terminology database to ensure terminology consistency in subsequent translation tasks. At the same time, the syntactic adjustment strategy is optimized to make the translation more consistent with the target language's expression habits, improving translation quality and readability.

[0018] The second object of the present invention is achieved through the following technical solutions: An intelligent document analysis device, comprising: A task creation and initialization module is used to obtain user input information, generate an analysis task based on the user input information, assign a task identifier to the analysis task, and store the analysis task in an analysis storage database, while setting the status of the analysis task to an initial state, wherein the initial state includes pending execution or executing; A task scheduling and execution module is used to determine the execution time of the analysis task based on a preset task scheduling strategy when the status of the analysis task is the pending execution, and update the status of the analysis task to the executing state; a text parsing and feature extraction module, configured to read corresponding user input information from the analysis storage database according to the task identifier when the status of the analysis task is in execution; The task type matching and process calling module is used to extract the semantic features and syntactic structure of the user input information, identify the key information in the text, and match the corresponding analysis task type according to the key information; An analysis calculation and result generation module is used to call the corresponding analysis process based on the analysis task type to obtain analysis calculation results, which include optimized copy, sentiment analysis results and translated text; The result storage and task status management module is used to store the analysis and calculation results in the analysis storage database and update the status of the analysis task, which includes the status of executing, completed and failed.

[0019] By adopting the above technical solution, based on a preset task scheduling strategy, the execution timing of analysis tasks is determined and the status of the analysis tasks is updated to "in progress." This ensures that high-priority tasks can be processed quickly and avoids inefficient sequential execution. Furthermore, based on the usage of computing resources, analysis tasks are scheduled for execution when resources are available and delayed if resources are unavailable, enabling multi-task parallel processing and improving computing throughput. In terms of data processing, semantic features and syntactic structures are extracted based on user input information, key information in the text is identified, and the corresponding analysis task type is matched. This accurately identifies task requirements, reduces invalid calculations, and improves execution efficiency. Furthermore, the analysis and calculation results are stored in an analysis storage database and the status of the analysis task is updated. This not only makes task results traceable but also allows them to be directly reused when similar tasks arise, reducing duplicate calculations and speeding up document analysis. Task execution and querying are independent of each other. The front-end only needs to query the task status to understand the progress, avoiding waiting for execution completion to obtain results and improving user response speed.

[0020] The third objective of this application is achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the intelligent document analysis method are implemented.

[0021] The fourth objective of this application is achieved through the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned intelligent copywriting analysis method.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. Based on the preset task scheduling strategy, determine the execution time of the analysis task and update the status of the analysis task to executing, ensuring that high-priority tasks can be processed quickly and avoiding inefficient sequential execution modes. At the same time, combined with the usage of computing resources, schedule the analysis task execution when resources are available, and delay execution if resources are not available, so as to achieve multi-task parallel processing and improve computing throughput. In terms of data processing, according to user input information, extract semantic features and syntactic structures, identify key information in the text, and match the corresponding analysis task type, accurately identify task requirements, reduce invalid calculations, and improve execution efficiency. In addition, the analysis calculation results are stored in the analysis storage database, and the status of the analysis task is updated, which not only makes the task results traceable, but also can be directly reused when similar tasks appear, reducing repeated calculations and speeding up the text analysis. Task execution and query are independent of each other. The front end only needs to query the task status to understand the progress, avoiding waiting for execution to complete to obtain results, and improving user response speed; 2. According to the analysis task type, the identification analysis process belongs to a type of copy generation, comment analysis or text translation, which can ensure that different types of tasks are correctly classified and avoid waste of resources or calculation errors caused by incorrect matching. In terms of copy generation, when the analysis process corresponding to the analysis task type is copy generation, based on the keywords, prompts and negative words of the user input information, according to the preset copy generation rules, a copy that conforms to grammatical specifications is constructed, and the syntactic structure and vocabulary selection of the copy are adjusted to obtain an optimized copy, ensuring that the generated copy not only conforms to the language expression specifications but also meets user needs, thereby improving the readability and accuracy of the copy. In terms of comment analysis, when the analysis process corresponding to the analysis task type is comment analysis, the user input information is parsed, the text sentiment analysis model is used to identify the text sentiment tendency, and the proportion of different sentiment categories is calculated to obtain the sentiment analysis results, which can quickly identify the text sentiment attitude, help users understand the sentiment distribution of the text, and improve the accuracy and execution efficiency of sentiment analysis. In terms of text translation, when the analysis task type corresponds to the analysis process of text translation, the user input information is translated through the translation model according to the target language set by the user, and the terminology consistency is adjusted based on the context information, and the syntactic structure is optimized to obtain the translation text that conforms to the expression standards of the target language. This ensures that the translation text meets the target language standards in terms of grammar, semantics and terminology consistency, thereby improving the accuracy and readability of the translation. 3. Based on the preset task scheduling strategy, the execution timing of the analysis task is determined and the status of the analysis task is updated to "in progress". This ensures that high-priority tasks can be processed quickly and avoids inefficient sequential execution modes. At the same time, based on the usage of computing resources, the analysis task is scheduled for execution when resources are available. If resources are not available, execution is delayed, enabling multi-task parallel processing and improving computing throughput. In terms of data processing, semantic features and syntactic structures are extracted based on user input information, key information in the text is identified, and the corresponding analysis task type is matched. This accurately identifies task requirements, reduces invalid calculations, and improves execution efficiency. In addition, the analysis and calculation results are stored in the analysis storage database and the status of the analysis task is updated. This not only makes the task results traceable, but also allows them to be directly reused when similar tasks arise, reducing repeated calculations and speeding up document analysis. Task execution and query are independent of each other. The front end only needs to query the task status to understand the progress, avoiding waiting for execution completion to obtain results and improving user response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of an intelligent document analysis method in one embodiment of the present application; Figure 2 This is a flowchart for implementing step S10 in the intelligent document analysis method in one embodiment of the present application; Figure 3 This is a flowchart for implementing step S20 in the intelligent document analysis method in one embodiment of the present application; Figure 4 This is a flowchart for implementing step S40 in the intelligent document analysis method in one embodiment of the present application; Figure 5 This is a flowchart for implementing step S50 in the intelligent document analysis method in one embodiment of the present application; Figure 6 This is a flowchart of an implementation after step S60 in the intelligent document analysis method in one embodiment of the present application; Figure 7 This is a flowchart for implementing step S602 in the intelligent document analysis method in one embodiment of the present application; Figure 8 This is a principle block diagram of an intelligent document analysis device in one embodiment of the present application; Figure 9 It is a schematic diagram of a device in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, if Figure 1 As shown, this application discloses an intelligent copywriting analysis method, which specifically includes the following steps: S10: Obtain user input information, generate an analysis task based on the user input information, assign a task identifier to the analysis task, and store the analysis task in an analysis storage database. At the same time, set the status of the analysis task to an initial state, which includes to be executed or being executed.

[0026] In this embodiment, the analysis storage database refers to a data management system for storing analysis task data, task status, and analysis calculation results.

[0027] Specifically, the user input information can be text content, keywords, instructions or task requirements, which can be obtained through the interface input box, API call or batch data import. When the user enters text content, the input box supports manual editing or pasting of content. When the user enters keywords, multiple keywords can be entered and separated by commas. When the user enters instructions, the user requirements can be parsed through the preset command format to generate the corresponding analysis task. The task identifier of the analysis task can be obtained through random generation, user customization or system internal rule setting. When stored in the analysis storage database, it can be classified and stored according to the task type, and the creation time, user ID and task status information can be attached to each task.

[0028] S20: When the status of the analysis task is to be executed, the execution timing of the analysis task is determined based on a preset task scheduling strategy, and the status of the analysis task is updated to being executed.

[0029] In this embodiment, the preset task scheduling strategy refers to a strategy for dynamically adjusting the order of task execution based on factors such as task priority, computing resource status, and task dependencies.

[0030] Specifically, the task scheduling strategy can comprehensively determine the execution time based on factors such as task priority, creation time, and availability of computing resources. The priority can be set according to task type, user settings, or system default rules. For example, urgent tasks have high priority and ordinary tasks have low priority. Task creation time can be used for sorting and scheduling, and execution is carried out in chronological order. When computing resources are limited, the execution order can be dynamically adjusted according to task priority to ensure that high-priority tasks are executed first. Resource availability can be judged based on the current CPU, memory, and storage usage. When computing resources are sufficient, the task is executed immediately. When resources are insufficient, the task can be temporarily stored and delayed for execution after resources are released to obtain the execution time of the task and update the task status.

[0031] S30: When the status of the analysis task is in execution, the corresponding user input information is read from the analysis storage database according to the task identifier. Specifically, the task identifier is used to uniquely identify the analysis task. When retrieving from the analysis storage database, index matching can be performed based on the task identifier to improve data query efficiency. The retrieval method can be direct query, fuzzy matching or batch reading. When the database adopts a relational structure, precise query can be performed based on the index field. When the database adopts a non-relational structure, data can be obtained based on key-value mapping. Data integrity must be ensured when reading data. If the task data is lost or damaged, the exception handling mechanism can be triggered, including re-requesting data, returning error information or marking the task as failed, to ensure that complete user input information is obtained to obtain user input information.

[0032] S40: Extract the semantic features and syntactic structure of the user input information, identify the key information in the text, and match the corresponding analysis task type according to the key information.

[0033] Specifically, semantic feature extraction can be performed based on text segmentation, part-of-speech tagging, syntactic analysis, etc. Word segmentation can split the text by rule matching or based on a dictionary. Part-of-speech tagging can be performed according to dictionary rules or statistical models. Syntactic analysis can parse the text structure based on dependency relationships. Key information can be identified through keyword matching or pattern recognition. For example, when the input information contains specific words or phrases, the corresponding task type can be matched. The task matching method can be based on preset rules or machine learning models. The rule matching method can be matched through regular expressions or pattern tree structures. The machine learning model can learn text features through training data and predict the corresponding task type based on text input to obtain a matching analysis task type.

[0034] S50: Based on the analysis task type, the corresponding analysis process is called to obtain analysis calculation results, which include optimized copy, sentiment analysis results, and translated text.

[0035] Specifically, the call of the analysis process determines the analysis method based on the task type. The copy generation task calls the text generation module, inputs keywords, prompts and negative words, and constructs a complete copy in combination with grammar rules. The sentiment analysis task calls the sentiment classification module to judge the sentiment tendency of the text and calculate the ratio of positive, neutral and negative sentiments. The translation task calls the translation processing module to perform translation according to the set target language, match terms for consistency during translation, and adjust the syntactic structure to ensure semantic fluency. After executing the analysis process, the corresponding analysis and calculation results are generated to obtain optimized copy, sentiment analysis results or translated text.

[0036] S60: The analysis and calculation results are stored in the analysis storage database, and the status of the analysis task is updated, including the status being executed, completed, and failed.

[0037] Specifically, when stored in the analysis storage database, the analysis and calculation results are archived and stored according to the task identifier, and a timestamp is attached to record the task execution time. When updating the task status, if the analysis task is completed successfully, the status is updated to completed, and the task completion time is marked. If an exception occurs during the task execution, such as input data error, calculation failure or timeout, the status is updated to failure, and an error log is recorded. Failed tasks can decide whether to re-execute or terminate the task based on the error type. Task status management can optimize storage space by regularly cleaning up expired data or archiving completed tasks to obtain stored analysis and calculation results and updated task status.

[0038] In one embodiment, if Figure 2 As shown, in step S10, user input information is obtained, an analysis task is generated based on the user input information, a task identifier is assigned to the analysis task, and the analysis task is stored in the analysis storage database. At the same time, the state of the analysis task is set to the initial state, and the initial state includes waiting to be executed or executing, including: S101: Use preset input verification rules to check the integrity and format of user input information. If the user input information is missing parameters, generate an error prompt message.

[0039] In this embodiment, the preset input verification rules refer to rules such as data integrity check, format constraint check, character legitimacy detection and content specification review formulated according to different types of user input information, which are used to ensure that the input information meets the processing requirements of the analysis task and avoid task execution failure or inaccurate calculation results due to input anomalies.

[0040] Specifically, when performing data integrity check, first obtain the field list entered by the user and compare it with the list of required fields required by the task. If it is found that a necessary field is missing, an error prompt message is generated, indicating the missing field name, and prompting the user to supplement it. When performing format specification verification, first parse the format of the user input content and match it with the preset format template. If the format does not match, an error prompt message is generated. For example, when the user enters an incorrect date format, the prompt "Date format is incorrect, please use YYYY-MM-DD format" is displayed. When the user enters numeric data out of range, the prompt "Input value exceeds the allowable range" is displayed. Please enter a value between 0 and 100. When performing character validity detection, the input content is scanned to see if it contains illegal characters, special symbols, or system-restricted words. If illegal characters are detected, an error message is generated, indicating the characters or words that are not allowed. When performing content standard review, the input information is analyzed to see if it complies with preset industry standards or semantic rules. For example, in the sensitive content filtering scenario, if the input text contains banned words, an error message is generated and modification suggestions are provided so that the user can adjust the input content to obtain user input information that complies with the input verification rules or the corresponding error message.

[0041] S102: If the user input information meets the preset input verification rules, the type of the user input information is analyzed to obtain an analysis task.

[0042] Specifically, task type analysis can be based on methods such as keyword matching, text pattern recognition, and semantic parsing. Keyword matching can be searched through a preset task keyword library. For example, when the user input contains words related to "generate copy", it can be identified as a copy generation task. Text pattern recognition can parse input information based on regular expressions or rule tree structures to determine whether the input content conforms to a specific task pattern. Semantic parsing can extract contextual semantic information of the text through natural language processing technology (NLP) and predict the most matching task type through a machine learning model. The determination of the task type can be combined with multiple recognition methods for comprehensive judgment to obtain an analysis task that meets the user input requirements.

[0043] S103: Assign a task identifier to the analysis task through the unique identifier, and store the analysis task and the task identifier in the analysis storage database.

[0044] Specifically, the generation of a unique identifier can be based on a timestamp, random number, hash calculation or serial number. For example, a timestamp combined with a user ID can be used to generate a unique identifier to ensure that the identifiers of different tasks will not be repeated. The identifier can also be allocated based on a UUID (universally unique identifier) or a database auto-increment ID. When stored in the analysis and storage database, it can be classified and stored according to the task type, and information such as the task creation time, user ID and task status can be attached. The storage method can use a relational database (such as MySQL) for structured storage, or a non-relational database (such as MongoDB, Redis) for flexible storage to obtain analysis task data with a unique identifier.

[0045] S104: According to the execution mode of the analysis task, the initial state is set to pending or executing. If the analysis task needs to queue for scheduling, the initial state is set to pending. If the analysis task does not need to queue and can be executed directly, the initial state is set to executing.

[0046] Specifically, the execution method of the task can be determined based on the urgency of the task, the availability of computing resources and the task dependencies. If the task needs to be executed immediately, the task status is set to executing and the calculation phase is entered immediately. If the task needs to wait for other tasks to be completed or queued for scheduling, the task status is set to pending and prioritized according to the preset scheduling strategy. It enters the execution state after the execution conditions are met. When the task status changes, a log can be stored to record the time point of the task status change to ensure that the task execution process is traceable, so as to obtain an analysis task with a correctly set initial state.

[0047] In one embodiment, if Figure 3 As shown, in step S20, the preset task scheduling strategy includes: S201: Assigning execution priorities to tasks based on the urgency, business requirements, and user levels of the analyzed tasks, and adjusting the execution order of the tasks according to the priorities.

[0048] Specifically, the urgency of the analysis task can be determined based on the task type and the task requirements set by the user. For example, tasks involving real-time data processing have a high priority, while batch analysis tasks that are executed regularly have a low priority. Business needs can be set based on different application scenarios. For example, corporate marketing tasks may have a higher priority than general text summary tasks. User levels can be divided according to the user's authority level or paid service level. For example, tasks of premium member users can be executed first, while tasks of ordinary users may enter the queue. Task priority allocation is stored in the form of priority queues. High-priority tasks are stored in the high-priority queue, and low-priority tasks are stored in the ordinary queue. When tasks are executed, the system extracts tasks from the queue in order of priority to obtain the task execution order adjusted according to priority.

[0049] S202: sorting the analysis tasks according to their creation time, and adjusting the execution order of the tasks using a first-in-first-out control strategy.

[0050] Specifically, when the tasks have the same priority, they are sorted by their creation time. The sorting method can use a timestamp sorting algorithm to record the timestamp when the task was created and arrange the tasks in the order of the timestamps. The task created first is executed first. When the task is executed, a first-in-first-out (FIFO) control strategy is adopted, that is, the tasks are executed in the order in which they enter the queue to ensure the fairness of task processing and avoid the unwarranted resource preemption of tasks submitted later. If the task fails to execute due to system abnormality or user cancellation, the task status can be recorded and re-sorted when the task is restarted or resubmitted to obtain the task execution order arranged in chronological order.

[0051] S203: Based on the usage of currently available computing resources, the analysis task is scheduled for execution when resources are available, and the execution is delayed if resources are not available, and the execution order of the tasks is dynamically adjusted according to the load balancing principle.

[0052] Specifically, the availability of computing resources can be judged by monitoring CPU usage, memory usage, disk I / O and task concurrency. If the computing resources are idle, the task will be executed immediately. If the computing resources are close to the upper limit, the task will be temporarily stored and entered into the resource waiting queue, waiting for the resources to be released before execution. When load balancing adjusts the order of task execution, it can be based on the dynamic load monitoring algorithm to adjust the task allocation according to the current task execution status and the load of the computing node. For example, when the load of a computing node is high, subsequent tasks can be assigned to nodes with lower load to avoid overloading of computing resources. At the same time, task slicing can be used to split large tasks into multiple small tasks and assign them to different computing nodes to improve task execution efficiency, so as to obtain the task execution order after dynamic adjustment of computing resources.

[0053] S204: When there is a dependency relationship between the analysis task and other tasks, parse the dependency relationship based on the task dependency tree, and schedule the execution of the analysis task after the dependency conditions are met.

[0054] Specifically, the task dependency relationship can be represented by a task dependency graph (DAG). Each task node records its predecessor tasks and successor tasks. When parsing the tasks, first traverse the task dependency tree to determine whether there are unfinished predecessor tasks for the current task. If the predecessor tasks are still being executed, temporarily store the current task in the dependency waiting queue. After the predecessor tasks are completed, notify the dependent tasks to enter the execution state. When executing the tasks, a batch execution mode can be adopted. That is, when multiple tasks share the same predecessor task, only trigger all dependent tasks after the predecessor task is completed to improve the task execution efficiency and reduce the repeated execution of tasks. If a dependent task fails, a rollback strategy can be triggered to reschedule the execution of the predecessor tasks to ensure that the dependent tasks are completed under the premise of meeting the dependency relationship and obtain the correct task execution order.

[0055] In one embodiment, as Figure 4 shown, in step S40, that is, extract the semantic features and syntactic structure of the user input information, identify the key information in the text, and match the corresponding analysis task type according to the key information, including: S401: Perform text preprocessing on the user input information to obtain a standardized text.

[0056] Specifically, the text preprocessing can include steps such as character normalization, stop word filtering, punctuation processing, case conversion, and text tokenization. Character normalization can convert special characters (such as full-width characters, abnormal spaces) in the input text into a standard format. Stop word filtering can remove common meaningless words (such as "de", "shi", "he", etc.). Punctuation processing can be used to delete redundant punctuation or standardize the punctuation form. Case conversion can convert all characters into a unified format to ensure text consistency. Text tokenization can adopt a rule-based tokenization method or a statistical tokenization model to segment the text according to language habits to make it have a good grammatical structure. If the text contains spelling mistakes or non-standard words, it can be further corrected through a spelling correction algorithm (such as the edit distance algorithm) to ensure that the text format is unified and standardized to obtain a standardized text.

[0057] S402: Based on the standardized text, use a pre-trained language model to extract the grammatical structure and semantic features, parse the sentence components, identify the syntactic relationship of the text, and identify the key information in the standardized text.

[0058] Specifically, grammatical structure extraction can use dependency analysis or component analysis to decompose the text into components such as subject, predicate, object, and modifiers. Semantic feature extraction can use pre-trained language models (such as BERT and GPT) to calculate the word vector representation of the text to capture the contextual information of the sentence. When parsing sentence components, the POS (part-of-speech tagging) algorithm can be used to annotate the part of speech of each word, such as noun, verb, adjective, etc. When identifying the syntactic relationship of the text, dependency parsing can be used to construct a syntactic dependency tree to represent the grammatical relationship between words. The identification of key information can be combined with TF-IDF, named entity recognition (NER) or attention mechanism to extract core vocabulary with important meaning from the text, such as time, place, person, product name, etc., to obtain a parsing result that contains grammatical structure, semantic features and key information.

[0059] S403: Match the corresponding analysis task type according to the key information.

[0060] Specifically, task type matching can be based on keyword matching, pattern recognition, classification models, etc. Keyword matching can be performed by querying a predefined task keyword library to find whether the text contains key phrases related to a certain task. Pattern recognition can be based on text rule matching (such as regular expressions) or text structure analysis (such as sentence analysis) to detect whether the input text conforms to a known task pattern. The classification model can use machine learning or deep learning models, such as support vector machines (SVM), random forests, or Transformer-based text classification models to classify and predict the input text and determine its most likely task type. When the matching degree of multiple task types is similar, a confidence scoring mechanism can be used to calculate the matching score of each task type, and the type with the highest score is selected as the final matching result to obtain the analysis task type that best matches the input content.

[0061] In one embodiment, if Figure 5 As shown, in step S50, based on the analysis task type, the corresponding analysis process is called to obtain the analysis calculation results, which include the optimized copy, sentiment analysis results and translated text, including: S501: According to the analysis task type, it is identified that the analysis process belongs to a type of copywriting generation, comment analysis or text translation.

[0062] Specifically, the identification of the analysis process can be based on task classification models, keyword matching, rule mapping and other methods. The task classification model can use text classification algorithms (such as Naive Bayes, SVM, Transformer pre-training model) to classify user input text and predict its most suitable analysis process. Keyword matching can be based on a predefined task keyword table to extract key phrases in the text. If the input contains words such as "generate copy" and "content optimization", it is matched as a copy generation task. If it contains "sentiment analysis" and "opinion recognition", it is matched as a comment analysis task. If it contains "translation" and "multi-language conversion", it is matched as a text translation task. Rule mapping can be based on regular expressions or task templates to determine whether the input text meets a certain task pattern, and determine the analysis process based on the degree of matching to obtain the correct analysis process category. S502: When the analysis process corresponding to the analysis task type is copy generation, based on the keywords, prompts and negative words of the user input information, according to the preset copy generation rules, a copy that conforms to the grammatical specifications is constructed, and the syntactic structure and vocabulary selection of the copy are adjusted to obtain an optimized copy.

[0063] Specifically, copywriting generation can be based on methods such as keyword filling, template generation, and deep learning text generation. The keyword filling method can use a rule-based text template to embed the keywords entered by the user into the preset template. For example, if the user enters "smartphone", "high performance", and "cost-effectiveness" as keywords, the copywriting "This smartphone has a high-performance processor, extremely high cost-effectiveness, and is suitable for all types of users" can be generated. The template generation method can be based on pattern matching and combined with prompts to construct complete sentences. If the user enters "smart home product introduction" as the prompt, a suitable template can be selected to generate the corresponding introductory copywriting. The deep learning text generation method can use a neural network model and a language model based on large-scale text data training to generate fluent copywriting that conforms to the context logic. During the generation process, the syntactic structure is adjusted, the vocabulary selection is optimized, and repeated expressions or lengthy sentences are avoided to obtain optimized copywriting that meets user needs. S503: When the analysis process corresponding to the analysis task type is comment analysis, parse the user input information, use the text sentiment analysis model to identify the text sentiment tendency, and calculate the proportion of different sentiment categories to obtain the sentiment analysis results.

[0064] Specifically, comment analysis can be carried out based on steps such as text preprocessing, sentiment classification, and sentiment intensity calculation. Text preprocessing may include removing stop words, punctuation processing, and text normalization to ensure that the input text format is unified. Sentiment classification can adopt a dictionary-based method or a deep learning sentiment analysis model. The dictionary-based method can use a predefined sentiment vocabulary to match the sentiment polarity of words in the text. For example, "good" and "like" can be judged as positive emotions, and "disappointment" and "bad" can be judged as negative emotions. The deep learning sentiment analysis model can be based on natural language processing models such as BERT and LSTM. Through large-scale annotated data training, it can predict the sentiment tendency of the text with higher accuracy. The sentiment intensity calculation can be based on sentiment score weighting, statistically analyzing the weights of positive and negative sentiment words in the text, and calculating the proportion of different sentiment categories to obtain sentiment analysis results.

[0065] S504: When the analysis process corresponding to the analysis task type is text translation, the user input information is translated through the translation model according to the target language set by the user, the term consistency is adjusted in combination with the context information, and the syntactic structure is optimized to obtain a translation text that conforms to the expression specifications of the target language.

[0066] Specifically, text translation can be performed based on technologies such as neural machine translation (NMT), term matching, and grammar adjustment. Neural machine translation can use deep learning translation models such as Transformer, Seq2Seq, and BERT to perform end-to-end translation of the input text, parse the contextual information of the text, and generate a translation that conforms to the target language. Term matching can be based on a term base or translation memory to ensure that professional terminology in a specific field remains consistent in the translation. For example, in the translation of technical documents, "artificial intelligence" should always be translated as "Artificial Intelligence" rather than other synonyms. Grammar adjustment can be based on syntactic analysis to perform secondary optimization of the translation to ensure that the grammar conforms to the expression standards of the target language. For example, in Chinese-English translation, the subject-verb-object structure can be adjusted to avoid grammatical errors caused by literal translation, so as to obtain a translated text that conforms to the standards of the target language.

[0067] In one embodiment, if Figure 6 As shown, after step S60, based on the analysis task type, the corresponding analysis process is called to obtain the analysis calculation results, which include the optimized copy, sentiment analysis results and translated text, and then also include: S601: When the analysis process corresponding to the analysis task type is user feedback optimization, obtain user interaction behavior information on the analysis calculation results.

[0068] Specifically, user interaction behavior information may include data such as likes, modifications, clicks, dwell time, feedback submission, etc. Likes may refer to users' recognition of the generated copy, translated text or sentiment analysis results. Modifications may record users' manual editing or replacement operations on the output content. Clicks can be used to track users' attention to different options or content recommendations. Dwell time can measure the depth of users' reading of the results. Feedback submissions may include suggestions, ratings or complaints actively filled out by users. When obtaining interaction behavior information, it can be collected through front-end event monitoring, log collection, database records, etc. Front-end event monitoring can capture users' clicks, scrolling, dwell time and other behaviors. Log collection can record users' access paths and operation tracks. Database records can store the timestamp, user ID and corresponding task ID of each interaction behavior to ensure that subsequent optimization can be adjusted based on real user behavior data to obtain users' interaction behavior information on analysis and calculation results.

[0069] S602: According to the interactive behavior information, adjust the strategy parameters of the analysis process to obtain the adjusted strategy parameters, and based on the adjusted strategy parameters, update the preset copy generation rules, the parameters of the text sentiment analysis model, or the parameters of the translation model.

[0070] Specifically, strategy parameter adjustments can be performed based on user modification behavior analysis, preference modeling, and model adaptive adjustment. User modification behavior analysis can count the frequency, content, and method of user modifications to copywriting, translated text, or sentiment analysis results. For example, when a user frequently replaces certain vocabulary or sentence structures, the copywriting generation rules can be adjusted to make subsequently generated copywriting more in line with user preferences. Preference modeling can analyze user usage habits based on historical interaction data and build a user-personalized model. For example, if a user has a long-term preference for a formal copywriting style, the generation strategy can be adjusted to make subsequent copywriting more in line with formal writing standards. Model adaptive adjustment can dynamically adjust the classification threshold of the sentiment analysis model, the term matching method of the translation model, and the sentence selection strategy for copywriting generation through a feedback-driven parameter optimization mechanism. If users report that certain sentiment classification results are inaccurate, the classification boundaries of the sentiment model are adjusted to improve the recognition accuracy of specific categories. If users repeatedly correct certain term translations in the translation results, the term matching rules are updated to ensure that the user-approved translation method is consistently used in subsequent translations to obtain optimized strategy parameters. Based on this, the copywriting generation rules, text sentiment analysis model parameters, or translation model parameters are updated.

[0071] In one embodiment, if Figure 7 As shown, in step S602, the strategy parameters of the analysis process are adjusted according to the interactive behavior information to obtain the adjusted strategy parameters. Based on the adjusted strategy parameters, the preset copy generation rules, the parameters of the text sentiment analysis model, or the parameters of the translation model are updated, including: S6021: Analyze the user's modification frequency, content adjustment method, and vocabulary replacement record of the generated copy based on the interactive behavior information, obtain adjusted copy generation strategy parameters, and update the preset copy generation rules based on the adjusted strategy parameters.

[0072] Specifically, the optimization of copy generation strategy parameters can be based on the analysis of users' editing behavior of generated copy. The modification frequency can be determined by counting the number of times users edit the same copy. If users frequently modify a certain type of copy, the generation rules of the copy may need to be adjusted. The content adjustment method can be based on the text comparison algorithm to analyze the changes in the copy before and after the user's modification, and determine whether the user tends to adjust the sentence structure, tone, grammar or wording. For example, if the user often splits long sentences into short sentences, the generation rules can be adjusted so that the subsequently generated copy uses shorter sentences. Vocabulary replacement records can be used to identify user preferred words. By constructing a user vocabulary replacement dictionary, the user's commonly used expressions in the copy are analyzed and applied to subsequent copy generation rules. For example, if the user frequently replaces "efficient" with "fast", the word "fast" can be given priority in the generation process. The updated copy generation rules are stored in the strategy database and loaded when a new task is executed to obtain a copy generation strategy that is more in line with user habits.

[0073] S6022: Optimize the classification threshold, feature extraction method, and sentiment dictionary of the text sentiment analysis model based on the interactive behavior information, obtain adjusted sentiment analysis strategy parameters, and update the parameters of the text sentiment analysis model based on the adjusted strategy parameters.

[0074] Specifically, sentiment analysis strategy parameters can be optimized based on user feedback on analysis results. Classification thresholds can be optimized by collecting user approval data on sentiment classification results, calculating the correct / incorrect feedback ratio, and dynamically adjusting the confidence interval of the classification. For example, if a user believes that some neutral sentiment text is incorrectly classified as positive, the threshold for positive sentiment classification can be lowered, increasing the judgment range of the neutral category. Feature extraction methods can be optimized based on user mislabeling in sentiment analysis tasks by weighting features of specific text categories. For example, if users frequently modify the sentiment classification of sarcastic text in comment analysis tasks, feature weights can be adjusted to more accurately identify sarcastic text. Sentiment lexicon optimization can accumulate new words based on user feedback data and assign new sentiment polarity to existing sentiment words. For example, if the word "stimulation" is more often used to express positive sentiment in a certain industry context, but the original lexicon defines it as negative, its sentiment category can be adjusted based on user feedback. The updated sentiment analysis model parameters are stored in the database and applied to subsequent sentiment analysis tasks to obtain more accurate sentiment analysis results.

[0075] S6013: Optimize term matching, syntactic adjustment, and context relevance of the translation model based on the interaction behavior information, update the terminology database, obtain adjusted translation strategy parameters, and update the parameters of the translation model based on the adjusted strategy parameters.

[0076] Specifically, the optimization of translation strategy parameters can be adjusted based on the user's modification records and terminology preferences for the translated text. The optimization of terminology matching can determine the translation method that the user prefers to use by counting the frequency of manual corrections made by users to specific terms. For example, if a user repeatedly rewrites "artificial intelligence" as "AI" in the translation of scientific and technological documents, "AI" can be preferentially stored in the terminology library as the default translation method. The optimization of syntactic adjustment can be based on grammatical structure analysis to identify the user's sentence adjustment trend for the translation results. For example, if the user tends to change the passive voice to the active voice, the translation model can be adjusted to give priority to the active voice expression in subsequent translations. The optimization of context relevance can ensure the consistency of terminology and expression style throughout the text by analyzing the user's adjustment pattern for paragraph translation. For example, if a user modifies the translation of a term multiple times in the same document, the global terminology matching rules can be updated to prevent different translation versions from appearing in subsequent paragraphs. The updated translation model parameters are stored in the database and applied to subsequent translation tasks to obtain more accurate translation results that conform to user habits.

[0077] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0078] In one embodiment, an intelligent document analysis device is provided, which corresponds to the intelligent document analysis method in the above embodiment. Figure 8 As shown, the intelligent document analysis device includes a task creation and initialization module, a task scheduling and execution module, a text parsing and feature extraction module, a task type matching and process call module, an analysis calculation and result generation module, and a result storage and task status management module. The functional modules are described in detail below: The task creation and initialization module is used to obtain user input information, generate analysis tasks based on the user input information, assign task identifiers to the analysis tasks, and store the analysis tasks in the analysis storage database. At the same time, the status of the analysis tasks is set to the initial state, which includes pending or executing; The task scheduling and execution module is used to determine the execution time of the analysis task based on the preset task scheduling strategy when the status of the analysis task is pending, and update the status of the analysis task to executing; The text parsing and feature extraction module is used to read the corresponding user input information from the analysis storage database according to the task identifier when the status of the analysis task is in execution; The task type matching and process calling module is used to extract the semantic features and syntactic structure of user input information, identify key information in the text, and match the corresponding analysis task type based on the key information; The analysis calculation and result generation module is used to call the corresponding analysis process based on the analysis task type and obtain the analysis calculation results, which include optimized copywriting, sentiment analysis results, and translated text; The result storage and task status management module is used to store the analysis and calculation results in the analysis storage database and update the status of the analysis task, which includes executing, completed and failed.

[0079] Optionally, the task creation and initialization module includes: The input verification and error prompt submodule is used to check the integrity and format of user input information using preset input verification rules. If the user input information is missing parameters, an error prompt message is generated; The task type analysis submodule is used to analyze the type of user input information and obtain an analysis task if the user input information meets the preset input verification rules; The task identifier allocation and storage submodule is used to allocate a task identifier to the analysis task through a unique identifier, and store the analysis task and the task identifier in the analysis storage database; The task status initialization submodule is used to set the initial status to pending or executing according to the execution mode of the analysis task. If the analysis task needs to queue for scheduling, the initial status is set to pending. If the analysis task does not need to queue and can be executed directly, the initial status is set to executing.

[0080] Optionally, the task scheduling and execution module includes: The task priority allocation submodule is used to allocate the execution priority of tasks according to the urgency of the analyzed tasks, business requirements and user levels, and adjust the execution order of tasks according to the priority level; The task sorting and scheduling submodule is used to sort the analysis tasks according to their creation time and adjust the execution order of the tasks using a first-in-first-out control strategy; The computing resource scheduling submodule is used to schedule the execution of analysis tasks when resources are available based on the current usage of available computing resources. If resources are not available, the execution is delayed and the execution order of tasks is dynamically adjusted according to the load balancing principle. The task dependency management submodule is used to resolve the dependency relationship based on the task dependency tree when the analysis task has a dependency relationship with other tasks, and arrange the execution of the analysis task after the dependency conditions are met.

[0081] Optionally, the task type matching and process calling modules include: The text preprocessing submodule is used to perform text preprocessing on user input information to obtain standardized text; The grammatical and semantic feature extraction submodule is used to extract grammatical structure and semantic features based on standardized text using a pre-trained language model, parse sentence components, identify syntactic relationships in the text, and identify key information in the standardized text; The task type matching submodule is used to match the corresponding analysis task type based on key information.

[0082] Optional analysis, calculation and result generation modules include: The analysis process identification submodule is used to identify the analysis process as copywriting generation, comment analysis or text translation based on the analysis task type; The copy generation processing submodule is used to construct a grammatically correct copy based on the keywords, prompts, and negation words in the user input information according to the preset copy generation rules when the analysis process corresponding to the analysis task type is copy generation. It also adjusts the syntactic structure and vocabulary selection of the copy to obtain an optimized copy. The sentiment analysis processing submodule is used to parse user input information when the analysis process corresponding to the analysis task type is comment analysis, identify the text sentiment tendency using the text sentiment analysis model, and calculate the proportion of different sentiment categories to obtain the sentiment analysis results; The text translation processing submodule is used to translate the user input information through the translation model according to the target language set by the user when the analysis process corresponding to the analysis task type is text translation, adjust the terminology consistency based on the context information, and optimize the syntactic structure to obtain a translated text that conforms to the expression specifications of the target language.

[0083] Optionally, the result storage and task status management module includes: The user interaction data acquisition module is used to obtain the user's interactive behavior information on the analysis calculation results when the analysis process corresponding to the analysis task type is user feedback optimization; The analysis process optimization module is used to adjust the strategy parameters of the analysis process according to the interactive behavior information, obtain the adjusted strategy parameters, and update the preset copy generation rules, the parameters of the text sentiment analysis model or the parameters of the translation model based on the adjusted strategy parameters.

[0084] Optional analysis process optimization modules include: The copy optimization strategy adjustment sub-block is used to analyze the user's modification frequency, content adjustment method and vocabulary replacement records of the generated copy based on the interaction behavior information, obtain the adjusted copy generation strategy parameters, and update the preset copy generation rules based on the adjusted strategy parameters; The sentiment analysis model optimization submodule is used to optimize the classification threshold, feature extraction method and sentiment dictionary of the text sentiment analysis model based on the interactive behavior information, obtain the adjusted sentiment analysis strategy parameters, and update the parameters of the text sentiment analysis model based on the adjusted strategy parameters; The translation optimization strategy adjustment submodule is used to optimize the term matching, syntactic adjustment and context relevance of the translation model based on the interaction behavior information, update the terminology library, obtain the adjusted translation strategy parameters, and update the parameters of the translation model based on the adjusted strategy parameters.

[0085] For the specific definition of the intelligent document analysis device, please refer to the definition of the intelligent document analysis method above, and will not be repeated here. The various modules in the above-mentioned intelligent document analysis device can be implemented in whole or in part by software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0086] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 9 As shown. The computer device includes a processor, memory, network interface and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to analyze and store the database. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, an intelligent document analysis method is implemented.

[0087] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed: Obtain user input information, generate an analysis task based on the user input information, assign a task identifier to the analysis task, and store the analysis task in an analysis storage database. At the same time, the status of the analysis task is set to an initial state, which may be pending or in progress. When the status of the analysis task is pending, the execution time of the analysis task is determined based on the preset task scheduling strategy, and the status of the analysis task is updated to executing; When the status of the analysis task is in execution, the corresponding user input information is read from the analysis storage database according to the task identifier; Extract the semantic features and syntactic structure of user input information, identify key information in the text, and match the corresponding analysis task type based on the key information; Based on the analysis task type, the corresponding analysis process is called to obtain the analysis and calculation results, which include the optimized copy, sentiment analysis results, and translated text; The analysis and calculation results are stored in the analysis storage database, and the status of the analysis task is updated, including the status of executing, completed and failed.

[0088] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: Obtain user input information, generate an analysis task based on the user input information, assign a task identifier to the analysis task, and store the analysis task in an analysis storage database. At the same time, the status of the analysis task is set to an initial state, which may be pending or in progress. When the status of the analysis task is pending, the execution time of the analysis task is determined based on the preset task scheduling strategy, and the status of the analysis task is updated to executing; When the status of the analysis task is in execution, the corresponding user input information is read from the analysis storage database according to the task identifier; Extract the semantic features and syntactic structure of user input information, identify key information in the text, and match the corresponding analysis task type based on the key information; Based on the analysis task type, the corresponding analysis process is called to obtain the analysis and calculation results, which include the optimized copy, sentiment analysis results, and translated text; The analysis and calculation results are stored in the analysis storage database, and the status of the analysis task is updated, including the status of executing, completed and failed.

[0089] Those skilled in the art will 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. 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. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0090] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0091] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. An intelligent copywriting analysis method, characterized in that: The intelligent copywriting analysis method includes: Obtaining user input information, generating an analysis task based on the user input information, assigning a task identifier to the analysis task, storing the analysis task in an analysis storage database, and setting the state of the analysis task to an initial state, wherein the initial state includes pending execution or executing; When the status of the analysis task is "to be executed", based on a preset task scheduling strategy, determining the execution timing of the analysis task, and updating the status of the analysis task to "in execution"; When the status of the analysis task is in execution, reading corresponding user input information from the analysis storage database according to the task identifier; Extracting semantic features and syntactic structures of the user input information, identifying key information in the text, and matching corresponding analysis task types according to the key information; Based on the analysis task type, the corresponding analysis process is called to obtain analysis and calculation results, which include optimized copywriting, sentiment analysis results, and translated text; The analysis calculation results are stored in the analysis storage database, and the status of the analysis task is updated, where the status includes being executed, completed, and failed.

2. The intelligent copywriting analysis method according to claim 1, characterized in that: The step of obtaining user input information, generating an analysis task based on the user input information, assigning a task identifier to the analysis task, storing the analysis task in an analysis storage database, and setting the state of the analysis task to an initial state, wherein the initial state includes pending execution or executing, includes: Use preset input validation rules to check the integrity and format of the user input information, and generate an error message if the user input information is missing parameters; If the user input information meets the preset input verification rule, analyzing the type of the user input information to obtain the analysis task; Assigning the task identifier to the analysis task through a unique identifier, and storing the analysis task and the task identifier in the analysis storage database; According to the execution mode of the analysis task, the initial state is set to the pending execution or the executing state. If the analysis task needs to queue for scheduling, the initial state is set to the pending execution. If the analysis task does not need to queue and can be executed directly, the initial state is set to the executing state.

3. The intelligent copywriting analysis method according to claim 1, characterized in that: The preset task scheduling strategy includes: Assigning execution priorities to tasks based on the urgency, business requirements, and user levels of the analysis tasks, and adjusting the execution order of tasks according to their priorities; Sort the analysis tasks according to their creation time and use a first-in-first-out control strategy to adjust the execution order of the tasks; Based on the current usage of available computing resources, the analysis tasks are scheduled for execution when resources are available, and delayed if resources are not available, and the execution order of tasks is dynamically adjusted according to the load balancing principle; When the analysis task has a dependency relationship with other tasks, the dependency relationship is parsed based on the task dependency tree, and the execution of the analysis task is arranged after the dependency conditions are met.

4. The intelligent copywriting analysis method according to claim 1, characterized in that: The extracting of semantic features and syntactic structures of the user input information, identifying key information in the text, and matching corresponding analysis task types according to the key information includes: Performing text preprocessing on the user input information to obtain the standardized text; Based on the standardized text, using a pre-trained language model to extract grammatical structure and semantic features, parse sentence components, identify syntactic relationships in the text, and identify the key information in the standardized text; According to the key information, the corresponding analysis task type is matched.

5. The intelligent copywriting analysis method according to claim 1, characterized in that: Based on the analysis task type, the corresponding analysis process is called to obtain analysis and calculation results, which include optimized copywriting, sentiment analysis results and translated text, including: According to the analysis task type, identifying whether the analysis process is a copywriting generation, comment analysis, or text translation process; When the analysis process corresponding to the analysis task type is the copywriting generation, based on the keywords, prompts and negation words in the user input information, according to the preset copywriting generation rules, a copywriting that conforms to grammatical specifications is constructed, and the syntactic structure and vocabulary selection of the copywriting are adjusted to obtain the optimized copywriting; When the analysis process corresponding to the analysis task type is the comment analysis, parsing the user input information, using a text sentiment analysis model to identify the sentiment tendency of the text, and calculating the proportion of different sentiment categories to obtain the sentiment analysis result; When the analysis process corresponding to the analysis task type is text translation, the user input information is translated through a translation model based on the target language set by the user, the terminology consistency is adjusted in combination with the context information, and the syntactic structure is optimized to obtain the translated text that conforms to the expression specifications of the target language.

6. The intelligent copywriting analysis method according to claim 1, characterized in that: Based on the analysis task type, the corresponding analysis process is called to obtain analysis and calculation results, which include optimized copywriting, sentiment analysis results and translated text, and then further include: When the analysis process corresponding to the analysis task type is user feedback optimization, obtaining the user's interactive behavior information on the analysis calculation result; According to the interactive behavior information, the strategy parameters of the analysis process are adjusted to obtain adjusted strategy parameters; based on the adjusted strategy parameters, the preset copy generation rules, the parameters of the text sentiment analysis model or the parameters of the translation model are updated.

7. The intelligent copywriting analysis method according to claim 6, characterized in that: The step of adjusting the strategy parameters of the analysis process according to the interaction behavior information to obtain the adjusted strategy parameters, and updating the preset copywriting generation rules, the parameters of the text sentiment analysis model, or the parameters of the translation model based on the adjusted strategy parameters, includes: Analyzing the user's modification frequency, content adjustment method, and vocabulary replacement record of the generated copy based on the interaction behavior information to obtain adjusted copy generation strategy parameters, and updating the preset copy generation rules based on the adjusted strategy parameters; Optimizing the classification threshold, feature extraction method, and sentiment dictionary of the text sentiment analysis model according to the interactive behavior information to obtain adjusted sentiment analysis strategy parameters, and updating the parameters of the text sentiment analysis model based on the adjusted strategy parameters; According to the interactive behavior information, term matching, syntax adjustment and context relevance of the translation model are optimized, a term library is updated, adjusted translation strategy parameters are obtained, and parameters of the translation model are updated based on the adjusted strategy parameters.

8. An intelligent document analysis device, characterized in that: The intelligent document analysis device includes: A task creation and initialization module is used to obtain user input information, generate an analysis task based on the user input information, assign a task identifier to the analysis task, and store the analysis task in an analysis storage database, while setting the status of the analysis task to an initial state, wherein the initial state includes pending execution or executing; A task scheduling and execution module is used to determine the execution time of the analysis task based on a preset task scheduling strategy when the status of the analysis task is the pending execution, and update the status of the analysis task to the executing state; a text parsing and feature extraction module, configured to read corresponding user input information from the analysis storage database according to the task identifier when the status of the analysis task is in execution; The task type matching and process calling module is used to extract the semantic features and syntactic structure of the user input information, identify the key information in the text, and match the corresponding analysis task type according to the key information; An analysis calculation and result generation module is used to call the corresponding analysis process based on the analysis task type to obtain analysis calculation results, which include optimized copy, sentiment analysis results and translated text; The result storage and task status management module is used to store the analysis and calculation results in the analysis storage database and update the status of the analysis task, which includes the status of executing, completed and failed.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the intelligent copywriting analysis method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent copywriting analysis method according to any one of claims 1 to 7 are implemented.

Images