File identification method and device based on semantic recognition of natural language processing technology

Abstract

The embodiment of the invention provides an archive identification method and device based on semantic recognition of a natural language processing technology, and relates to the technical field of archive management, and the method comprises the steps: carrying out the preliminary metadata extraction and storage of an inputted archive; semantic analysis is conducted on the archive, semantic elements in the archive are recognized, and the semantic elements comprise entities, relations and events; setting an authentication rule, automatically authenticating the archive according to the preset authentication rule, and allocating an authentication task; setting a process model, receiving real-time operation, and automatically updating and displaying the state of the file in the corresponding process node; generating a detailed identification report and related statistical information according to the identification result, and storing all data of the whole file processing process; semantic information of archive content is deeply understood, so that the identification result is accurate and comprehensive, and the requirements of modern archive management for efficient and accurate identification are met.

Description

Technical Field

[0001] The present application relates to the technical field of archive management, and in particular to an archive identification method and device based on semantic recognition of natural language processing technology. Background Art

[0002] Traditional archives appraisal work faces many challenges, such as low efficiency, strong subjectivity, and difficulty in standardization of manual appraisal. With the development of information technology, although there are some archives management systems, there is still a lack of intelligent, automated and high-precision solutions for archives appraisal. Existing archives appraisal systems often cannot deeply understand the semantic information of archives content, resulting in inaccurate and incomplete appraisal results, which is difficult to meet the needs of modern archives management for efficient and accurate appraisal. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method and device for archive identification based on semantic recognition of natural language processing technology, so as to solve the problem that the current archive identification methods are often unable to deeply understand the semantic information of the archive content, resulting in the identification results being inaccurate and incomplete, and unable to meet the needs of modern archive management for efficient and accurate identification.

[0004] In a first aspect, the present application provides a method for file identification based on semantic recognition of natural language processing technology, including: Extract and store preliminary metadata of the entered archives; Perform semantic analysis on the archives to identify semantic elements in the archives, where the semantic elements include entities, relations and events; Set up identification rules, automatically identify files according to the preset identification rules, and assign identification tasks; Set up the process model and receive real-time operations, automatically update the status of the archive at the corresponding process node and display it; Generate detailed identification reports and relevant statistical information based on the identification results, and store all data of the entire archive processing process.

[0005] In the above implementation process, preliminary metadata of the entered archives is extracted and stored; semantic analysis is performed on the archives to identify semantic elements in the archives, where the semantic elements include entities, relationships and events; identification rules are set, and archives are automatically identified and identification tasks are assigned according to the preset identification rules; a process model is set, and real-time operations are received to automatically update the status of the archives at the corresponding process nodes and display them; a detailed identification report and related statistical information are generated based on the identification results, and all data of the entire archive processing process is stored; the semantic information of the archive content is deeply understood, so that the identification results are accurate and comprehensive, meeting the needs of modern archive management for efficient and accurate identification.

[0006] Furthermore, the preliminary metadata extraction and storage of the input archives includes: Input files, including paper files and electronic document files; Convert the entered files into a unified format; After preliminary metadata extraction of the entered archives, they are stored in the database, where the metadata includes the archive type, creation time and source.

[0007] In the above implementation process, the input archival data will be converted into a unified format and metadata will be extracted to facilitate subsequent processing.

[0008] Furthermore, the semantic analysis of the archive is performed to identify semantic elements in the archive, wherein the semantic elements include entities, relationships and events, including: Construct a dictionary containing various entity types and process the archive to match words in the archive text with the dictionary and words with entities of the corresponding type; Determine the location and grammatical structure of the entity according to the grammatical rules of the language; Extracting features for entity recognition from the text, wherein the features include: features of the word itself, context features of the word, and part-of-speech features; Using the labeled entity training data, the supervised learning model is used to train the entity recognition model, and the trained entity recognition model is used to predict the entity and type of the newly input archive text; Using dependency syntactic analysis tools, a syntactic tree is constructed for the sentence to show the dependency relationship between words. According to the dependency relationship type and path in the syntactic tree, rules are formulated to extract the relationship between entities. By training the relation recognition model on texts with annotated semantic roles, the relation recognition model learns to recognize the semantic role played by each component in the sentence; According to the results of semantic role labeling, establish the relationship between entities based on semantic roles; Pre-build a knowledge graph, where the knowledge graph contains various entities and relationship types between entities; When processing archival text, the identified entities are matched with the entities in the knowledge graph. When the entities in the text are successfully matched with the entities in the knowledge graph, the relationships in the knowledge graph are used to infer the relationships between the entities in the text and to supplement the missing relationship information. Determine the keywords used to trigger the event. If the keywords are detected in the archive text, it is preliminarily determined that an event exists; According to different types of events, event templates are constructed, and the content surrounding the trigger words in the text is matched with the event template to identify complete event information; Using the theory of frame semantics, each event is mapped to a semantic frame with different slots. The semantic frame corresponding to the event is identified in the archival text and the information of each slot is filled. Use the pre-trained word vector model to convert words in the text into vector representations, where the word vector includes the semantic information of the words and the semantic relationship between the words is represented by vector operations; The context of the text is modeled using a set network model to capture the order information and contextual dependencies of words in a sentence; Analyze the chapter structure of the text and complete event identification.

[0009] In the above implementation process, it is possible to identify semantic elements such as entities, relationships, events, etc. in the text, and understand the overall semantics and context of the text, which is conducive to subsequent archive identification.

[0010] Furthermore, the setting of identification rules, automatically identifying files according to preset identification rules, and assigning identification tasks include: Set up rules for assigning identification tasks; When evaluating a new file text, the relevant information of the file is automatically read, and according to the evaluation task assignment rules, qualified personnel are searched in the evaluation personnel database to assign the evaluation task; Collect the competency data of the appraisers and establish a competency model for each appraiser, where the competency data includes the completion quality and efficiency of past appraisal tasks, professional skills, and experience level; When a new appraisal task is received, the characteristics of the task are matched with the competency model of the appraiser, a machine learning algorithm is used to calculate the matching score between the task and each appraiser, and the appraiser is selected based on the score.

[0011] In the above implementation process, the rules cover multiple aspects such as the importance, timeliness, and integrity of the archives. At the same time, the rule base is extensible and maintainable.

[0012] Furthermore, the process model is set, and real-time operations are received, and the status of the archive at the corresponding process node is automatically updated and displayed, including: Predefine the process model of archive appraisal, where the process model includes specific operations and requirements at each stage, and sets different states for each process node; After receiving the real-time operation on the file, the status of the file in the corresponding process node is automatically updated; The real-time status of the archives identification process is displayed through a graphical interface; According to business experience and process requirements, abnormal situation rules are set, and various data in the archive identification process are monitored in real time. When situations that meet the abnormal rules are found, the early warning mechanism is automatically triggered.

[0013] In the above implementation process, during the identification process, the progress of each stage is monitored in real time to ensure that the identification work proceeds smoothly according to the process; at the same time, it facilitates communication and information sharing between identification personnel to improve identification efficiency and accuracy.

[0014] Furthermore, after the process model is set, and real-time operations are received, and the status of the archive at the corresponding process node is automatically updated and displayed, the following is further included: When receiving real-time operations on archives, automatically record the timestamp of the operations; Based on the timestamp data, calculate the residence time of each file at each process node and the overall identification progress, and display the progress information in the form of a dashboard or report; Among them, there are associations and dependencies between multiple tasks in the archives appraisal process; When the status of a task changes, the progress and status of related dependent tasks are automatically updated based on the task association relationship.

[0015] During the above implementation process, the progress of each stage is monitored in real time to ensure that the identification work proceeds smoothly according to the process.

[0016] Furthermore, a detailed identification report and relevant statistical information are generated based on the identification results, and all data of the entire archive processing process are stored, including: Obtain basic data of archives; Determine the final conclusion of the appraisal based on the appraisal results and comprehensive judgment of each stage of the appraisal process, and write the conclusion into the report; Query the rules related to the file identification process from the identification rule library; If natural language processing technology is used to conduct semantic analysis during the identification process, the results of the semantic analysis shall be included in the basis for identification; According to the appraisal conclusion, put forward corresponding treatment measures; Query all the files that have been appraised from the database, classify and count them according to the file type, and display the statistical results in the form of tables or charts; Search the database for files that have been appraised, classify and count them according to the appraisal conclusions, and display the statistical results in the form of a pie chart; Stores all data of the entire archive processing process, including original archive data, metadata, NLP semantic analysis results, identification rules, identification process records and identification results.

[0017] In the above implementation process, a detailed appraisal report and related statistical information are generated based on the appraisal results to provide data support for archive management decisions; efficient data storage and retrieval technology is used to ensure the security, integrity and rapid access performance of the data.

[0018] In a second aspect, the embodiment of the present application provides a file identification device for semantic recognition based on natural language processing technology, including: The pre-entry module is used to extract and store preliminary metadata of the entered archives; A semantic analysis module is used to perform semantic analysis on the archives and identify semantic elements in the archives, wherein the semantic elements include entities, relations and events; The identification processing module is used to set identification rules, automatically identify files according to the preset identification rules, and assign identification tasks; The process processing module is used to set the process model and receive real-time operations, automatically update the status of the archive at the corresponding process node and display it; The result generation module is used to generate a detailed identification report and related statistical information based on the identification results, and store all data of the entire archive processing process.

[0019] In a third aspect, an embodiment of the present application provides an electronic device, including: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the archive identification method based on semantic recognition based on natural language processing technology as described above.

[0020] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a server, implements the archive identification method for semantic recognition based on natural language processing technology as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A flowchart of a method for file identification based on semantic recognition of natural language processing technology provided in an embodiment of the present application; Figure 2It is a structural schematic diagram of a file identification device for semantic recognition based on natural language processing technology provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0024] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0025] NLP (Natural Language Processing) technology, that is, natural language processing technology, is an important direction in the fields of computer science and artificial intelligence.

[0026] Please see Figure 1 , Figure 1 A flowchart of a method for file identification based on semantic recognition of natural language processing technology provided in an embodiment of the present application. The method for file identification based on semantic recognition of natural language processing technology includes: 100. Extract and store preliminary metadata of the entered archives.

[0027] It should be noted that the archive identification system based on semantic recognition of natural language processing technology includes an archive entry module, an NLP semantic analysis module, an identification rule base module, an identification process management module, a result output module and a database module.

[0028] 110. Enter files, including paper files and electronic document files.

[0029] Understandably, paper archives need to be scanned into electronic format before being entered, and electronic document archives can be directly imported into the archive entry module.

[0030] 120. Convert the entered files into a unified format.

[0031] Optionally, determine a unified format suitable for all archives, which can be determined based on the type of archive (such as text, photos, audio, video, etc.) and the specific needs within the organization. For example, for text archives, you can choose formats such as layout files, OFD, DOCX, XML, PDF, etc., because these formats have high compatibility and readability.

[0032] 130. Perform preliminary metadata extraction on the entered archives and store them in the database, where the metadata includes the archive type, creation time and source.

[0033] 200. Perform semantic analysis on the archives to identify semantic elements in the archives, where the semantic elements include entities, relationships and events.

[0034] It should be noted that the NLP semantic analysis module is one of the core modules of the system, which uses advanced NLP technology to perform in-depth semantic analysis of the archive content. It can identify semantic elements such as entities, relationships, events, etc. in the text, and understand the overall semantics and context of the text. For example, for a contract file, this module can accurately identify key information such as the parties to the contract, the subject matter of the contract, the signing date, the content of the terms, and analyze the logical relationship and potential risks between the terms. In addition, the module also has a sentiment analysis function, which can judge the emotional tendency expressed in the archive text, which is of great significance for some archive appraisals involving subjective evaluations or opinions.

[0035] 211. Construct a dictionary containing various entity types and process the archives to match the words in the archive text with the dictionary and match the words with the corresponding type of entity.

[0036] Specifically, a dictionary containing various entity types (such as names of people, places, organizations, products, etc.) is constructed; when processing archive text, the words in the text are matched with the dictionary, and when a match occurs, it is identified as an entity of the corresponding type. For example, for contract archives, there is a dictionary containing common company names, and when these company names appear in the text, they are identified as one of the two parties in the contract.

[0037] 212. Determine the location and grammatical structure of the entity based on the grammatical rules of the language.

[0038] Specifically, according to the grammatical rules of the language, determine the possible position and grammatical structure of the entity. For example, in Chinese, a noun phrase in the subject position is more likely to be an entity; by analyzing the grammatical structure of the sentence, find the noun phrase that meets the entity characteristics, such as "Party A (XX Company) agrees to purchase Party B's products", through grammatical analysis, it can be determined that "XX Company" is an entity.

[0039] 213. Extract features for entity recognition from the text, where the features include: features of the word itself, context features of the word, and part-of-speech features.

[0040] Specifically, features for entity recognition are extracted from the text, such as the features of the word itself (whether it is capitalized, whether it contains numbers, etc.), the context features of the word (what are the words before and after), part-of-speech features, etc. For example, for name entity recognition, the surname and first name usually have certain word usage characteristics, and some identity-related words (such as "Mr." and "Ms.") may appear around the name.

[0041] 214. Use the labeled entity training data and the supervised learning model to train the entity recognition model, and predict the entity and type of the newly input archive text based on the trained entity recognition model.

[0042] Specifically, using training data with labeled entities (such as labeling entities such as the contracting parties and contract subject matter in the contract text), and using supervised learning models (such as hidden Markov models, conditional random fields (CRF), BiLSTM-CRF in deep learning, etc.) for training, the trained model can predict entities and their types based on the features in the new text. For example, in a new contract text, the model can accurately identify the name entities of the contracting parties based on the previously learned features and rules.

[0043] 215. Using dependency syntactic analysis tools, a syntactic tree is constructed for the sentence to show the dependency relationship between words. According to the dependency type and path in the syntactic tree, rules are formulated to extract the relationship between entities.

[0044] Specifically, we use dependency parsing tools to construct a syntactic tree for sentences and display the dependency relationship between words. For example, in the sentence "Party A sells the product to Party B", we can get that "Party A" is the subject of the action "sell" through dependency syntactic analysis, and "Party B" is the object of the action. This subject-action-object relationship helps to identify the relationship between entities.

[0045] In addition, according to the dependency type and path in the syntactic tree, rules are formulated to extract the relationship between entities. For example, if there is a dependency path such as "nsubj (subject) - verb - dobj (direct object)", and both the subject and the object are entities, it can be extracted that there is an action relationship between the two entities, which is the relationship of "Party A - Seller - Party B" in the above example.

[0046] 221. By training the relation recognition model on texts labeled with semantic roles, the relation recognition model learns to recognize the semantic role played by each component in the sentence.

[0047] Specifically, by training on a large number of texts with annotated semantic roles (such as agent, patient, time, place, etc.), the model learns to identify the semantic role played by each component in the sentence. For example, for the sentence "Party A signed a contract with Party B at the company's headquarters on January 1, 2023", the model can recognize that "Party A" is the agent of the event "signing", "Party B" is the patient, "January 1, 2023" is the time, and "Company headquarters" is the location.

[0048] 222. According to the results of semantic role labeling, establish semantic role-based relationships between entities.

[0049] Specifically, according to the results of semantic role labeling, the relationship between entities based on semantic roles is established. For example, in the contract signing scenario, the agent-patient relationship between the two parties in the "signing" event is identified, as well as the relationship between the time and place of the event and the two parties to the contract.

[0050] 223. Pre-build a knowledge graph, where the knowledge graph contains various entities and relationship types between entities.

[0051] Specifically, an existing knowledge graph is pre-built or referenced, and the knowledge graph contains various entities and types of relationships between entities (such as partnerships between companies, kinship between people, etc.). When processing archival texts, the identified entities are matched with entities in the knowledge graph.

[0052] 224. When processing archival text, the identified entities are matched with the entities in the knowledge graph. When the entities in the text are successfully matched with the entities in the knowledge graph, the relationships in the knowledge graph are used to infer the relationships between the entities in the text and to supplement the missing relationship information.

[0053] For example, in a business cooperation file, if two company entities are identified, their specific relationship in the cooperation scenario described in the file text can be inferred through the cooperation relationship information about the two companies in the knowledge graph.

[0054] 231. Determine the keywords used to trigger the event. If the keywords are detected in the archive text, it is preliminarily determined that an event exists.

[0055] Specifically, we identify keywords that can trigger events, such as verbs that indicate actions ("buy", "sell", "sign", etc.), words that indicate changes in status ("change", "terminate", etc.). After detecting these trigger words in the text, we can preliminarily determine that there may be an event. For example, in a contract file, the trigger word "sign a contract" indicates a contract signing event.

[0056] 232. According to different types of events, event templates are constructed, and the content surrounding the trigger words in the text is matched with the event template to identify complete event information.

[0057] Specifically, event templates are constructed based on different types of events (such as contract signing events, product delivery events, etc.). The templates include the entity roles involved in the event (such as the contracting parties in the contract signing event), the conditions for the event to occur (such as meeting the contract terms), and the results of the event (such as the contract taking effect). The content surrounding the trigger words in the text is matched with the event template to identify the complete event information. For example, for the contract signing event template, when the text contains the names of the contracting parties, the signing date, and the content related to the contract taking effect, it can be matched successfully and a complete contract signing event can be identified.

[0058] 233. Using the theory of frame semantics, each event is mapped to a semantic frame with different slots. The semantic frame corresponding to the event is identified in the archival text and the information of each slot is filled.

[0059] Specifically, using the theory of frame semantics, each event corresponds to a semantic frame, which has different slots (such as the participants, time, location, method, etc. of the event). The semantic frame corresponding to the event is identified in the text, and the information of each slot is filled. For example, for a meeting event frame, the slot information such as the participants of the meeting (through entity recognition), the meeting time (through time information extraction), and the meeting location (through location entity recognition) are identified, so as to completely identify the meeting event.

[0060] 234. Use a pre-trained word vector model to convert words in the text into vector representations, where the word vector includes the semantic information of the words and the semantic relationship between words is represented by vector operations.

[0061] Specifically, pre-trained word vector models (such as Word2Vec, BERT, etc.) are used to convert words in the text into vector representations. These word vectors include the semantic information of the words, and the semantic relationship between words can be represented by vector operations. For example, by calculating the similarity of word vectors, the semantic proximity of two words (such as "purchase" and "purchase") can be determined. Then, the vectors of all words in the text are combined to obtain the semantic representation of the text, which is used to understand the overall semantics of the text.

[0062] 235. The context of the text is modeled using a set network model to capture the order information and contextual dependencies of words in a sentence.

[0063] Optionally, a recurrent neural network (RNN) or its variants (such as LSTM, GRU), or the self-attention mechanism in the Transformer architecture, is used to model the context of the text. These models can capture the order information and contextual dependencies of words in a sentence, thereby better understanding the semantics of the text. For example, in a sentence, context modeling can help understand the specific entity referred to by a pronoun, or understand the exact meaning of a polysemous word in the current context.

[0064] 236. Analyze the chapter structure of the text and complete event identification.

[0065] Specifically, analyze the text's chapter structure, such as the relationship between paragraphs (general-specific, cause-effect, parallel, etc.), the means of connection between sentences (conjunctions, pronoun reference, etc.), and grasp the overall logic and context of the text by understanding the chapter structure. For example, in a contract file, analyze the paragraph relationship between clauses, determine which clauses are main clauses and which are subsidiary clauses, and whether there is a causal relationship between clauses (such as what consequences will result from violating a main clause), so as to understand the context and semantics of the entire contract text.

[0066] 300. Set identification rules, automatically identify files according to the preset identification rules, and assign identification tasks.

[0067] It should be noted that the identification rule library module stores a wealth of archive identification rules, which are formulated by archive management experts based on the characteristics of different types of archives and relevant laws, regulations, industry standards, etc. The rules cover the importance, timeliness, integrity and other aspects of the archives. For example, for controlled archives, the rule library contains sensitive keywords, control level determination criteria, controlled period and other rules; for important archives, there are rules for determining the degree of their impact on organizational decision-making, business operations and other aspects. The rule library is scalable and maintainable, and administrators can add, modify and delete rules based on actual conditions.

[0068] 310. Set up rules for assigning identification tasks.

[0069] Optionally, a series of appraisal task allocation rules can be pre-set in the system. For example, according to the file type (such as document files, scientific and technological files, financial files, etc.), the tasks can be allocated to appraisers in different professional fields; according to the principle of workload balance among appraisers, the tasks can be allocated to those with less workload at present; or according to the importance and control level of the files, the tasks can be allocated to appraisers with corresponding authority and experience.

[0070] 320. When evaluating a new archival text, the relevant information of the archive is automatically read, and according to the evaluation task assignment rules, qualified personnel are searched in the evaluation personnel database to assign the evaluation task.

[0071] Optionally, when a new file needs to be appraised, the system automatically reads the relevant information of the file (such as type, importance, etc.), and then searches for qualified personnel in the appraisal personnel database according to the preset rules, and automatically assigns the appraisal task to them. For example, for a scientific and technological research and development file, the system will identify its type as a scientific and technological file, and then assign it to an appraisal personnel with professional knowledge in the field of science and technology.

[0072] 330. Collect the competency data of appraisers and establish a competency model for each appraiser. The competency data includes the completion quality and efficiency of past appraisal tasks, professional skills, and experience level.

[0073] Specifically, we collect the professional skills, experience level, quality and efficiency of past appraisal tasks of the appraisers, and other competence data to establish a competence model for each appraiser. For example, by analyzing the types of files that the appraiser has appraised in the past, the accuracy of the appraisal results, the time to complete the appraisal task, and other information, we generate a competence vector for each person, which contains competence indicator values ​​of different dimensions.

[0074] 340. Upon receiving a new appraisal task, the characteristics of the task are matched with the competency model of the appraiser, a machine learning algorithm is used to calculate the matching score between the task and each appraiser, and the appraiser is selected based on the score.

[0075] Specifically, when a new appraisal task is received, the characteristics of the task (such as the professional field of the file, the complexity, etc.) are matched with the ability model of the appraiser. Among them, machine learning algorithms (such as collaborative filtering, content-based recommendation algorithms, etc.) can be used to calculate the matching score between the task and each appraiser, and the appraiser with the highest score will be assigned the task. For example, for a file involving complex financial data, the system will give priority to assigning it to the appraiser with a high ability score in the financial field and a moderate amount of recent tasks.

[0076] 400. Set up a process model and receive real-time operations, automatically update the status of the archive at the corresponding process node and display it.

[0077] It should be noted that the appraisal process management module is responsible for managing the entire file appraisal process, including appraisal task allocation, process monitoring and progress tracking. This module assigns files to corresponding appraisers or automatically triggers the intelligent appraisal process according to the preset appraisal process (such as preliminary review, re-review, final review, etc.). During the appraisal process, it monitors the progress of each stage in real time to ensure that the appraisal work proceeds smoothly according to the process. At the same time, the module also provides a communication and collaboration platform to facilitate communication and information sharing between appraisers, and improve the efficiency and accuracy of appraisal.

[0078] 410. A process model for archive appraisal is predefined, wherein the process model includes specific operations and requirements at each stage and sets different states for each process node.

[0079] Optionally, a process model for archive appraisal can be pre-defined in the system, including various stages (such as preliminary review, re-review, final review, etc.) and the specific operations and requirements of each stage, and different statuses can be set for each process node, such as "pending", "processing", "completed", etc.

[0080] 420. After receiving the real-time operation on the archive, the status of the archive in the corresponding process node is automatically updated.

[0081] Optionally, after receiving real-time operations on the file (such as starting preliminary review, submitting review opinions, etc.), the status of the file at the corresponding process node will be automatically updated. For example, when the appraiser clicks the "Start preliminary review" button, the system updates the status of the file at the preliminary review node to "Processing" and records the start time.

[0082] 430. The real-time status of the archive appraisal process is displayed through a graphical interface.

[0083] Optionally, the real-time status of the file identification process can be displayed to administrators and relevant personnel through a graphical interface (such as a flowchart, progress bar, etc.). The administrator can intuitively see which stage each file is in, as well as the number and completion status of tasks in each stage. For example, in a flowchart, nodes of different colors represent different states, and the administrator can quickly see which files are stuck in the preliminary review stage and which have entered the final review stage.

[0084] 440. Set up abnormal situation rules based on business experience and process requirements, and monitor all data in the archive identification process in real time. When situations that meet the abnormal rules are found, the early warning mechanism is automatically triggered.

[0085] Specifically, a series of exception rules are set based on business experience and process requirements. For example, the processing time of a process node exceeds the preset deadline (for example, the initial review should be completed within 3 working days, if it exceeds the deadline, it will be considered an exception); or at a certain stage, the appraisal personnel have modified the appraisal results multiple times and the modification range is large.

[0086] Specifically, the data in the file identification process is monitored in real time. When a situation that meets the abnormal rules is found, the early warning mechanism is automatically triggered. The early warning method can be a notification within the system, an email reminder, a text message reminder, etc., to notify relevant personnel (such as administrators, identification personnel at this stage, etc.) to handle the abnormal situation in a timely manner. For example, if a file is not completed in the review stage for more than 5 working days, the system will send a warning email to the reviewer and administrator to remind them to pay attention to the identification progress of the file.

[0087] 450. When receiving a real-time operation on a file, automatically record the timestamp of the operation.

[0088] Specifically, when real-time operations on archives are received (such as opening archives, submitting appraisal opinions, passing audits, etc.), the timestamps of the operations are automatically recorded. Optionally, these timestamps are accurate to seconds to provide basic data for subsequent progress statistics.

[0089] 460. Based on the timestamp data, calculate the residence time of each file at each process node and the overall identification progress, and display the progress information in the form of a chart.

[0090] Specifically, based on the timestamp data, the time each file stays at each process node, the overall appraisal progress (such as the proportion of completed stages to the total number of stages), and other information are calculated, and these progress information are displayed to relevant personnel through dashboards, reports, etc. For example, an administrator can view a file from entering the appraisal process to the current time, staying in the preliminary review stage for 2 days, staying in the review stage for 1 day, and the overall progress is 60% (assuming there are a total of 5 stages, and 3 stages have been completed).

[0091] 470. Among them, there is a relationship and dependency between multiple tasks in the archive appraisal process.

[0092] Specifically, for some complex archive appraisal processes, there may be associations and dependencies between multiple tasks. For example, in the review stage, it may be necessary to wait for some supplementary materials submitted by the initial reviewer before starting; or the final review result needs to refer to the reviewer's opinions for comprehensive judgment. The system defines the preconditions and dependencies between these tasks.

[0093] 480. When the status of a task changes, the progress and status of related dependent tasks are automatically updated according to the task association relationship.

[0094] Specifically, when the status of a task changes, the progress and status of related dependent tasks are automatically updated according to the task association relationship. For example, if the initial reviewer submits supplementary materials, the system will automatically update the status of the review task from "waiting for materials" to "can start", and recalculate the estimated start time and overall progress of the review task, so that relevant personnel can clearly understand the progress changes caused by the dependency relationship between tasks, so as to arrange work and resources reasonably.

[0095] 500. Generate detailed identification reports and relevant statistical information based on the identification results, and store all data of the entire archive processing process.

[0096] It should be noted that the result output module generates a detailed appraisal report and related statistical information based on the appraisal results. The appraisal report includes the basic information of the archive, the appraisal conclusion (such as maintenance, opening, control, etc.), the appraisal basis (specific appraisal rules and semantic analysis results), and the recommended treatment measures. The statistical information can be analyzed by dimensions such as archive type, appraisal conclusion, and time to provide data support for archive management decisions.

[0097] It should be noted that the identification subject library module stores all data generated during the operation of the system, including original archive data, metadata, NLP semantic analysis results, identification rules, identification process records, identification results, etc. The database uses efficient data storage and retrieval technology to ensure data security, integrity and fast access performance.

[0098] 510. Obtain basic data of archives.

[0099] Specifically, the basic data of the archive is obtained, such as the file number, file name, creation time, creator, department, file type (such as document files, scientific and technological files, financial files, etc.), etc. This information is usually stored in a specific field of the database and can be extracted through a query statement.

[0100] For example, for a scientific and technological file named "XX Project R&D Report", its file number is "001-KJ01-XM005-ML001-AJ003", the creation time is May 1, 2023, the creator is Zhang San, the department is the R&D department, and the file type is scientific and technological file. This information will serve as the beginning of the appraisal report to provide background for subsequent content.

[0101] 520. Based on the appraisal results and comprehensive judgment at each stage of the appraisal process, the final appraisal conclusion shall be determined and written into the report.

[0102] Specifically, the final conclusion is determined based on the results and comprehensive judgment of each stage of the appraisal process. The conclusions may include maintenance, opening, control, etc. For example, after preliminary review, re-review and final review, the conclusion of a file whose control period has expired and whose content is no longer sensitive is open; while the conclusion of a file involving important business information and within its validity period is control.

[0103] Specifically, the appraisal conclusion will be clearly written into the report, usually expressed in a concise and clear sentence, such as "After appraisal, the appraisal conclusion of this archive is: open."

[0104] 530. Query the rules related to the file identification process from the identification rule library.

[0105] Specifically, the rules related to the archive identification process are queried from the identification rule library. For example, for an archive, it may involve sensitive keyword matching rules, level determination standard rules, etc., and the specific content of these rules and how to apply these rules in the archive identification are explained in detail. For example, during the identification process, by scanning the archive text for keywords, it is found that it contains sensitive keywords such as "XX core technical parameters". According to the level determination standard, the archive belongs to a non-public archive within the level, and the period is 5 years from the date of creation. As of this identification, the period has expired, and after expert review, the content is no longer sensitive, so it is determined to be no longer controlled.

[0106] 540. If natural language processing technology is used to conduct semantic analysis during the identification process, the results of the semantic analysis shall be included in the basis for identification.

[0107] Optionally, if natural language processing (NLP) technology is used for semantic analysis during the identification process, the results of the semantic analysis are also included in the identification basis. For example, through semantic analysis, it is found that the description of the market share of a certain product in the archives is consistent with public reports and there is no sensitive information. This result supports the final conclusion. It can be expressed as "After NLP semantic analysis, the description of the market share in the archives is consistent with public information, and no sensitive content is found, which further supports the identification conclusion."

[0108] 550. Based on the appraisal conclusion, put forward corresponding treatment measures.

[0109] Specifically, according to the appraisal conclusion, corresponding treatment measures are proposed. For example, for open archives, it is recommended to remove them from the controlled archives, manage and use them as ordinary archives, and update the archive status and related metadata in the system; for controlled archives, it is recommended to continue to store, borrow and approve them according to the management requirements of controlled archives, and conduct regular reviews. These treatment measures are written in detail in the report, such as "Since the archive is no longer controlled, it is recommended to transfer it from the controlled archives to the ordinary archives, update the status of the archive in the system to 'open', and indicate the opening date and related information in the archive catalog for subsequent management and use."

[0110] 560. Query all the files that have been appraised from the database, classify and count them according to the file type, and display the statistical results in the form of tables or charts.

[0111] Specifically, query all the files that have been appraised from the database and classify and count them according to the file type. For example, count the number and proportion of various types of files such as document files, scientific and technological files, and financial files. You can use SQL query statements, such as: SELECT COUNT(*) AS count, type FROM archives WHERE status='completed' GROUP BY type, where archives is the archive table, status is the archive identification status field, and type is the archive type field.

[0112] Optionally, the statistical results can be displayed in the form of a table or chart, such as a bar chart, with the horizontal axis representing the file type and the vertical axis representing the number, to intuitively present the identification status of different types of files.

[0113] 570. Query the files that have been appraised from the database, classify and count them according to the appraisal conclusions, and display the statistical results in the form of a pie chart.

[0114] Specifically, query the files that have been appraised from the database and classify and count them according to the appraisal conclusions (maintain, open, control, etc.). For example, count the number of controlled files, the number of maintained files, the number of open files, etc. Use similar SQL query statements, such as: SELECT COUNT(*)AS count, conclusion FROM archives WHERE status='completed' GROUP BY conclusion.

[0115] Optionally, the statistical results can be displayed in the form of a pie chart, with each sector area representing a different identification conclusion and its proportion, so that managers can quickly understand the distribution of various identification conclusions.

[0116] For example, statistics can be collected by time dimension such as year, month, week, etc. For example, to count the number of files that have been appraised each year, the query statement is as follows: SELECT YEAR(completed_date) AS year, COUNT(*) AS count FROM archivesWHERE status='completed 'GROUP BY YEAR(completed_date), where completed_date is the time field when the archive appraisal was completed.

[0117] Optionally, the statistical results of the time dimension can be displayed in a line graph, with the horizontal axis being time (such as year) and the vertical axis being the number of archives. The trend of the line graph can be used to observe the changing trends of the appraisal work in different time periods, providing data support in the time dimension for archive management decisions, such as whether it is necessary to increase the number of appraisal personnel or optimize the appraisal process in certain time periods.

[0118] 580. Store all data of the entire archive processing process, including original archive data, metadata, NLP semantic analysis results, identification rules, identification process records and identification results.

[0119] Therefore, the database adopts efficient data storage and retrieval technology to ensure the security, integrity and fast access performance of data.

[0120] For example, taking the appraisal of official documents and archives as an example, the scenario is as follows: a city needs to appraise a large number of official documents and archives, including notices, announcements, reports, requests for instructions, etc., to determine whether they can be opened to the public and the scope and method of opening. The content of official documents and archives involves all aspects of work, such as policies and regulations, government activities, people's livelihood affairs, etc.

[0121] Specifically, the implementation process includes: File entry and preprocessing: enter the official document files into the file entry module by scanning paper documents or importing the official document data into the e-government system. The system automatically extracts the metadata of the official document, such as the official document number, issuing authority, issuing date, etc., and preprocesses the official document text, including removing irrelevant information such as red headers and official seals, and performing text cleaning and word segmentation operations.

[0122] NLP semantic analysis and feature extraction: The NLP semantic analysis module conducts in-depth analysis of official document texts, identifies semantic information such as policy themes, departments and institutions involved, relevant legal and regulatory references, event descriptions, etc., extracts the subject terms of official documents, such as "urban construction", "environmental protection", "social security", etc., as well as keywords such as specific policy measures, project names, funding amounts, etc., and generates semantic vectors to represent the semantic features of official documents.

[0123] Rule-based identification: The identification process management module matches the semantic features and metadata of official documents with the rules in the identification rule library in accordance with the process and requirements of archive identification. For example, for official documents involving sensitive information, if the official document contains specific keywords or involves major decisions, important strategies, etc., and meets the criteria for determining sensitive information in the identification rules, the system will preliminarily determine that the official document is a sensitive document and push it to the relevant department for review. For some general notices and announcements, if the semantic analysis results show that the content does not involve sensitive information and complies with the provisions of information disclosure, the system can preliminarily determine that it is a public document.

[0124] Intelligent appraisal and manual intervention: For some policy and regulatory documents, the system can automatically perform intelligent appraisal based on the NLP semantic analysis results and the appraisal rule base, such as judging the timeliness and scope of application of the policy, and generate an appraisal conclusion. For some complex documents, such as those involving multi-department coordination, policy adjustments, or controversial documents, the system will trigger the manual intervention process. With the help of the NLP semantic analysis tools provided by the system, appraisers can deeply analyze the semantic content and potential impact of the documents, communicate and discuss with leaders and experts of relevant departments, and finally determine the appraisal conclusion of the documents, including whether they can be made public, the scope of disclosure (such as full text disclosure, partial disclosure, etc.) and the time of disclosure.

[0125] Result output and feedback: The appraisal result output module generates a detailed official document appraisal report based on the appraisal conclusion, including the basic information of the official document, the appraisal conclusion, the appraisal basis (specific appraisal rules and semantic analysis results involved), and the recommended treatment measures (such as the requirements for the storage and use of official documents, the channels and methods for publishing public documents, etc.). The appraisal report is stored in the archive management system database and fed back to relevant departments and leaders. At the same time, the system collects the feedback and opinions of the appraisal personnel on the appraisal results, as well as the public feedback after the actual disclosure, etc., which is used to continuously optimize the appraisal rule base and NLP semantic analysis model, improve the scientificity and accuracy of official document and archive appraisal, and better serve the work of information disclosure and government affairs management.

[0126] As described above, the embodiment of the present application performs preliminary metadata extraction and storage on the entered archives; performs semantic analysis on the archives to identify semantic elements in the archives, wherein the semantic elements include entities, relationships, and events; sets identification rules, automatically identifies the archives according to the preset identification rules, and assigns identification tasks; sets a process model, receives real-time operations, automatically updates the status of the archives at the corresponding process nodes and displays them; generates detailed identification reports and related statistical information based on the identification results, and stores all data of the entire archive processing process; deeply understands the semantic information of the archive content, so that the identification results are accurate and comprehensive, meeting the needs of modern archive management for efficient and accurate identification.

[0127] Specifically, the embodiment of the present application is based on the application of deep NLP semantic recognition technology: it innovatively integrates advanced NLP semantic recognition technology into the process of archive identification, and can accurately understand the semantic information of archive texts, breaking through the limitations of traditional archive identification based only on surface information and simple keyword matching. Through the comprehensive identification of semantic elements such as entities, relationships, and events and the extraction of semantic features, a richer and more accurate basis is provided for archive identification, greatly improving the accuracy and reliability of identification. For example, in the identification of a historical document archive, the system can not only identify the basic information such as people, places, and time in the document, but also understand the background, process, and impact of the historical events described in the document, so as to more accurately judge the historical value and importance of the document.

[0128] The embodiments of the present application realize the deep integration of technology and break through the traditional limitations: The limitations of traditional archives appraisal: Traditional archives appraisal methods are often limited to simple processing of archives surface information, such as preliminary classification and judgment based only on limited information such as archives titles and keywords. This method cannot deeply explore the intrinsic meaning and potential value of archives content, and it is easy to cause omissions and misjudgments of important information. For example, for some historical document archives, it may not be possible to accurately judge their importance and influence in a specific historical period only by the title, and simple keyword matching may not be able to fully reflect the true content and value of the archives due to unreasonable keyword selection and weight setting. Deep integration of NLP semantic recognition technology: The embodiments of the present application innovatively integrate advanced NLP semantic recognition technology into the archives appraisal process, achieving a comprehensive and in-depth understanding of archives texts. This technology can perform detailed grammatical analysis, semantic parsing and contextual understanding of archives texts, not just staying on the surface form of the text, but going deep into the semantic level of the text. For example, when processing a historical document archive, the system can accurately identify entity information such as names of people, places, and time, which can also be done by traditional methods, but it can also analyze the relationship between these entities, such as kinship and affiliation between characters, the sequence of events, and causal relationships. At the same time, the system can understand the background, process, and impact of the historical events described in the text. For example, by analyzing the description of a historical event in the document, including the cause of the event, the various forces involved, the development of the event, and the impact on the society, politics, and economy at the time, the value and importance of the document in historical research can be more accurately evaluated. This deep integration makes archive appraisal no longer a simple information screening, but an intelligent analysis process based on a comprehensive and in-depth understanding of the archive content.

[0129] The embodiment of the present application is about comprehensive identification and feature extraction of semantic elements: comprehensive identification of entities, relationships, and events: using NLP technology to accurately identify entities in archive texts, covering various types of entities, such as names (including historical figures, modern figures, fictional characters, etc.), place names (cities, regions, specific locations, etc.), institution names (institutions, enterprises, social organizations, etc.), time (specific date, time period, historical period, etc.), item names, etc. Taking a corporate archive as an example, the system can accurately identify the name, address, main person in charge, partners, product name, production time and other entity information of the enterprise. On the basis of identifying entities, the relationship between entities is further analyzed, such as the cooperative relationship between enterprises and partners, the corresponding relationship between products and production time, etc. At the same time, various events described in the archives can also be identified, such as the establishment of the enterprise, the development of products, marketing events, etc., and the time sequence, causal relationship, etc. of these events can be analyzed. For example, by analyzing corporate archives, it can be found that the successful development of a certain product is due to the cooperation between the enterprise and a scientific research institution, and the launch of the product has driven the growth of the enterprise's market share. The relationship and influence between these events can be accurately captured and analyzed by the system. Rich extraction of semantic features: In addition to the recognition of entities, relationships, and events, rich semantic features can also be extracted to provide a multi-dimensional basis for archive identification. These semantic features include but are not limited to subject terms, keywords, semantic vectors, etc. Subject terms can summarize the core content and subject areas of archives. For example, the subject terms of a scientific and technological archive may be "artificial intelligence", "machine learning", "data mining", etc.; keywords focus more on highlighting key information points in the archive, such as specific technical terms, important parameters, key events, etc.; semantic vectors represent the semantic information of the archive text in the form of vectors, which is convenient for operations such as semantic similarity calculation and cluster analysis. Through the extraction of these semantic features, the system can classify, compare and evaluate archives more accurately. For example, when identifying a batch of scientific and technological archives, archives related to a specific research direction can be quickly screened out based on subject terms and keywords, and then the similarity between archives can be calculated through semantic vectors to find archives with similar research content or technical routes, providing more comprehensive and accurate reference materials for scientific researchers, and also helping to discover the potential value and related information in archives, and improve the accuracy and reliability of archive identification.

[0130] The specific embodiment of the present application to improve the accuracy and reliability of identification: more accurate value judgment: taking historical documents and archives as an example, the traditional method may only classify them as general historical materials based on the title of the document or some surface information, but through the in-depth analysis of the content of the document by NLP semantic recognition technology, the system can discover the important information and value hidden therein. For example, for a seemingly ordinary diary document, the system can accurately judge that the document has important reference value for studying social life, interpersonal relationships, etc. in the historical period by identifying the social customs, folk sufferings, and details of the author's interactions with important historical figures at that time mentioned therein, combined with the semantic understanding and background analysis of these information, thereby avoiding the underestimation of value caused by the limitations of traditional identification methods. More comprehensive information consideration: when identifying archives, traditional methods may ignore other relevant factors because they only focus on some key information, thereby affecting the accuracy of the identification results. The NLP semantic recognition technology of the embodiment of the present application can comprehensively consider various information in the archives, including subtle expressions in the text, implicit logical relationships, etc. For example, when identifying a legal file, the system will not only focus on the basic facts of the case and the citation of legal provisions, but also understand factors such as the tone and emotional tendencies in the statements of the parties and the logical coherence between different evidence through semantic analysis, so as to more comprehensively and accurately judge the nature of the case, the reliability of the evidence and the accuracy of the application of the law, provide a more reliable basis for the identification of legal files, and reduce the risk of misjudgment due to incomplete information.

[0131] The embodiments of the present application are adapted to complex and diverse file types: different types of files have different characteristics and identification requirements, and traditional methods are often difficult to adapt to this diversity. The NLP semantic recognition technology of the embodiments of the present application has strong versatility and adaptability, and can effectively handle various types of files, whether historical documents, scientific and technological reports, legal documents, corporate archives or official documents. For example, for scientific and technological report archives, the system can accurately understand the complex content such as professional terms, technical processes and experimental data, and conduct identification based on this information; for official document archives, the system can identify the type of document, purpose of writing, policy content and scope of influence, etc., so as to provide accurate basis for the classification, level determination and public scope demarcation of official documents. This good adaptability to complex and diverse file types enables the embodiments of the present application to play an important role in different fields and scenarios, greatly improving the overall accuracy and reliability of file identification.

[0132] The embodiment of the present application integrates intelligent identification rules and NLP: the identification rule base is closely integrated with NLP semantic analysis to achieve the organic integration of rule-driven and semantic understanding. The identification rules are no longer isolated, static clauses, but are interconnected and mutually verified with the semantic content of the archives. The system can automatically match and apply corresponding identification rules according to the results of NLP semantic analysis. At the same time, the identification personnel can also better understand and interpret the application of identification rules through the results of semantic analysis, thereby improving the operability and adaptability of the identification rules. For example, for the identification of controlled archives, the system can identify keywords and their contextual contexts based on semantic analysis, accurately judge the controlled level and scope, and give reasonable identification conclusions based on the corresponding identification rules.

[0133] Specifically, the organic fusion of rules and semantics: Limitations of traditional identification rules: In traditional archive identification, identification rules are usually a series of pre-set isolated and static clauses. These rules are often formulated based on experience and general standards, and lack deep interaction with the specific content of the archives. For example, for the identification of controlled archives, traditional rules may simply list some keywords or stipulate that certain types of archives are controlled, but in practical applications, these rules are difficult to accurately respond to the diversity and complexity of archive content. Because relying solely on keyword matching may result in misjudgment, for example, a keyword may have different meanings in different contexts, and traditional rules cannot take into account such contextual differences, resulting in inaccurate identification results. Innovative way of intelligent integration: The embodiment of the present application closely combines the identification rule library with NLP semantic analysis, breaking the isolation and static nature of traditional rules. The identification rules are no longer rigid clauses, but are dynamically associated with the semantic content of the archives. The system uses NLP technology to conduct an in-depth analysis of the archive text and understand the semantic information therein, including entities, relationships, events, and contextual contexts. Then, according to these semantic analysis results, the corresponding identification rules are automatically matched and applied. For example, in the identification of controlled archives, the system can not only identify sensitive keywords, but also analyze the context in which these keywords appear. If the keywords appear in a text describing an ordinary workflow, and the context does not indicate that it is controlled, the system may not simply determine that the archive is controlled; but if the keywords appear in the context of key content such as core technology and strategic decisions, the system will combine relevant identification rules to accurately determine the level and scope of control. This organic fusion of rules and semantics makes the identification process more intelligent and precise, and can better adapt to the changes and diversity of archive content.

[0134] Automatic matching and application of rules in the embodiment of the present application: Efficient and accurate rule application: The system can automatically and quickly match the most suitable identification rules according to the real-time results of NLP semantic analysis, greatly improving the efficiency and accuracy of the identification work. Taking corporate archives as an example, when a new contract archive enters the system, the NLP semantic analysis module will quickly identify the key information in the contract, such as the parties to the contract, the subject matter of the contract, the contract amount, the terms, etc. Then, the system will automatically search for relevant rules in the identification rule library based on these semantic information, such as the importance judgment rules when the contract amount reaches a certain value, the rules involving specific technologies or business fields, etc. This process of automatic matching and application of rules does not require manual intervention, avoids omissions and errors that may occur in manual search and judgment of rules, and ensures the comprehensive and accurate application of identification rules. For example, for a contract involving large-scale capital transactions and containing sensitive technical information, the system can quickly identify these key elements, and accurately classify the contract as an important and controlled archive based on the corresponding identification rules such as importance, which provides a reliable basis for subsequent management and processing. Dynamic adaptation to different archives: Due to the different contents and characteristics of different archives, traditional static rules are difficult to meet diverse identification needs. The automatic matching and application rule mechanism of the embodiment of the present application has strong dynamic adaptability. Whether the archive is about historical events, scientific research, or commercial transactions, the system can extract its unique semantic features through NLP semantic analysis, and then flexibly match and apply identification rules based on these features. For example, for a historical document archive, the system may match relevant rules such as historical data importance assessment, authenticity judgment, and preservation value identification based on the semantic information such as the detailed description of historical events, the complexity of character relationships, and the source and age of the document; and for a scientific research report archive, the system will automatically apply the identification rules on the innovation and scientificity of scientific research results based on the semantic content such as the research methods, experimental data, research results, and the association with existing scientific research results in the report. This dynamic adaptability enables the system to be widely used in various types of archive identification, improving the versatility and practicality of the system.

[0135] The embodiment of the present application enhances the operability and adaptability of the identification rules: it is easy for identification personnel to understand and operate: for identification personnel, traditional identification rules may be relatively abstract and difficult to understand, especially when facing complex archives, it is difficult to accurately grasp the application conditions and scope of the rules. The embodiment of the present application provides identification personnel with more intuitive and clear operation guidance by combining identification rules with NLP semantic analysis results. Identifiers can better understand the content and semantic characteristics of archives by viewing the system's semantic analysis results of archives, and then understand why certain identification rules are applied and how these rules are related to the content of archives. For example, in the identification of controlled archives, identification personnel can see the keywords identified by the system and their contextual analysis, as well as how this information corresponds to the specific terms in the identification rules. This enables identification personnel to understand and execute identification rules more accurately, reduces identification differences caused by inconsistent understanding of rules, and improves the quality and consistency of identification work. The adaptability of rules is improved: with the continuous changes in archive content and management requirements, identification rules also need to be continuously updated and adjusted. The intelligent fusion mechanism of the embodiment of the present application enables identification rules to better adapt to these changes. On the one hand, the system can automatically discover the possible deficiencies or inapplicability of existing identification rules based on the new semantic patterns and features reflected in the results of NLP semantic analysis, and provide a basis for the optimization and updating of rules. For example, if the system finds that the keywords and semantic relationships in some emerging technology fields are not fully considered in the existing identification rules when analyzing a large number of new scientific and technological archives, it can prompt the administrator to supplement and improve the relevant rules. On the other hand, when the policies, regulations or business processes of archive management change, the administrator can quickly verify and adjust the effectiveness and applicability of the new rules by updating the identification rule library and combining the results of NLP semantic analysis. For example, when new regulations on the control of official documents are issued, the administrator can convert the new regulations into identification rules and enter them into the system, and then through semantic analysis and rule application tests on some typical official documents, ensure that the new rules can be accurately applied to the actual archive identification work, thereby improving the adaptability and timeliness of the identification rules, so that the system can always maintain an efficient and accurate identification state.

[0136] The collaborative mechanism between intelligent identification and manual identification in the embodiment of the present application: a working mode in which intelligent identification and manual identification cooperate and complement each other is established. Under the premise of ensuring the efficiency of identification, the high efficiency of artificial intelligence and the professional knowledge and experience of human experts are fully utilized. For simple archives with clear rules, the system can automatically complete the identification, which greatly improves work efficiency; for complex archives, manual identification personnel can use the NLP semantic analysis tools and intelligent identification results provided by the system to conduct more in-depth and accurate identification, ensuring the accuracy and authority of the identification results. This collaborative mechanism not only improves the overall efficiency of archive identification, but also ensures the quality of identification, and maximizes the advantages of man and machine.

[0137] Feedback optimization mechanism of the appraisal results of the embodiment of the present application: It has a complete appraisal result feedback optimization mechanism, which can continuously optimize the appraisal rule base and NLP semantic analysis model according to the feedback of the appraisers and the actual appraisal situation. Through the accumulation and analysis of a large number of appraisal cases, the system can automatically learn and adjust the weights, thresholds and parameters of the NLP model of the appraisal rules, so that it can better adapt to the appraisal needs of different types of archives, and realize the self-evolution and continuous improvement of the system. This mechanism enables the system to continuously improve the accuracy and intelligence level of appraisals as time goes by and the frequency of use increases, and provide users with better quality archive appraisal services.

[0138] Specifically, the comprehensiveness and systematicness of the feedback optimization mechanism: the shortcomings of the traditional system: the traditional archives appraisal system often lacks an effective feedback mechanism. Once the appraisal rules and models are set, it is difficult to dynamically adjust and optimize according to the actual usage. This results in the system being unable to make adaptive changes in a timely manner when faced with new file types, changing appraisal standards, or constantly accumulated appraisal experience, thereby affecting the accuracy and effectiveness of the appraisal. For example, over time, the importance of certain archives may change, or the introduction of new laws and regulations may lead to adjustments to the appraisal standards of controlled archives, but the traditional system cannot automatically perceive these changes and make corresponding optimizations, and can only rely on manual updating of rules, which is not only inefficient, but also prone to omissions and errors. The innovation of the embodiment of the present application: The embodiment of the present application has a complete appraisal result feedback optimization mechanism, covering multiple dimensions from feedback from appraisers to analysis of actual appraisal situations. As front-line users, the experience and judgment of appraisers are crucial to system optimization. The system can collect direct feedback from appraisers on the appraisal results, including recognition, questioning, and modification suggestions for the appraisal conclusions. At the same time, the system will automatically record various data in the actual identification process, such as the distribution of identification results of different types of archives, the triggering frequency and effect of identification rules, the performance of NLP semantic analysis models in different scenarios, etc. Through the comprehensive analysis of this rich information, the system can fully understand its own strengths and weaknesses in identification work, and provide a solid data foundation and direction for subsequent optimization.

[0139] Automatic learning and adjustment based on data accumulation: Implementation of automatic learning: The system uses advanced machine learning algorithms to accumulate and analyze a large number of identification cases to achieve automatic learning and adjustment of the identification rule base and NLP semantic analysis model. Specifically, for the identification rule base, the system will automatically adjust the weight and threshold of the identification rules according to the accuracy and reliability of different identification results and the feedback of the identification personnel. For example, in the identification of controlled archives, if a keyword is judged as controlled by the system in a specific context with a high accuracy, then the weight of the identification rule corresponding to the keyword will increase accordingly; conversely, if a rule often leads to misjudgment, the system will automatically reduce its weight or adjust the threshold to improve the applicability of the rule. For the NLP semantic analysis model, the system will automatically adjust the parameters of the model according to the semantic analysis effect of the archive text, such as the accuracy of entity recognition, the completeness of relationship extraction, the validity of the semantic vector, etc. For example, when the system finds that the entity recognition accuracy of a certain type of archive is low, it will automatically adjust the parameters related to this type of archive in the model, such as the dimension of the word vector, the number of layers and nodes of the neural network, etc., to improve the performance of the model when processing this type of archive.

[0140] Advantages of adapting to different types of archives: This automatic learning and adjustment mechanism based on data accumulation enables the system to quickly adapt to the identification needs of different types of archives. Whether it is historical documents, scientific and technological reports, commercial contracts or official documents, the system can gradually optimize its identification capabilities in the process of continuously processing and analyzing these archives. For example, for historical document archives, as the number of documents processed by the system increases, it will automatically learn the language characteristics, common vocabulary and expressions of different historical periods, so as to more accurately identify entities, relationships and events in the documents, and provide a more accurate basis for the identification of historical documents; for scientific and technological report archives, the system can optimize the NLP semantic analysis model based on the continuously accumulated professional terms, technical processes and research methods, and better understand the content and value of scientific and technological reports, and then adjust the identification rules to ensure that the identification of scientific and technological archives is in line with its professional characteristics and actual value. This adaptive ability enables the system to always maintain a high level of identification accuracy and intelligence when facing a variety of archives, providing users with more personalized and accurate archive identification services.

[0141] Self-evolution and continuous improvement of the system of the embodiment of the present application: The process and effect of self-evolution: As time goes by and the frequency of use increases, the embodiment of the present application can continuously evolve and improve itself. In the initial stage, the system may work based on some general identification rules and NLP models, but with the accumulation of a large number of identification cases and the continuous integration of feedback information, the system will gradually form a set of optimized identification rule bases and NLP semantic analysis models for specific user groups and file types. For example, a system dedicated to enterprise archive management will continuously optimize itself during long-term use according to the business characteristics of the enterprise, the style of archive content, and the identification standards and habits within the enterprise. For common contract archives of enterprises, the system will more accurately identify key information and potential risks in contract terms, and the judgment of sensitive information will be more in line with the actual control requirements of the enterprise; for the R&D project archives of the enterprise, the system can better understand the professional content in the technical documents and accurately evaluate the innovation and importance of the project. This self-evolution enables the system to deeply integrate with the needs and usage scenarios of users, and provide more and more high-quality archive identification services. Value enhancement brought by continuous improvement: The continuous improvement of the system is not only reflected in the improvement of identification accuracy, but also in work efficiency, user experience and other aspects. By optimizing the identification rules and NLP models, the system can process archives faster, reduce the number of misjudgments and manual interventions, and thus improve overall work efficiency. At the same time, more accurate and intelligent identification results can also provide users with more valuable references, helping them to better manage and utilize archive resources. For example, for archive managers, the system can more accurately screen out important archives and controlled archives, reduce their workload, and improve the standardization and security of archive management; for scientific researchers or corporate decision makers, the system's accurate identification and analysis of relevant archives can provide them with more comprehensive and accurate information support, helping them make more scientific decisions. This ability to continuously improve makes the system more vital and competitive, able to always maintain a leading position in the ever-changing archive management environment, and create greater value for users.

[0142] The above steps are not to be performed in a strict order as described in the numbers, but should be understood as an overall solution.

[0143] In the second aspect, based on the above embodiments, Figure 2 A schematic diagram of the structure of a file identification device based on semantic recognition of natural language processing technology provided in an embodiment of the present application. Figure 2 The archive identification device based on semantic recognition of natural language processing technology provided in this embodiment specifically includes: an input advance module 201, a semantic analysis module 202, an identification processing module 203, a process processing module 204 and a result generation module 205.

[0144] Among them, the advance entry module 201 is used to extract and store preliminary metadata of the entered archives; the semantic analysis module 202 is used to perform semantic analysis on the archives and identify the semantic elements in the archives, wherein the semantic elements include entities, relationships and events; the identification processing module 203 is used to set identification rules, automatically identify archives according to preset identification rules, and assign identification tasks; the process processing module 204 is used to set the process model, and receive real-time operations, automatically update the status of the archives at the corresponding process nodes and display them; the result generation module 205 is used to generate a detailed identification report and related statistical information according to the identification results, and store all data of the entire archive processing process.

[0145] As described above, the embodiment of the present application performs preliminary metadata extraction and storage on the entered archives; performs semantic analysis on the archives to identify semantic elements in the archives, wherein the semantic elements include entities, relationships, and events; sets identification rules, automatically identifies the archives according to the preset identification rules, and assigns identification tasks; sets a process model, receives real-time operations, automatically updates the status of the archives at the corresponding process nodes and displays them; generates detailed identification reports and related statistical information based on the identification results, and stores all data of the entire archive processing process; deeply understands the semantic information of the archive content, so that the identification results are accurate and comprehensive, meeting the needs of modern archive management for efficient and accurate identification.

[0146] The archive appraisal device for semantic recognition based on natural language processing technology provided in the embodiment of the present application can be used to execute the archive appraisal method for semantic recognition based on natural language processing technology provided in the above embodiment, and has corresponding functions and beneficial effects.

[0147] In a third aspect, an embodiment of the present application further provides an electronic device that can integrate the archive identification device for semantic recognition based on natural language processing technology provided in an embodiment of the present application. Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device includes: an input device 33, an output device 34, a memory 32 and one or more processors 31; the memory 32 is used to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the archive identification method based on semantic recognition of natural language processing technology as provided in the above embodiment. The input device 33, the output device 34, the memory 32 and the processor 31 can be connected by a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0148] The processor 31 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 32, that is, realizes the above-mentioned archive identification method of semantic recognition based on natural language processing technology.

[0149] The electronic device provided above can be used to execute the archive identification method based on semantic recognition of natural language processing technology provided in the above embodiment, and has corresponding functions and beneficial effects.

[0150] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the archive identification method based on semantic recognition based on natural language processing technology as described above, and can achieve the same beneficial effects.

[0151] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present application is not limited to the archive identification method based on semantic recognition of natural language processing technology as described above, and can also execute related operations in the archive identification method based on semantic recognition of natural language processing technology provided in any embodiment of the present application.

[0152] In a fifth aspect, the embodiments of the present application also provide a computer program product, and the methods described in the various embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on a computer, the processes or functions described in the various embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model) or other programmable device.

[0153] The computer program or instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program or instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it may also be an optical medium, such as a digital video disk; it may also be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0154] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0155] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0156] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0157] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0158] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0159] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method for file identification based on semantic recognition of natural language processing technology, characterized in that: The method comprises: Extract and store preliminary metadata of the entered archives; Perform semantic analysis on the archives to identify semantic elements in the archives, where the semantic elements include entities, relations and events; Set up identification rules, automatically identify files according to the preset identification rules, and assign identification tasks; Set up the process model and receive real-time operations, automatically update the status of the archive at the corresponding process node and display it; Generate detailed identification reports and relevant statistical information based on the identification results, and store all data of the entire archive processing process.

2. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: The preliminary metadata extraction and storage of the input archives includes: Input files, including paper files and electronic document files; Convert the entered files into a unified format; After preliminary metadata extraction of the entered archives, they are stored in the database, where the metadata includes the archive type, creation time and source.

3. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: The semantic analysis of the archive is performed to identify the semantic elements in the archive, wherein the semantic elements include entities, relationships and events, including: Construct a dictionary containing various entity types and process the archive to match words in the archive text with the dictionary and words with entities of the corresponding type; Determine the location and grammatical structure of the entity according to the grammatical rules of the language; Extracting features for entity recognition from the text, wherein the features include: features of the word itself, context features of the word, and part-of-speech features; Using the labeled entity training data, the supervised learning model is used to train the entity recognition model, and the trained entity recognition model is used to predict the entity and type of the newly input archive text; Using dependency syntactic analysis tools, a syntactic tree is constructed for the sentence to show the dependency relationship between words. According to the dependency relationship type and path in the syntactic tree, rules are formulated to extract the relationship between entities. By training the relation recognition model on texts with annotated semantic roles, the relation recognition model learns to recognize the semantic role played by each component in the sentence; According to the results of semantic role labeling, establish the relationship between entities based on semantic roles; Pre-build a knowledge graph, where the knowledge graph contains various entities and relationship types between entities; When processing archival text, the identified entities are matched with the entities in the knowledge graph. When the entities in the text are successfully matched with the entities in the knowledge graph, the relationships in the knowledge graph are used to infer the relationships between the entities in the text and to supplement the missing relationship information. Determine the keywords used to trigger the event. If the keywords are detected in the archive text, it is preliminarily determined that an event exists; According to different types of events, event templates are constructed, and the content surrounding the trigger words in the text is matched with the event template to identify complete event information; Using the theory of frame semantics, each event is mapped to a semantic frame with different slots. The semantic frame corresponding to the event is identified in the archival text and the information of each slot is filled. Use the pre-trained word vector model to convert words in the text into vector representations, where the word vector includes the semantic information of the words and the semantic relationship between the words is represented by vector operations; The context of the text is modeled using a set network model to capture the order information and contextual dependencies of words in a sentence; Analyze the chapter structure of the text and complete event identification.

4. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: The setting of identification rules, automatically identifying files according to preset identification rules, and assigning identification tasks include: Set up rules for assigning identification tasks; When evaluating a new file text, the relevant information of the file is automatically read, and according to the evaluation task assignment rules, qualified personnel are searched in the evaluation personnel database to assign the evaluation task; Collect the competency data of the appraisers and establish a competency model for each appraiser, where the competency data includes the completion quality and efficiency of past appraisal tasks, professional skills, and experience level; When a new appraisal task is received, the characteristics of the task are matched with the competency model of the appraiser, a machine learning algorithm is used to calculate the matching score between the task and each appraiser, and the appraiser is selected based on the score.

5. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: The process model is set, and real-time operations are received, and the status of the archive at the corresponding process node is automatically updated and displayed, including: Predefine the process model of archive appraisal, where the process model includes specific operations and requirements at each stage, and sets different states for each process node; After receiving the real-time operation on the file, the status of the file in the corresponding process node is automatically updated; The real-time status of the archives identification process is displayed through a graphical interface; According to business experience and process requirements, abnormal situation rules are set, and various data in the archive identification process are monitored in real time. When situations that meet the abnormal rules are found, the early warning mechanism is automatically triggered.

6. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: After the process model is set, and real-time operations are received, and the status of the archive at the corresponding process node is automatically updated and displayed, the following steps are also included: When receiving real-time operations on archives, automatically record the timestamp of the operations; Based on the timestamp data, calculate the residence time of each file at each process node and the overall identification progress, and display the progress information in the form of a dashboard or report; Among them, there are associations and dependencies between multiple tasks in the archives appraisal process; When the status of a task changes, the progress and status of related dependent tasks are automatically updated based on the task association relationship.

7. The archive identification method based on semantic recognition of natural language processing technology according to claim 1 is characterized in that: The above generates a detailed identification report and related statistical information based on the identification results, and stores all data of the entire archive processing process, including: Obtain basic data of archives; Determine the final conclusion of the appraisal based on the appraisal results and comprehensive judgment of each stage of the appraisal process, and write the conclusion into the report; Query the rules related to the file identification process from the identification rule library; If natural language processing technology is used to conduct semantic analysis during the identification process, the results of the semantic analysis shall be included in the basis for identification; According to the appraisal conclusion, put forward corresponding treatment measures; Query all the files that have been appraised from the database, classify and count them according to the file type, and display the statistical results in the form of tables or charts; Search the database for files that have been appraised, classify and count them according to the appraisal conclusions, and display the statistical results in the form of a pie chart; Stores all data of the entire archive processing process, including original archive data, metadata, NLP semantic analysis results, identification rules, identification process records and identification results.

8. A file identification device based on semantic recognition of natural language processing technology, characterized in that: include: The pre-entry module is used to extract and store preliminary metadata of the entered archives; A semantic analysis module is used to perform semantic analysis on the archives and identify semantic elements in the archives, wherein the semantic elements include entities, relations and events; The identification processing module is used to set identification rules, automatically identify files according to the preset identification rules, and assign identification tasks; The process processing module is used to set the process model and receive real-time operations, automatically update the status of the archive at the corresponding process node and display it; The result generation module is used to generate a detailed identification report and related statistical information based on the identification results, and store all data of the entire archive processing process.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the archive identification method based on semantic recognition based on natural language processing technology as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a server, implements the archive identification method based on semantic recognition of natural language processing technology as described in any one of claims 1-7.

Images