Enterprise-level digital clone running method and device based on personalized increment, equipment and medium

Abstract

The application discloses an enterprise-level digital clone running method and device based on personalized increment, equipment and medium, relates to the technical field of artificial intelligence, including: under the condition of meeting the preset authorization, preprocessing the original file of the target personnel, and classifying the sample; the preset authorization condition is that the authorization of the target personnel for the distillation operation has been previously obtained; according to the classification result, the working skills and personality style of the target personnel are extracted and sorted respectively to obtain the working skills draft and personality style draft; the draft is checked and the personal increment package draft to be audited is generated; the audit result is obtained by feeding back the personal increment package draft, if the audit is passed, the target personal increment package is constructed and stored based on the personal increment package draft, so that when the enterprise-level digital clone system runs, the runtime control package is constructed by using the target personal increment package, the preset system public capability package and the preset business field capability package, and the enterprise business service is provided.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, and medium for operating enterprise-level digital clones based on personalized incremental changes. Background Technology

[0002] In digital avatar, intelligent question answering, and personalized intelligent agent systems, existing solutions can typically generate a personalized performance based on an individual's data. This includes solutions that generate chatbots based on person data, making the chatbots more similar to a specific person in style and content. However, these solutions tend to focus more on the chat style and can only imitate the person's tone of voice, making it difficult to meet the compliance control and standardized operation requirements of professional scenarios such as finance and consulting. Personality / skill distillation solutions can only complete the one-time generation from multi-source data to a single-layer skill / personality text package, focusing on the distillation result output. The result is more like a text-based skill or personality summary rather than a runtime controllable configuration object. The avatar has poor fidelity, controllability, and maintainability, and its adaptability to enterprise scenarios with high requirements for professional accuracy is low.

[0003] In summary, optimizing the operation method of enterprise-level digital clones to adapt to enterprise-level operational requirements and reduce the risk of black-box generation are urgent problems that need to be solved. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for operating enterprise-level digital avatars based on personalized incremental changes, which can optimize the operation method of enterprise-level digital avatars to adapt to enterprise-level operational requirements and reduce the risk of black-box generation. The specific solution is as follows: Firstly, this application provides a method for operating an enterprise-level digital clone based on personalized incremental changes, including: Under the preset authorization conditions, the original file data of the target personnel is preprocessed to obtain preprocessed file data, and the preprocessed file data is sampled and classified to obtain the corresponding classification results; the preset authorization conditions are that the authorization of the target personnel for distillation operation has been obtained in advance. Based on the classification results, the work skills and personality styles of the target personnel are extracted from the preprocessed archives and then sorted and merged to obtain merged drafts of work skills and personality styles. The job skills draft and the personality style draft are verified and checked to obtain the corresponding verification results, and a personal incremental package draft to be reviewed is generated to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft, and the personality style draft; Feedback is provided on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft of the personal incremental package. When the enterprise-level digital clone system is running, the target personal incremental package, the preset system common capability package, and the preset business domain capability package are used to construct a runtime control package, and enterprise business services are provided based on the runtime control package.

[0005] Optionally, the preprocessing of the original personnel files to obtain preprocessed files includes: The original files of the target personnel are deduplicated, desensitized, and denoising to obtain the denoising files. The denoised archival data is segmented into segments, and the segmented segments are scored for quality. The segments are then sorted in descending order of quality score to obtain the preprocessed archival data.

[0006] Optionally, the step of classifying the preprocessed archival data to obtain corresponding classification results includes: According to the preset classification rules, each data fragment in the preprocessed archive data is classified into samples to determine the corresponding sample categories and obtain the classification results. The sample categories include first candidate samples that only match the job skills, second candidate samples that only match the personality style, third candidate samples that match both the job skills and personality style, and invalid samples that do not match either the job skills or personality style.

[0007] Optionally, the step of extracting the target personnel's work skills and personality style from the preprocessed archival data based on the classification results includes: The job skills of the target personnel are extracted from the first candidate sample and the third candidate sample to obtain several job skill candidate items and corresponding confidence scores; The personality styles of the target person are extracted from the second and third candidate samples to obtain several personality style candidate items and corresponding confidence scores.

[0008] Optionally, the process of sorting and merging the materials to obtain merged drafts of job skills and personality style includes: Identify candidate item pairs among the target candidate items whose semantic similarity is higher than a preset similarity threshold; The candidate item pairs in the target candidate items are merged, and the candidate items with confidence scores lower than the first preset confidence score threshold are removed, and deduplication is performed to obtain the sorted target candidate items. Detect conflicting candidate entries in the sorted target candidate entries and mark the conflicting candidate entries to obtain the merged target draft; Wherein, if the target candidate item is one of the several job skill candidate items, then the target draft is the job skill draft; if the target candidate item is one of the several personality style candidate items, then the target draft is the personality style draft.

[0009] Optionally, the verification and checking of the job skills draft and the personality style draft to obtain corresponding verification results includes: Detect conflict items in the target draft to obtain a list of conflict items; the target draft includes the job skills draft and the personality style draft; the conflict items include internal contradictions within the target draft and conflicts between the target draft and the preset system common capability package and the preset business domain capability package, respectively; Filter out low-confidence items in the target draft whose confidence scores are lower than the second preset confidence score threshold to obtain a list of low-confidence items; Risk items in the target draft that meet preset violation conditions are filtered to obtain a verification result including the list of conflict items and the list of low-confidence items.

[0010] Optionally, before generating the draft of the individual incremental package to be reviewed, the process further includes: The usability of the job skills draft and the personality style draft is verified by using preset test questions to obtain corresponding evaluation results; Accordingly, generating the draft of the individual incremental package to be reviewed includes: Based on the assessment results, the verification results, the job skills draft, and the personality style draft, a draft of the personal incremental package to be reviewed is generated.

[0011] Secondly, this application provides an enterprise-level digital avatar operating device based on personalized incremental updates, comprising: The sample classification module is used to preprocess the original file data of the target personnel under the premise of meeting the preset authorization conditions to obtain the preprocessed file data, and to classify the preprocessed file data to obtain the corresponding classification results; the preset authorization conditions are that the target personnel have been granted authorization for distillation operation in advance. The sorting and merging module is used to extract the work skills and personality styles of the target personnel from the preprocessed archive data according to the classification results, and sort and merge them to obtain the merged work skills draft and personality style draft. The draft generation module is used to verify and check the job skills draft and the personality style draft to obtain the corresponding verification results, and generate a personal incremental package draft to be reviewed in order to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft and the personality style draft; The incremental package storage module is used to provide feedback on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft of the personal incremental package. In order to build a runtime control package using the target personal incremental package, the preset system public capability package and the preset business domain capability package when the enterprise-level digital clone system is running, and provide enterprise business services based on the runtime control package.

[0012] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned enterprise-level digital avatar operation method based on personalized incremental changes.

[0013] Fourthly, this application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned enterprise-level digital clone operation method based on personalized incremental changes.

[0014] In this application, under preset authorization conditions, the original personnel file data of the target personnel is preprocessed to obtain preprocessed file data, and the preprocessed file data is sampled and classified to obtain corresponding classification results; the preset authorization conditions are that the authorization of the target personnel for distillation operation has been obtained in advance; according to the classification results, the work skills and personality style of the target personnel are extracted from the preprocessed file data, and sorted and merged to obtain merged work skill drafts and personality style drafts; the work skill drafts and personality style drafts are verified and checked to obtain corresponding verification results. As a result, a draft of a personal incremental package to be reviewed is generated to complete the distillation operation for the target personnel. The draft personal incremental package includes the verification result, the work skills draft, and the personality style draft. Feedback is provided on the draft personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft personal incremental package. When the enterprise-level digital avatar system is running, the target personal incremental package, the preset system common capability package, and the preset business domain capability package are used to construct a runtime control package, and enterprise business services are provided based on the runtime control package. As can be seen from the above, this application, under the premise of obtaining authorization for the distillation operation of the target personnel, first preprocesses and classifies their original file data, extracts and integrates the target personnel's work skills draft and personality style draft based on the classification results, verifies the two drafts and generates a draft of the personal incremental package to be reviewed containing the verification results, completes the data distillation of the target personnel, and after feedback and approval, generates and stores the target personal incremental package from the draft of the personal incremental package. When the enterprise-level digital avatar system runs, it integrates the incremental package, the preset system public capability package and the preset business domain capability package to assemble the runtime control package, and provides enterprise business services to the outside world based on the control package.In this way, through the above-described process of this application, preset authorization conditions for the distillation operation of target personnel are set as a pre-operation threshold, avoiding data compliance risks caused by unauthorized collection and processing of personal files, and protecting personnel privacy. The original files are preprocessed before sample classification, which standardizes the format of the original files and removes invalid and redundant information, thereby improving the recognition accuracy of subsequent skill and personality feature extraction. Work skills and personality styles are extracted and merged independently according to the classification results, realizing the decoupled storage of the two types of personnel features, which makes it easy for the digital clone to call the corresponding feature dimensions as needed. The two drafts are checked and the verification results are included in the incremental package draft, intercepting problems such as incomplete feature extraction, logical conflicts, and information distortion in advance, reducing the cost of manual secondary correction. During the operation phase, the personal incremental package, system common capability package, and business domain capability package are integrated to assemble the runtime control package, decoupling and encapsulating the personalized characteristics, general basic capabilities, and industry business capabilities of personnel. This allows for flexible adaptation to different enterprise business scenarios to quickly generate digital clone operation carriers, supporting the implementation of various enterprise business services, thereby optimizing the enterprise-level digital clone operation method to adapt to enterprise-level operation requirements and reducing the risk of black box generation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This application discloses a flowchart of an enterprise-level digital clone operation method based on personalized incremental changes; Figure 2 This is a system architecture block diagram of an enterprise-level digital clone operation method disclosed in this application; Figure 3 This is a schematic diagram of the structure of a personal incremental package disclosed in this application; Figure 4 This is a schematic diagram of the relationship of a runtime assembly disclosed in this application; Figure 5 This is a schematic diagram of the structure of an enterprise-level digital clone operation device based on personalized incremental operation disclosed in this application; Figure 6 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In digital avatar, intelligent question answering, and personalized intelligent agent systems, existing solutions can typically generate a personalized performance based on an individual's data. This includes solutions that generate chatbots based on person data, making the chatbots more similar to a specific person in style and content. However, these solutions tend to focus on mimicking the person's conversational style and tone of voice, rather than equipping them with enterprise-level professional capabilities. This makes it difficult to meet the compliance control and standardized operation requirements of professional scenarios such as finance and consulting. Personality / skill distillation solutions can only complete the one-time generation from multi-source data to a single-layer skill / personality text package, focusing on the distillation result. The result is more like a text-based skill or personality summary than a runtime controllable configuration object. The avatar has poor fidelity, controllability, and maintainability, making it unsuitable for enterprise scenarios with high requirements for professional accuracy.

[0019] To overcome the aforementioned technical problems, this application provides an enterprise-level digital clone operation method based on personalized incremental changes, which can optimize the enterprise-level digital clone operation method to adapt to enterprise-level operational requirements and reduce the risk of black-box generation.

[0020] See Figure 1 As shown, this embodiment of the invention discloses an enterprise-level digital avatar operation method based on personalized incremental updates, including: Step S11: Under the preset authorization conditions, preprocess the original file data of the target personnel to obtain preprocessed file data, and perform sample classification on the preprocessed file data to obtain the corresponding classification results; the preset authorization conditions are that the authorization of the target personnel for distillation operation has been obtained in advance.

[0021] In this embodiment, provided that the target personnel have been authorized to perform distillation operations and the preset authorization conditions are met, the original personnel file data is preprocessed to obtain regularized preprocessed file data. Then, the preprocessed file data is sample-classified, and the corresponding classification results are output. It is understood that the original file data is legally sourced and obtained with authorization, meeting data privacy compliance requirements.

[0022] It should be noted that, in order to transform a teacher's multi-source unstructured data into a reviewable, publishable, versionable, and runtime-assembleable personal incremental package for subsequent digital avatar execution, this application proposes an enterprise-level digital avatar operation method based on personalized incremental data, addressing the shortcomings of existing technologies. The key lies in providing a method for distilling, reviewing, publishing, and runtime assembly of a personal incremental package. The core is that it does not directly generate a chatbot from the original data, nor does it directly generate a prompt message; instead, it generates a personal incremental package. This personal incremental package consists of a work skill coverage package and a personality coverage package, and can be dynamically assembled with public capability packages and domain capability packages at runtime. Figure 2 The diagram shown is a system architecture block diagram of an enterprise-level digital avatar operation method provided in this application. The method includes: a teacher profile creation module, a raw data access module, a distillation task arrangement module, a work skill extraction module, a personality extraction module, a candidate merging module, a verification module, a draft generation module, an approval and release module, and a runtime assembly module. The system first establishes teacher profiles and then integrates raw data. The distillation task orchestration module drives subsequent steps. The job skill extraction module and personality extraction module execute in parallel or sequentially to generate two types of candidate results. The candidate merging module deduplicates and merges multiple candidates. The verification module detects conflicts, out-of-bounds errors, and low-confidence items. The draft generation module creates a reviewable draft. The review and release module releases the draft as a formal personal incremental package. The runtime assembly module then combines this personal incremental package with other capability packages for use in the digital avatar execution chain. This allows for the extraction of runnable, reusable, and reviewable personal incremental capabilities from multi-source raw data. The system separates and represents teachers' professional abilities and personalized styles, forming structured objects. The distillation process is controlled by a task state machine, supporting preprocessing, classification, extraction, verification, review, release, and rollback. Teachers and administrators can participate in review and correction, avoiding black-box generation. The released personal incremental package can be dynamically assembled with public capability packages and domain capability packages at runtime, thus serving the digital avatar execution chain.

[0023] Specifically, the distillation task orchestration module controls the entire process of generating individual incremental packages. This module maintains at least one distillation task object and a set of distillation step objects. The distillation task object includes at least: task identifier, teacher identifier, target version identifier, current task status, number of retries, error messages, start time, and end time. The distillation step objects include at least: step identifier, corresponding task identifier, step name, step order, current status, input reference, output reference, model identifier used, and prompt template version used. The distillation task orchestration module advances the task according to a preset state machine: task creation, data access, preprocessing, sample classification, job skill extraction, personality extraction, candidate merging, verification, draft generation, automatic evaluation, waiting for review, review, and release. If a step fails, it enters a failure state and performs step-level retries according to the retry strategy, without having to redo the entire process. The distillation task status is illustrated below: CREATED (Task created) → INGESTING (Raw data being imported) → PREPROCESSING (Preprocessing in progress) → SAMPLE_CLASSIFYING (Sample classification in progress) → WORK_SKILL_EXTRACTING (Work skills extraction in progress) → PERSONA_EXTRACTING (Personality style extraction in progress) → MERGING (Candidate merging in progress) → VALIDATING (Validation check in progress) → GENERATING_DRAFT (Draft generation in progress) → AUTO_EVALUATING (Automatic evaluation in progress) → WAITING_REVIEW (Waiting for manual review) → REVIEWING (Review in progress) → APPROVED (Review passed) → PUBLISHED (Published). Abnormal task statuses include: FAILED (Execution failed), RETRYING (Retrying in progress), FAILED_PERMANENT (Permanent failure (cannot be retried)), NEEDS_REVISION (Revision required), and REGENERATING (Regeneration). After a task is created, it goes through the following steps in sequence: data access, preprocessing, sample classification, job skill extraction, personality extraction, candidate merging, validation, draft generation, automatic evaluation, and manual review. Once approved, it is published; if it fails, it enters a retry or permanent failure state; if the review requires modifications, it enters a revision state and triggers regeneration.

[0024] Furthermore, the teacher profile creation module is used to create profiles for target teachers (the target personnel) to be integrated into the digital avatar system. The teacher profile includes at least: teacher identifier, teacher name, domain identifier, basic introduction, optional contact identifier or organization identifier, and currently enabled personal incremental package version information. The purpose of this module is to bind a unique entity for subsequent raw data access and distillation tasks. The raw data access module is used to receive raw data related to the teacher (the raw profile data). The raw data may include at least: personal profile information, FAQs (Frequently Asked Questions), documents, training minutes, case studies, chat logs, or message records. It is understood that this embodiment only processes work-related data regarding chat logs or message records to protect the privacy of the target personnel. After access, each piece of original data forms an original data object, recording at least: original data identifier, data type, title, content, time, tags, source channel, and optional speaker identifier. The output of this module is a set of original data, which serves as the input for the distillation task. The job skills extraction module extracts teachers' reusable professional abilities from job skills candidate samples, essentially structuring what they can do and how they do it. The personality extraction module extracts teachers' stable personality and behavioral characteristics from personality candidate samples, essentially structuring who they resemble and how they speak. The candidate merging module merges multiple candidate entries that may be generated from different data fragments. The verification module checks the merged draft. The draft generation module generates a draft object of a personal incremental package to be reviewed based on the merging and verification results. The automatic evaluation module verifies the usability of the draft using standard test questions. The review and release module allows teachers and administrators to review the personal incremental package draft. The runtime assembly module synthesizes the runtime control package when the digital clone is executed.

[0025] It should be noted that the preprocessing process for the original personnel files to obtain preprocessed files is as follows: The original personnel files are deduplicated, anonymized, and denoised to obtain denoised files; the denoised files are then segmented into segments, and the segmented segments are scored for quality. The segments are then sorted according to their quality scores from highest to lowest to obtain the preprocessed files. In other words, the preprocessing step involves cleaning and formatting the original data. The preprocessing steps include at least: deduplication, anonymization, segmentation into segments suitable for model processing, noise removal, and prioritizing the extraction of high-quality output segments from the teacher. Specifically, the original personnel files are sequentially deduplicated, anonymized and controlled for sensitive information, denoised, segmented into segments suitable for model processing, and their quality scored. The segments are then sorted from highest to lowest score, and high-quality output segments from the teacher are extracted first, ultimately resulting in the preprocessed files.

[0026] It should be further noted that the process of classifying the preprocessed archival data to obtain corresponding classification results is as follows: Each data segment in the preprocessed archival data is classified according to preset classification rules to determine the corresponding sample category and obtain the classification result. The sample categories include first candidate samples that only match job skills, second candidate samples that only match personality styles, third candidate samples that match both job skills and personality styles, and invalid samples that do not match either job skills or personality styles. That is, the sample classification step is used to determine which distillation chain a data segment is more suitable for, dividing the samples into the following categories: job skill candidate samples, personality candidate samples, samples that match both, and invalid samples. The classification result forms the sample classification object. Specifically, the data segments in the preprocessed archival data are divided into sample categories according to preset classification rules to obtain the classification result. The categories are divided into four types: first candidate samples used only for extracting job skills, second candidate samples used only for extracting personality styles, third candidate samples containing both types of information, and invalid samples with no utilization value. In this way, this embodiment sets the pre-obtaining of personnel distillation authorization as a pre-authorization verification condition, constraining the processing permissions of personal files from the process entry point, meeting data privacy compliance requirements, and eliminating compliance risks of unauthorized private parsing of personnel files; the original files first undergo preprocessing operations to reduce the contamination of digital clones by noise samples and reduce the impact of chatter, duplicate text, and invalid fragments on subsequent distillation; after preprocessing, samples are classified to avoid distilling all materials together, thereby improving the subsequent extraction accuracy; through the distillation task state machine, access, preprocessing, classification, extraction, merging, verification, review, and release are linked into a complete technical process, and through the review and release module, teachers and administrators can participate in reviewing, returning, revising, and releasing, making the generation process of personalized capabilities reviewable, traceable, modifiable, and versioned, rather than a one-time black-box output of the model, reducing the risk of black-box generation, and improving the accuracy, controllability, and maintainability of personalized digital clones.

[0027] Step S12: Based on the classification results, extract the work skills and personality styles of the target personnel from the preprocessed archive data, and sort and merge them to obtain merged work skill drafts and personality style drafts.

[0028] In this embodiment, based on the classification results of the data fragments, the work skills and personality style information of the target personnel are extracted from the preprocessed archive data by the work skills extraction module and the personality extraction module, respectively. The information is then summarized and organized to generate a work skills draft and a personality style draft.

[0029] It should be noted that the process for extracting the target personnel's work skills and personality styles from the preprocessed archival data based on the classification results is as follows: Work skills of the target personnel are extracted from the first and third candidate samples to obtain several candidate work skills and corresponding confidence scores; personality styles of the target personnel are extracted from the second and third candidate samples to obtain several candidate personality styles and corresponding confidence scores. That is, as... Figure 3 The diagram shows a structural schematic of a personal incremental package provided in this application. The personal incremental package consists of two parts: a job skills coverage package, representing what the teacher can do and how they do it; and a personality coverage package, representing who the teacher resembles and how they speak. These two parts are extracted separately during distillation and jointly influence the digital avatar result during runtime. The job skills extraction module is used to extract the teacher's reusable professional abilities, i.e., job skills, from the job skills candidate samples including the first candidate sample and the third candidate sample. The extraction results, including several job skills candidate items and corresponding confidence scores, form the job skills coverage package candidate objects. The job skills coverage package includes at least: a capability range: used to describe in which tasks the teacher has stable experience; a methodological framework: used to describe the teacher's professional judgment framework; a process template: used to describe the teacher's step sequence for handling a certain type of task; experience rules: used to describe the teacher's commonly used heuristic judgment rules; prohibited modes: used to describe practices the teacher explicitly does not use; and a case index: used to reference reusable cases. The personality extraction module is used to extract the teacher's stable personality and behavioral characteristics from the personality candidate samples including the second candidate sample and the third candidate sample. The extracted personality style candidate items and corresponding confidence scores form the candidate personality coverage package. The personality coverage package adopts a five-layer structure: hard rules: inviolable boundaries; identity: how teachers view their role; expression style: tone, organization, and preferred expression; decision-making pattern: typical judgment order and trade-off preferences; interpersonal behavior: behavioral styles such as questioning, correcting, and encouraging others.

[0030] It should be further noted that the processing flow for sorting and merging to obtain the merged job skills draft and personality style draft is as follows: Identify candidate item pairs with semantic similarity higher than a preset similarity threshold among the target candidate items; merge the candidate item pairs among the target candidate items, remove candidate items with confidence scores lower than a first preset confidence score threshold, and perform deduplication to obtain the sorted target candidate items; detect conflicting candidate items in the sorted target candidate items and mark the conflicting candidate items to obtain the merged target draft; wherein, if the target candidate items are the several job skills candidate items, then the target draft is the job skills draft; if the target candidate items are the several personality style candidate items, then the target draft is the personality style draft. That is, since different data fragments may generate multiple candidate items, the candidate merging module is responsible for merging semantically similar items; deleting duplicate items; retaining items with higher confidence and more specific details; and marking potential conflicting items. The merged results form the job skills draft and personality style draft, respectively. Specifically, for candidate items in work skills or personality styles, the process first filters and merges groups of items with semantic similarity exceeding a threshold, then removes low-confidence items with confidence levels below a specified threshold and performs deduplication, resulting in a refined set of target candidate items. Next, it identifies and marks conflicting items with logical contradictions, ultimately generating drafts for work skills and personality styles respectively. In this way, this embodiment stably and repeatedly extracts runnable, reusable, and auditable personalized incremental capabilities from multi-source heterogeneous raw data, rather than simply generating one-time prompts or text summaries, improving the auditability and maintainability of personalized capabilities. It separates and represents the teacher's professional abilities and personalized styles, forming structured objects that can be directly invoked in subsequent execution chains, improving structural clarity and thus enhancing the accuracy and controllability of personalized digital avatar assembly. During the distillation process, sample classification and merging reduce the interference of noisy materials, incidental expressions, conflicting materials, and outdated data on the final result, avoiding the direct solidification of single contexts, accidental emotions, or low-quality chat fragments into personal digital avatar capabilities.

[0031] Step S13: Verify and check the job skills draft and the personality style draft to obtain the corresponding verification results, and generate a personal incremental package draft to be reviewed in order to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft and the personality style draft.

[0032] In this embodiment, the work skills draft and the personality style draft are verified separately and the verification results are obtained. The two drafts, together with the verification results, are packaged into a personal incremental package draft to be reviewed, thus completing the data distillation operation of the target personnel.

[0033] It should be noted that the processing flow for verifying and checking the work skill draft and the personality style draft to obtain the corresponding verification results is as follows: Detecting conflict items in the target draft to obtain a conflict item list; the target draft includes the work skill draft and the personality style draft; the conflict items include internal contradictions within the target draft and conflicts between the target draft and the preset system common capability package and the preset business domain capability package respectively; filtering out low-confidence items in the target draft whose confidence scores are lower than a second preset confidence score threshold to obtain a low-confidence item list; filtering out risk items in the target draft that meet preset violation conditions to obtain a verification result including the conflict item list and the low-confidence item list. That is, the verification module is used to check the merged drafts. The check content includes at least: whether it conflicts with public rules; whether it conflicts with domain capability packages; whether there are internal contradictions; whether there are insufficient evidence; and whether there is overstepping boundaries or inappropriate content for publication. The verification module outputs a verification result object, which includes at least: the overall verification result; a list of conflicting items; a list of low-confidence items; and modification suggestions. Specifically, it examines the work skill draft and the personality style draft, identifies self-contradictory content within the drafts, and incompatible content between the drafts and system common capability packages or business domain capability packages, forming a list of conflicting items. It then filters out items with a built-in confidence level below a second threshold to generate a list of low-confidence items, filters out risky content in the drafts that meets the violation standards, and integrates the list of conflicting items and the list of low-confidence items to form a complete verification result.

[0034] It should be further noted that before generating the draft of the personal incremental package to be reviewed, an automatic draft evaluation can be performed. The process is as follows: The usability of the work skills draft and the personality style draft is verified using preset test questions to obtain corresponding evaluation results. Correspondingly, the process for generating the draft of the personal incremental package to be reviewed is as follows: Based on the evaluation results, the verification results, the work skills draft, and the personality style draft, a draft of the personal incremental package to be reviewed is generated. That is, the automatic evaluation module is used to verify the usability of the draft using standard test questions. The evaluation content includes at least: whether it reflects the teacher's style; whether it retains professional competence; whether it conflicts with public rules; and whether it has incremental differentiation from domain standard competence. The automatic evaluation results serve as auxiliary input for the review and release module. Correspondingly, the draft generation module generates a draft object of the personal incremental package to be reviewed based on the merging results and verification results. This draft object includes at least: a work skills draft, a personality draft, verification results, and optional automatic evaluation results. The draft object is not the final runtime package, but rather an intermediate object for review and revision. In this way, this embodiment reduces the impact of noise, conflicts, occasional contexts, and low-quality samples on distillation results through sample classification, candidate merging, and verification modules, thereby improving the efficiency of data-to-operational capability conversion. It generates structured, runtime-configurable personal incremental packages, which are more suitable for enterprise-level digital avatar systems. Before formal packaging, a pre-assessment operation is introduced to verify usability through preset test questions. Standardized test cases simulate digital avatar call scenarios, verifying in advance whether skills and personality traits can correctly match the interaction logic, avoiding the hidden problem of complete draft content but actual operational failure. The evaluation results, verification results, and two feature drafts are uniformly packaged into a personal incremental package draft. Reviewers can simultaneously view three types of assessment information: data conflicts, low confidence risk, and scenario usability, improving manual review efficiency and reducing the probability of abnormal operation of the digital avatar after deployment.

[0035] Step S14: Provide feedback on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, then construct and store the target personal incremental package based on the draft of the personal incremental package so that when the enterprise-level digital clone system is running, the runtime control package can be constructed using the target personal incremental package, the preset system common capability package and the preset business domain capability package, and the enterprise business services can be provided based on the runtime control package.

[0036] In this embodiment, the draft of the personal incremental package undergoes manual feedback review to obtain a review result. Upon approval, the draft personal incremental package is used to generate and persistently store the final target personal incremental package. This is so that during the operation of the digital clone system, the target personal incremental package, the preset system common capability package, and the preset business domain capability package are integrated to assemble a runtime control package, which is then used to deliver enterprise business services externally. Figure 4The diagram illustrates a runtime assembly mechanism provided in this application, showing how the personal incremental package functions during runtime. The personal incremental package is not an independent execution unit, but rather is combined with the public capability package and domain capability package by the runtime assembly module to form a runtime control package, which then controls knowledge retrieval, expert execution, presentation, and risk control.

[0037] Specifically, the review and release module is used to allow teachers and administrators to review draft personal incremental packages. The review and release module supports at least the following operations: approve; reject; request modification; partial editing; release as official version; rollback to previous version. Once the draft is approved, an official personal incremental package object is generated and enters the release state. The final released personal incremental package object includes at least: teacher identifier; domain identifier; version identifier; job skill coverage package; personality coverage package; source snapshot information; confidence score or quality rating; and release time. The personal incremental package object is stored as structured data, not just as a piece of text. The runtime assembly module, when the digital avatar executes, reads: common capability package; domain capability package; personal incremental package; and synthesizes them into a runtime control package. The role of the personal incremental package at runtime includes influencing the final system prompts; influencing the scope and filtering conditions of knowledge retrieval; influencing the organization and expression style of results; and influencing the interpretation of remote expert results. It should be noted that the personal incremental package is not an independently executing expert, but a personalized increment on the domain capability package. This allows the system to retain both standard domain capabilities and reflect the differences between individual teachers.

[0038] It is understood that this application is not limited to the above-described implementation methods, and several alternative solutions exist that can achieve the same purpose. Alternative solutions for original data access methods include: uploads by the teacher; synchronization from the knowledge base by the system; collection from messaging systems, email systems, document systems, and training systems; and import of historical data by the administrator. As long as the original data object is ultimately formed, it is considered an alternative solution. Alternative solutions for sample classification methods include: classification by a large model; initial screening by rules followed by model verification; and implementation by a supervised learning classifier. As long as the candidate samples for work skills and personality can be distinguished, it is considered an alternative solution. Alternative solutions for the structure of work skills and personality: the work skills coverage package can be further broken down into a method framework package, a process template package, and a case package; the personality coverage package can adopt a five-layer structure or an equivalent multi-layer structure, as long as it still covers the functions of boundaries, identity, expression, decision-making, and interpersonal behavior. Alternative solutions for verification methods include: rule-based verification; model-based verification; a hybrid rule + model verification; and single-round or multi-round verification. Alternatives to the review and release method: Review can be conducted by the teacher; it can be reviewed by the administrator; it can be reviewed jointly by multiple people; it supports partial editing before release, and also supports regeneration. Alternatives to the runtime assembly method: Assembly can be performed immediately upon request; pre-compilation can be performed during release; multiple versions can be assembled according to user, scenario, and domain. As long as the released personal incremental package can be used in combination with other capability layers, it is considered an alternative. In this way, this embodiment treats the released personal incremental package as a runtime assembleable object, combining it with the public capability package and domain capability package to form a runtime control package. This allows the personal incremental package to be dynamically assembled with the public capability package and domain capability package at runtime, thereby serving the digital avatar execution chain, uniformly controlling knowledge retrieval, execution strategies, output styles, and constraints, and improving the accuracy and controllability of personalized assembly for the digital avatar.

[0039] As can be seen from the above, in this embodiment of the application, under the premise of obtaining authorization for the distillation operation of the target personnel, the original file data is first preprocessed and the samples are classified. Based on the classification results, the work skills draft and personality style draft of the target personnel are extracted and integrated. After verifying the two drafts, a personal incremental package draft containing the verification results is generated to complete the data distillation of the target personnel. After the feedback review is approved, the target personal incremental package is generated and stored from the personal incremental package draft. When the enterprise-level digital avatar system runs, it integrates the incremental package, the preset system public capability package and the preset business domain capability package to assemble the runtime control package, and provides enterprise business services to the outside world based on the control package. In this way, through the above-described process of this application embodiment, preset authorization conditions for the distillation operation of target personnel are set as a pre-operation threshold to avoid data compliance risks caused by unauthorized collection and processing of personal files and to protect personnel privacy. The original files are preprocessed before sample classification, the original file format is standardized and invalid and redundant information is removed, so as to improve the recognition accuracy of subsequent skill and personality feature extraction. Work skills and personality styles are extracted and merged independently according to the classification results, so as to achieve decoupled storage of the two types of personnel features, which makes it easy for the digital clone to call the corresponding feature dimensions as needed. The two drafts are checked and the verification results are included in the incremental package draft to intercept problems such as incomplete feature extraction, logical conflicts, and information distortion in advance, reducing the cost of manual secondary correction. During the operation phase, the personal incremental package, system public capability package, and business domain capability package are integrated to assemble the runtime control package, decoupling and encapsulating the personalized characteristics, general basic capabilities, and industry business capabilities of personnel. It can flexibly adapt to different enterprise business scenarios to quickly generate digital clone operation carriers, support the implementation of various enterprise business services, and optimize the enterprise-level digital clone operation method to adapt to enterprise-level operation requirements and reduce the risk of black box generation.

[0040] Accordingly, see Figure 5 As shown in the illustration, this application also provides an enterprise-level digital avatar operating device based on personalized incremental updates, comprising: The sample classification module 11 is used to preprocess the original file data of the target personnel under the condition of meeting the preset authorization, so as to obtain the preprocessed file data, and to classify the preprocessed file data to obtain the corresponding classification results; the preset authorization condition is that the authorization of the target personnel for distillation operation has been obtained in advance. The sorting and merging module 12 is used to extract the work skills and personality styles of the target personnel from the preprocessed archive data according to the classification results, and sort and merge them to obtain the merged work skills draft and personality style draft. The draft generation module 13 is used to verify and check the job skills draft and the personality style draft to obtain the corresponding verification results, and generate a personal incremental package draft to be reviewed in order to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft and the personality style draft; The incremental package storage module 14 is used to provide feedback on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft of the personal incremental package. In order to build a runtime control package using the target personal incremental package, the preset system public capability package and the preset business domain capability package when the enterprise-level digital clone system is running, and to provide enterprise business services based on the runtime control package.

[0041] In some specific embodiments, the sample classification module 11 may specifically include: The data denoising unit is used to perform deduplication, desensitization, and denoising on the original archival data of the target personnel to obtain denoised archival data. The segment sorting unit is used to segment the denoised archival data into segments, score the quality of the segmented data segments, and sort the data segments in descending order of quality score to obtain preprocessed archival data.

[0042] In some specific embodiments, the sample classification module 11 may specifically include: The sample classification unit is used to classify each data fragment in the preprocessed archive data according to the preset classification rules to determine the corresponding sample category and obtain the classification result. The sample category includes a first candidate sample that only matches the job skills, a second candidate sample that only matches the personality style, a third candidate sample that matches both the job skills and the personality style, and an invalid sample that does not match either the job skills or the personality style.

[0043] In some specific embodiments, the sorting and merging module 12 may specifically include: The skill extraction unit is used to extract the work skills of the target personnel from the first candidate sample and the third candidate sample to obtain several candidate items of work skills and corresponding confidence scores; The style extraction unit is used to extract the personality style of the target person from the second candidate sample and the third candidate sample to obtain several personality style candidate items and corresponding confidence scores.

[0044] In some specific embodiments, the sorting and merging module 12 may specifically include: The item pair determination unit is used to determine candidate item pairs among the target candidate items whose semantic similarity is higher than a preset similarity threshold; The item deduplication unit is used to merge the candidate item pairs in the target candidate items, remove the candidate items whose confidence scores are lower than the first preset confidence score threshold, and perform deduplication processing to obtain the sorted target candidate items. A conflict marking unit is used to detect conflict candidate entries in the sorted target candidate entries and mark the conflict candidate entries to obtain the merged target draft. Wherein, if the target candidate item is one of the several job skill candidate items, then the target draft is the job skill draft; if the target candidate item is one of the several personality style candidate items, then the target draft is the personality style draft.

[0045] In some specific embodiments, the draft generation module 13 may specifically include: A conflict detection unit is used to detect conflict items in the target draft to obtain a list of conflict items; the target draft includes the work skills draft and the personality style draft; the conflict items include internal contradictions within the target draft and conflicts between the target draft and the preset system common capability package and the preset business domain capability package, respectively; The low confidence filtering unit is used to filter out low confidence items in the target draft whose confidence scores are lower than a second preset confidence score threshold, so as to obtain a list of low confidence items. The risk item filtering unit is used to filter risk items in the target draft that meet preset violation conditions to obtain a verification result including the conflict item list and the low confidence item list.

[0046] In some specific embodiments, the enterprise-level digital avatar operating device based on personalized incremental changes may further include: The usability verification unit is used to verify the usability of the job skills draft and the personality style draft using preset test questions, so as to obtain the corresponding evaluation results. Accordingly, the draft generation module 13 may specifically include: The draft generation unit is used to generate a draft of the personal incremental package to be reviewed based on the evaluation results, the verification results, the work skills draft, and the personality style draft.

[0047] Furthermore, embodiments of this application also disclose an electronic device, Figure 6This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the enterprise-level digital avatar operation method based on personalized incremental updates disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0048] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0049] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0050] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the personalized incremental enterprise-level digital avatar operation method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0051] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned disclosed enterprise-level digital avatar operation method based on personalized incremental changes. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0053] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0055] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0056] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for operating enterprise-level digital clones based on personalized incremental changes, characterized in that, include: Under the preset authorization conditions, the original file data of the target personnel is preprocessed to obtain preprocessed file data, and the preprocessed file data is sampled and classified to obtain the corresponding classification results; the preset authorization conditions are that the authorization of the target personnel for distillation operation has been obtained in advance. Based on the classification results, the work skills and personality styles of the target personnel are extracted from the preprocessed archives and then sorted and merged to obtain merged drafts of work skills and personality styles. The job skills draft and the personality style draft are verified and checked to obtain the corresponding verification results, and a personal incremental package draft to be reviewed is generated to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft, and the personality style draft; Feedback is provided on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft of the personal incremental package. When the enterprise-level digital clone system is running, the target personal incremental package, the preset system common capability package, and the preset business domain capability package are used to construct a runtime control package, and enterprise business services are provided based on the runtime control package.

2. The enterprise-level digital clone operation method based on personalized incremental updates according to claim 1, characterized in that, The preprocessing of the original personnel files to obtain preprocessed files includes: The original files of the target personnel are deduplicated, desensitized, and denoising to obtain the denoising files. The denoised archival data is segmented into segments, and the segmented segments are scored for quality. The segments are then sorted in descending order of quality score to obtain the preprocessed archival data.

3. The enterprise-level digital clone operation method based on personalized incremental updates according to claim 1, characterized in that, The step of classifying the preprocessed archival data to obtain corresponding classification results includes: According to the preset classification rules, each data fragment in the preprocessed archive data is classified into samples to determine the corresponding sample categories and obtain the classification results. The sample categories include first candidate samples that only match the job skills, second candidate samples that only match the personality style, third candidate samples that match both the job skills and personality style, and invalid samples that do not match either the job skills or personality style.

4. The enterprise-level digital clone operation method based on personalized incremental updates according to claim 3, characterized in that, The step of extracting the target personnel's work skills and personality styles from the preprocessed archival data based on the classification results includes: The job skills of the target personnel are extracted from the first candidate sample and the third candidate sample to obtain several job skill candidate items and corresponding confidence scores; The personality styles of the target person are extracted from the second and third candidate samples to obtain several personality style candidate items and corresponding confidence scores.

5. The enterprise-level digital clone operation method based on personalized incremental updates according to claim 4, characterized in that, The process of sorting and merging the materials to obtain merged drafts of job skills and personality styles includes: Identify candidate item pairs among the target candidate items whose semantic similarity is higher than a preset similarity threshold; The candidate item pairs in the target candidate items are merged, and the candidate items with confidence scores lower than the first preset confidence score threshold are removed, and deduplication is performed to obtain the sorted target candidate items. Detect conflicting candidate entries in the sorted target candidate entries and mark the conflicting candidate entries to obtain the merged target draft; Wherein, if the target candidate item is one of the several job skill candidate items, then the target draft is the job skill draft; if the target candidate item is one of the several personality style candidate items, then the target draft is the personality style draft.

6. The enterprise-level digital clone operation method based on personalized incremental updates according to claim 1, characterized in that, The verification and checking of the job skills draft and the personality style draft to obtain the corresponding verification results includes: Detect conflict items in the target draft to obtain a list of conflict items; the target draft includes the job skills draft and the personality style draft; the conflict items include internal contradictions within the target draft and conflicts between the target draft and the preset system common capability package and the preset business domain capability package, respectively; Filter out low-confidence items in the target draft whose confidence scores are lower than the second preset confidence score threshold to obtain a list of low-confidence items; Risk items in the target draft that meet preset violation conditions are filtered to obtain a verification result including the list of conflict items and the list of low-confidence items.

7. The enterprise-level digital cloning method based on personalized incremental operation according to any one of claims 1 to 6, characterized in that, Before generating the draft of the individual incremental package to be reviewed, the following steps are also included: The usability of the job skills draft and the personality style draft is verified by using preset test questions to obtain corresponding evaluation results; Accordingly, generating the draft of the individual incremental package to be reviewed includes: Based on the assessment results, the verification results, the job skills draft, and the personality style draft, a draft of the personal incremental package to be reviewed is generated.

8. An enterprise-level digital avatar operating device based on personalized incremental changes, characterized in that, include: The sample classification module is used to preprocess the original file data of the target personnel under the premise of meeting the preset authorization conditions to obtain the preprocessed file data, and to classify the preprocessed file data to obtain the corresponding classification results; the preset authorization conditions are that the target personnel have been granted authorization for distillation operation in advance. The sorting and merging module is used to extract the work skills and personality styles of the target personnel from the preprocessed archive data according to the classification results, and sort and merge them to obtain the merged work skills draft and personality style draft. The draft generation module is used to verify and check the job skills draft and the personality style draft to obtain the corresponding verification results, and generate a personal incremental package draft to be reviewed in order to complete the distillation operation for the target personnel; the personal incremental package draft includes the verification results, the job skills draft and the personality style draft; The incremental package storage module is used to provide feedback on the draft of the personal incremental package to obtain the corresponding review result. If the review result indicates that the review is passed, the target personal incremental package is constructed and stored based on the draft of the personal incremental package. In order to build a runtime control package using the target personal incremental package, the preset system public capability package and the preset business domain capability package when the enterprise-level digital clone system is running, and provide enterprise business services based on the runtime control package.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the enterprise-level digital clone operation method based on personalized incremental claims as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the enterprise-level digital clone operation method based on personalized incremental as described in any one of claims 1 to 7.

Images