Digital employee assisting method and system and electronic equipment

Through vector similarity calculation and large language model display, combined with local document preprocessing and permission verification, it solves the problem that employees in the field of high security find it difficult to master the workflow, and provides an adaptive digital employee assistance system to achieve fast and accurate Answers and information security.

CN118964550APending Publication Date: 2024-11-15STATE NUCLEAR POWER AUTOMATION SYST ENGCO

Patent Information

Application Number
CN202410955529.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In areas with high security requirements, it is difficult for a single employee to fully master all workflow procedures, or the learning cost is too high, making it difficult to provide answers and guidance for employees' specific application scenarios.

Method used

By obtaining the user's query vector data and the document data in the vector library, the trained large language model displays similar text blocks, and combining the preprocessing and permission verification of local documents, digital employee assistance methods and systems are provided. The adaptive device conditions automatically select the model to ensure information security and relevance of solutions.

Benefits of technology

It realizes a digital employee assistance system that is familiar with work procedures for employees in the field of high security, answers questions quickly and accurately, reduces learning and use costs, avoids the difficulty of information extraction caused by different file formats, and ensures information security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964550A_ABST
    Figure CN118964550A_ABST
Patent Text Reader

Abstract

The invention provides a digital employee assisting method and system and electronic equipment. The method comprises the steps of obtaining query vector data of a user; performing vector similarity calculation on the query vector data and document data stored in a vector library to obtain a plurality of similar text blocks and the similarity of each similar text block; and displaying the similar text blocks of which the similarity is greater than a preset threshold value through the trained large language model according to a preset prompt word. Through application of model tools such as a large language model, a digital employee auxiliary system which is familiar with related working programs and can perform solution and guidance for specific application scene problems of the employees is provided for the employees, and proper solution and guidance contents are extracted from numerous documents with different formats; a proper model is adopted for adaptive equipment conditions; due to local deployment, the leakage risk of data in the transmission process is avoided, and the information security is guaranteed; relevance and accuracy of answers are ensured, and convenience is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a digital employee assistance method, system and electronic device. Background Art

[0002] In areas with high security requirements, such as nuclear power, information security, complex work processes, large amounts of data, and regular revisions and iterations are common, which makes it difficult for a single employee to fully master all work processes or the learning cost is too high. With the maturity of related technologies such as large language models, the actual business processes in these areas require a "digital employee" who can adapt to the above characteristics, be familiar with relevant work procedures, and be able to answer and guide employees' specific application scenario questions. Summary of the invention

[0003] The technical problem to be solved by the present disclosure is to overcome the defects in the prior art that it is difficult for a single employee to fully master all work flow procedures in fields with high security requirements, or the learning cost is too high, and to provide a digital employee assistance method, system and electronic device.

[0004] The present invention solves the above technical problems through the following technical solutions:

[0005] In a first aspect, the present disclosure provides a digital employee assistance method, the digital employee assistance method comprising:

[0006] Obtain the user's query vector data;

[0007] Performing vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each of the similar text blocks;

[0008] According to the preset prompt words, the similar text blocks with similarity greater than a preset threshold are displayed through the trained large language model.

[0009] Optionally, obtaining the user's query vector data includes:

[0010] Get the user's query data;

[0011] The query data is input into the trained first word embedding model to obtain the query vector data.

[0012] Optionally, the digital employee assistance method further includes:

[0013] Extract text content from local documents to be stored;

[0014] Input the text content into a trained word segmentation model to obtain a text block;

[0015] Input the text block into the trained second word embedding model to obtain a text block vector;

[0016] The text block vectors are divided into individual documents and stored in the vector library as document data.

[0017] Optionally, the large language model includes one or more of Xinference, Ollama, and GPT-4;

[0018] The word segmentation model includes a word segmenter supporting one or more languages;

[0019] The first word embedding model includes one or more of Word2Vec, GloVe and FastText;

[0020] The second word embedding model includes one or more of Word2Vec, GloVe and FastText;

[0021] The vector library includes one or more of Milvus, Faiss and Annoy.

[0022] Optionally, the digital employee assistance method further includes:

[0023] Get the device parameters of the currently running device;

[0024] Determine the target large language model to be used for operation according to the device parameters; and / or, determine the target first word embedding model to be used for operation according to the device parameters; and / or, determine the target second word embedding model to be used for operation according to the device parameters.

[0025] Optionally, the digital employee assistance method further includes:

[0026] The target word segmentation model to be run is determined according to the language of the text content.

[0027] Optionally, extracting text content from the local document to be stored includes:

[0028] Preprocessing the local document to complete text cleaning, format standardization and error correction;

[0029] The text content is extracted from the preprocessed local document through a document loader.

[0030] Optionally, the digital employee assistance method further includes:

[0031] In response to the user's preview operation on the displayed similar text block, determining source document data associated with the target similar text block to be previewed, and completely displaying the source document data;

[0032] and / or,

[0033] The digital employee assistance method is applied in an enterprise local area network, the vector library is isolated from the Internet, and the digital employee assistance method further includes:

[0034] The authority of the user is verified, and when the verification is passed, the user is allowed to access the vector library.

[0035] In a second aspect, the present disclosure provides a digital employee assistance system, the digital employee assistance system comprising: an acquisition module, a similarity calculation module and a prompt display module;

[0036] The acquisition module is used to acquire the user's query vector data;

[0037] The similarity calculation module is used to perform vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each of the similar text blocks;

[0038] The prompt display module is used to display the similar text blocks whose similarity is greater than a preset threshold through a trained large language model according to the preset prompt words.

[0039] Optionally, the acquisition module is also used to acquire the user's query data;

[0040] The acquisition module is also used to input the query data into a trained first word embedding model to obtain the query vector data.

[0041] Optionally, the digital employee assistance system further comprises: a text extraction module, a word segmentation module, a text block vector extraction module and a storage module;

[0042] The text extraction module is used to extract text content from the local document to be stored;

[0043] The word segmentation module is used to input the text content into a trained word segmentation model to obtain a text block;

[0044] The text block vector extraction module is used to input the text block into the trained second word embedding model to obtain a text block vector;

[0045] The storage module is used to divide the text block vector into individual documents as document data and store them in the vector library.

[0046] Optionally, the large language model includes one or more of Xinference, Ollama, and GPT-4;

[0047] The word segmentation model includes a word segmenter supporting one or more languages;

[0048] The first word embedding model includes one or more of Word2Vec, GloVe and FastText;

[0049] The second word embedding model includes one or more of Word2Vec, GloVe and FastText;

[0050] The vector library includes one or more of Milvus, Faiss and Annoy.

[0051] Optionally, the digital employee assistance system further comprises: a model determination module;

[0052] The acquisition module is also used to obtain device parameters of the currently running device;

[0053] The model determination module is used to determine the target large language model to be run according to the device parameters; and / or, the model determination module is used to determine the target first word embedding model to be run according to the device parameters; and / or, the model determination module is used to determine the target second word embedding model to be run according to the device parameters.

[0054] Optionally, the model determination module is further used to determine a target word segmentation model to be run according to a language of the text content.

[0055] Optionally, the text extraction module is further used to pre-process the local document to complete text cleaning, format standardization and error correction;

[0056] The text extraction module is also used to extract the text content from the pre-processed local document through a document loader.

[0057] Optionally, the digital employee assistance system further includes: a document preview module;

[0058] The document preview module is used to determine the source document data associated with the target similar text block to be previewed in response to the user's preview operation on the displayed similar text block, and to fully display the source document data;

[0059] and / or,

[0060] The digital employee assistance system is applied in the enterprise local area network, the vector library is isolated from the Internet, and the digital employee assistance system further includes: an authority verification module;

[0061] The authority verification module is used to verify the authority of the user, and when the verification is passed, the user is allowed to access the vector library.

[0062] In a third aspect, the present disclosure provides an electronic device, including a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein the processor implements the aforementioned digital employee assistance method when executing the computer program.

[0063] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present disclosure.

[0064] The positive and progressive effects of the present disclosure are: in the field of high security requirements, through the application of model tools such as large language models, a digital employee assistance system is provided for employees (users) who are familiar with relevant work procedures and can answer and guide employees' specific application scenario questions, so as to realize the extraction of appropriate answers and guidance content from a large number of documents in various file formats within the enterprise, avoiding the difficulty of information extraction caused by different file formats; adaptively adapting to the current device conditions, automatically selecting and adopting the appropriate model (uploading local documents, querying, etc.); local deployment, that is, deployment within the enterprise LAN, not connecting to the Internet, avoiding the risk of data leakage during transmission, and ensuring information security; automatically performing text cleaning, format standardization and error correction by pre-processing local documents to ensure the quality of document data in the vector library; accurately matching user questions with reference file paragraphs in the vector library to ensure the relevance and accuracy of the answers; providing a file preview function, so that users can quickly view the file content on which the answers are based without downloading the entire file. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flowchart of a digital employee assistance method provided in Embodiment 1 of the present disclosure;

[0066] Figure 2 A flowchart of a specific implementation of step S11 of a digital employee assistance method provided in Example 1 of the present disclosure;

[0067] Figure 3 A flowchart of a specific implementation of a digital employee assistance method provided in Example 1 of the present disclosure;

[0068] Figure 4 A flowchart of another specific implementation of a digital employee assistance method provided in Embodiment 1 of the present disclosure;

[0069] Figure 5 A flowchart of another specific implementation of a digital employee assistance method provided in Embodiment 1 of the present disclosure;

[0070] Figure 6 A flowchart of another specific implementation of a digital employee assistance method provided in Embodiment 1 of the present disclosure;

[0071] Figure 7 A flowchart of a specific implementation of step S21 of a digital employee assistance method provided in Example 1 of the present disclosure;

[0072] Figure 8 A question-and-answer example of a digital employee assistance method provided in Embodiment 1 of the present disclosure;

[0073] Fig. 9 A module schematic diagram of a digital employee assistance system provided in Embodiment 2 of the present disclosure;

[0074] Fig.10 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present disclosure. DETAILED DESCRIPTION

[0075] The present disclosure is further described below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0076] Prefixes such as "first" and "second" are used in the embodiments of the present disclosure only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers to distinguish description objects in the embodiments of the present disclosure does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and no unnecessary limitation should be constituted due to the use of such prefixes. In addition, in the description of the present embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0077] Example 1

[0078] Figure 1 A flowchart of a digital employee assistance method provided by an exemplary embodiment of the present disclosure, the digital employee assistance method comprising:

[0079] S11. Obtain the user's query vector data.

[0080] S12, performing vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each similar text block.

[0081] S13. According to the preset prompt words, similar text blocks with similarity greater than a preset threshold are displayed through the trained large language model.

[0082] The query vector data is query data in the form of vectors. The document data in the vector library is stored in the form of vectors.

[0083] The vector library can store all documents within the enterprise, covering knowledge in technology, processes, security and other aspects. The vector library is the knowledge base within the enterprise.

[0084] The large language model can support, including but not limited to, Xinference, Ollama, and GPT-4. It is an existing technology that the large language model displays content according to the prompt words, which will not be described here.

[0085] It is possible to conduct question-and-answer queries on document data in a database (such as a vector library), including periodically converting relational database content into local files and then parsing and storing them in a vector database. Alternatively, it is possible to combine the contextual understanding capabilities of a large language model to intelligently construct database queries based on the context of the user's query.

[0086] By using large language models to answer employees' questions, we can quickly and accurately answer employees' questions, effectively reducing the time and errors of manual queries.

[0087] An approachable front-end answering interface can be used, that is, different digital employee avatars, as well as colloquial self-introductions and answering methods can be designed and displayed according to different user scenarios to improve the affinity between the employee and the questioner during the answering process.

[0088] In this embodiment, in the field of high security requirements, through the application of model tools such as large language models, a digital employee assistance system is provided for employees (users) who are familiar with relevant work procedures and can answer and guide employees' specific application scenario questions, so as to achieve the extraction of appropriate answers and guidance content from a large number of documents in various file formats within the enterprise, avoiding the difficulty of information extraction caused by different file formats; adaptively adapting to the current device conditions, automatically selecting and adopting the appropriate model (uploading local documents, querying, etc.); local deployment, that is, deployment within the enterprise LAN, not connected to the Internet, avoiding the risk of data leakage during transmission, and ensuring information security; automatically performing text cleaning, format standardization and error correction by pre-processing local documents to ensure the quality of document data in the vector library; accurately matching user questions with reference file paragraphs in the vector library to ensure the relevance and accuracy of the answers; providing a file preview function, so that users can quickly view the file content based on the answer without downloading the entire file.

[0089] In one embodiment, referring to Figure 2 , step S11 comprises:

[0090] S111. Obtain user query data.

[0091] S112, input the query data into the trained first word embedding model to obtain query vector data.

[0092] Among them, the first word embedding model can support including but not limited to Word2Vec, GloVe and FastText.

[0093] In one embodiment, referring to Figure 3 , digital employee assistance methods also include:

[0094] S21. Extract text content from the local document to be stored.

[0095] S22. Input the text content into the trained word segmentation model to obtain a text block.

[0096] S23. Input the text block into the trained second word embedding model to obtain a text block vector.

[0097] S24. Divide the text block vector into individual documents as document data and store them in a vector library.

[0098] Among them, local documents can support file formats including but not limited to HTML, MHTML, Markdown, JSON, CSV, PDF, Word, PowerPoint, PNG, JPEG, BMP, TXT, Excel, etc., and can extract text content from local documents of different file formats. The text content is separated from the file format of the original document to a certain extent for centralized management, storage and processing, and can effectively integrate multi-source data and improve data utilization.

[0099] In the case where it is necessary to retain the file format in the vector library, for example, when the text contents are related to each other and some formats need to be retained (such as tables), the document data can also support file formats including but not limited to HTML, MHTML, Markdown, JSON, CSV, PDF, Word, PowerPoint, PNG, JPEG, BMP, TXT, Excel, etc.

[0100] The word segmentation model can support word segmentation models (word segmenters) including but not limited to Chinese, English, German, etc.

[0101] The second word embedding model may support including but not limited to Word2Vec, GloVe, and FastText.

[0102] In one embodiment, the large language model includes one or more of Xinference, Ollama, and GPT-4.

[0103] A word segmentation model includes word segmenters that support one or more languages.

[0104] The first word embedding model includes one or more of Word2Vec, GloVe, and FastText.

[0105] The second word embedding model includes one or more of Word2Vec, GloVe, and FastText.

[0106] The vector library includes one or more of Milvus, Faiss and Annoy.

[0107] In one embodiment, referring to Figure 4 , digital employee assistance methods also include:

[0108] S31. Obtain device parameters of the currently running device.

[0109] S32. Determine the target large language model to be run according to the device parameters.

[0110] In this embodiment, it is able to adapt to user needs and hardware conditions and automatically select the most suitable large language model. If the device parameters (hardware conditions) permit, the large language model that requires the most hardware resources can be used as the target large language model without the need to purchase expensive hardware equipment. At the same time, it also has the ability to dynamically load new models and can be updated to the latest large language model in real time without restarting or reconfiguring. It can be compatible with multiple large language model frameworks through a set of common interfaces, support multiple word embedding models and word segmentation models and multiple vector databases, greatly improving the availability and compatibility of the system, automatically selecting and updating the most suitable large language model, without the need for users to manually select and update, and greatly reducing the cost of learning and use.

[0111] In one embodiment, referring to Figure 5 , digital employee assistance methods also include:

[0112] S31. Obtain device parameters of the currently running device.

[0113] S33. Determine the target first word embedding model to be run according to the device parameters.

[0114] In this embodiment, it is able to adapt to user needs and hardware conditions and automatically select the most appropriate first word embedding model. When device parameters (hardware conditions) permit, the first word embedding model that requires the most hardware resources can be used as the target first word embedding model without the need to purchase expensive hardware equipment. It also has the ability to dynamically load new models and can be updated to the latest first word embedding model in real time without restarting or reconfiguring. It can be compatible with a variety of large language model frameworks and support a variety of word embedding models through a set of common interfaces, greatly improving the availability and compatibility of the system, automatically selecting and updating the most appropriate first word embedding model without the need for users to manually select and update, greatly reducing the cost of learning and use.

[0115] In one embodiment, referring to Figure 6 , digital employee assistance methods also include:

[0116] S31. Obtain device parameters of the currently running device.

[0117] S34. Determine the target second word embedding model to be run according to the device parameters.

[0118] In this embodiment, it can adapt to user needs and hardware conditions and automatically select the most appropriate second word embedding model. When device parameters (hardware conditions) permit, the second word embedding model that requires the most hardware resources can be used as the target second word embedding model, without the need to purchase expensive hardware equipment. It also has the ability to dynamically load new models and can be updated to the latest second word embedding model in real time without restarting or reconfiguring. It can be compatible with multiple large language model frameworks and support multiple word embedding models through a set of common interfaces, greatly improving the availability and compatibility of the system, automatically selecting and updating the most appropriate second word embedding model, without the need for users to manually select and update, greatly reducing the cost of learning and use.

[0119] In one embodiment, the digital employee assistance method further includes:

[0120] The target word segmentation model to be run is determined based on the language of the text content.

[0121] In this embodiment, it can adapt to user needs and automatically select the most appropriate word segmentation model; it also has the ability to dynamically load new models, and can be updated to the latest word segmentation model in real time without restarting or reconfiguring. It can be compatible with multiple large language model frameworks through a set of common interfaces and support multiple word segmentation models, which greatly improves the availability and compatibility of the system, automatically selects and updates the most appropriate word segmentation model, and does not require users to manually select and update, which greatly reduces the cost of learning and use.

[0122] In one embodiment, referring to Figure 7 , step S21 comprises:

[0123] S211. Preprocess local documents to complete text cleaning, format standardization and error correction.

[0124] S212. Extracting text content from the preprocessed local document through a document loader.

[0125] In this embodiment, by preprocessing local documents, text cleaning, format standardization and error correction are automatically performed to ensure the quality of the knowledge base, so that user questions can be more accurately matched with reference document paragraphs in the vector library (knowledge base) to ensure the relevance and accuracy of the answers.

[0126] In one embodiment, the digital employee assistance method further includes:

[0127] In response to a user's preview operation on the displayed similar text block, source document data associated with the target similar text block to be previewed is determined, and the source document data is fully displayed.

[0128] In this embodiment, a file preview function is provided so that the user can quickly view the content of the file on which the answer is based without downloading the entire file.

[0129] In one embodiment, the digital employee assistance method is applied in an enterprise local area network, and the vector library is isolated from the Internet. The digital employee assistance method further includes:

[0130] Verify the user's permissions and allow the user to access the vector library if the verification passes.

[0131] In this embodiment, the digital employee assistance method is applied in the enterprise LAN, and the system is completely run on the local device without connecting to an external server, thereby avoiding the risk of data leakage during transmission. All document data stored in the local vector library is encrypted, and only users verified by the system can decrypt and access the data. In addition, the system implements fine-grained access control and user authentication mechanisms, as well as a complete audit log function, to ensure high security requirements.

[0132] The following uses an example to illustrate the implementation of the digital employee assistance method and designs a system for implementing the digital employee assistance method.

[0133] Hardware configuration and model selection: First, the system will automatically select the most appropriate large language model based on user needs and hardware conditions. For example, if the user's hardware conditions are good, the system may select a larger model, such as GPT-4; if the hardware conditions are poor, the system may select a smaller model, such as GPT-2. In addition, the system also has the ability to dynamically load new models, and can be updated to the latest large language model in real time without restarting or reconfiguring.

[0134] Data Integration: The system supports a variety of local document formats, including but not limited to HTML, MHTML, Markdown, JSON, CSV, PDF, Word, PowerPoint, PNG, JPEG, BMP, TXT, Excel, etc. Users can upload these files to the system, which will automatically parse them and store the parsing results in the vector database. In addition, the system can also perform question-and-answer queries on database data, including regularly converting relational database content into local files and then parsing and storing them in the vector database, or the system can combine the contextual understanding capabilities of the large language model to intelligently construct database queries based on the context of the user's query.

[0135] Offline deployment and information security: The system is designed to run entirely on local devices without connecting to external servers, thus avoiding the risk of data leakage during transmission. All knowledge base data stored locally is encrypted, and only users verified by the system can decrypt and access the data. In addition, the system implements fine-grained access control and user authentication mechanisms, as well as complete audit log functions.

[0136] Knowledge question and answer optimization: The system can pre-process knowledge base files and automatically perform text cleaning, format standardization, and error correction to ensure the quality of the knowledge base. The system can accurately match user questions with reference document paragraphs in the knowledge base to ensure the relevance and accuracy of the answers. In addition, the system provides a file preview function, so users can quickly view the content of the file on which the answer is based without downloading the entire file.

[0137] Figure 8 A question and answer session is shown. The question asks which departments are in the company's organizational structure, and the system lists the names of the departments.

[0138] Example 2

[0139] Corresponding to the aforementioned digital employee assistance method embodiment, the present disclosure also provides an embodiment of a digital employee assistance system.

[0140] Fig. 9 A module diagram of a digital employee assistance system provided for an exemplary embodiment of the present disclosure, the system comprising: an acquisition module 1, a similarity calculation module 2 and a prompt display module 3.

[0141] The acquisition module 1 is used to obtain the user's query vector data.

[0142] The similarity calculation module 2 is used to perform vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each similar text block.

[0143] The prompt display module 3 is used to display similar text blocks with similarity greater than a preset threshold through a trained large language model according to preset prompt words.

[0144] In this embodiment, in the field of high security requirements, through the application of model tools such as large language models, a digital employee assistance system is provided for employees (users) who are familiar with relevant work procedures and can answer and guide employees' specific application scenario questions, so as to achieve the extraction of appropriate answers and guidance content from a large number of documents in various file formats within the enterprise, avoiding the difficulty of information extraction caused by different file formats; adaptively adapting to the current device conditions, automatically selecting and using appropriate models (uploading local documents, queries, etc.) and vector libraries (locally running vector libraries are required); local deployment, that is, deployment within the enterprise LAN, not connected to the Internet, avoiding the risk of data leakage during transmission, and ensuring information security; automatically performing text cleaning, format standardization and error correction by pre-processing local documents to ensure the quality of document data in the vector library; accurately matching user questions with reference file paragraphs in the vector library to ensure the relevance and accuracy of the answers; providing a file preview function, so that users can quickly view the file content based on the answer without downloading the entire file.

[0145] In one embodiment, the acquisition module 1 is also used to acquire the user's query data.

[0146] The acquisition module 1 is also used to input the query data into the trained first word embedding model to obtain query vector data.

[0147] In one embodiment, referring to Fig. 9 The digital employee assistance system also includes: a text extraction module 4, a word segmentation module 5, a text block vector extraction module 6 and a storage module 7.

[0148] The text extraction module 4 is used to extract text content from the local documents to be stored.

[0149] The word segmentation module 5 is used to input text content into a trained word segmentation model to obtain a text block.

[0150] The text block vector extraction module 6 is used to input the text block into the trained second word embedding model to obtain the text block vector.

[0151] The storage module 7 is used to divide the text block vectors into individual documents as document data and store them in a vector library.

[0152] In one embodiment, the large language model includes one or more of Xinference, Ollama, and GPT-4.

[0153] A word segmentation model includes word segmenters that support one or more languages.

[0154] The first word embedding model includes one or more of Word2Vec, GloVe, and FastText.

[0155] The second word embedding model includes one or more of Word2Vec, GloVe, and FastText.

[0156] The vector library includes one or more of Milvus, Faiss and Annoy.

[0157] In one embodiment, referring to Fig. 9 , the digital employee assistance system also includes: a model determination module 8.

[0158] The acquisition module 1 is also used to acquire the device parameters of the currently running device.

[0159] The model determination module 8 is used to determine the target large language model to be run according to the device parameters.

[0160] In one embodiment, referring to Fig. 9 , the digital employee assistance system also includes: a model determination module 8.

[0161] The acquisition module 1 is also used to acquire the device parameters of the currently running device.

[0162] The model determination module 8 is used to determine the target first word embedding model to be run according to the device parameters.

[0163] In one embodiment, referring to Fig. 9 , the digital employee assistance system also includes: a model determination module 8.

[0164] The acquisition module 1 is also used to acquire the device parameters of the currently running device.

[0165] The model determination module 8 is used to determine the target second word embedding model to be run according to the device parameters.

[0166] In one embodiment, the model determination module 8 is further used to determine the target word segmentation model to be run according to the language of the text content.

[0167] In one embodiment, the text extraction module 4 is also used to pre-process the local document to complete text cleaning, format standardization and error correction.

[0168] The text extraction module 4 is also used to extract text content from the pre-processed local document through the document loader.

[0169] In one embodiment, referring to Fig. 9 The digital employee assistance system also includes: a document preview module 9.

[0170] The document preview module 9 is used to respond to the user's preview operation on the displayed similar text block, determine the source document data associated with the target similar text block to be previewed, and fully display the source document data.

[0171] In one embodiment, the digital employee assistance system is applied in the enterprise local area network, and the vector library is isolated from the Internet. Fig. 9 The digital employee assistance system also includes: an authority verification module 10.

[0172] The authority verification module 10 is used to verify the authority of the user, and allows the user to access the vector library when the verification is passed.

[0173] As for the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The system embodiment described above is only illustrative, wherein the units described as separate components may or may not be physically separated, and the components as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution.

[0174] Example 3

[0175] Fig.10 This is a schematic diagram of the structure of an electronic device showing an example embodiment of the present disclosure, wherein the electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor, and when the processor executes the computer program, the digital employee assistance method described in any of the above embodiments is implemented. Fig.10 The electronic device 90 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0176] like Fig.10 As shown, the electronic device 90 may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device 90 may include, but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 connecting different system components (including the memory 92 and the processor 91).

[0177] The bus 93 includes a data bus, an address bus, and a control bus.

[0178] The memory 92 may include a volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922 , and may further include a read-only memory (ROM) 923 .

[0179] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) of program modules 924, such program modules 924 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0180] The processor 91 executes various functional applications and data processing by running the computer programs stored in the memory 92, such as the digital employee assistance method provided in any of the above embodiments.

[0181] The electronic device 90 may also communicate with one or more external devices 94 (e.g., keyboards, pointing devices, etc.). Such communication may be performed via an input / output (I / O) interface 95. Furthermore, the electronic device 90 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 96. As shown, the network adapter 96 communicates with other modules of the electronic device 90 via a bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0182] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0183] Although the specific embodiments of the present disclosure are described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, but these changes and modifications all fall within the protection scope of the present disclosure.

Claims

1. A digital employee assistance method, characterized in that: The digital employee assistance method includes: Obtain the user's query vector data; Performing vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each of the similar text blocks; According to the preset prompt words, the similar text blocks with similarity greater than a preset threshold are displayed through the trained large language model.

2. The digital employee assistance method according to claim 1, characterized in that: The obtaining of the user's query vector data includes: Get the user's query data; The query data is input into the trained first word embedding model to obtain the query vector data.

3. The digital employee assistance method according to claim 2, characterized in that: The digital employee assistance method further includes: Extract text content from local documents to be stored; Input the text content into a trained word segmentation model to obtain a text block; Input the text block into the trained second word embedding model to obtain a text block vector; The text block vectors are divided into individual documents and stored in the vector library as document data.

4. The digital employee assistance method according to claim 3, characterized in that: The large language model includes one or more of Xinference, Ollama, and GPT-4; The word segmentation model includes a word segmenter supporting one or more languages; The first word embedding model includes one or more of Word2Vec, GloVe and FastText; The second word embedding model includes one or more of Word2Vec, GloVe and FastText; The vector library includes one or more of Milvus, Faiss and Annoy.

5. The digital employee assistance method according to claim 4, characterized in that: The digital employee assistance method further includes: Get the device parameters of the currently running device; Determine the target large language model to be used for operation according to the device parameters; and / or, determine the target first word embedding model to be used for operation according to the device parameters; and / or, determine the target second word embedding model to be used for operation according to the device parameters.

6. The digital employee assistance method according to claim 4, characterized in that: The digital employee assistance method further includes: The target word segmentation model to be run is determined according to the language of the text content.

7. The digital employee assistance method according to claim 3, characterized in that: The extracting text content from the local document to be stored includes: Preprocessing the local document to complete text cleaning, format standardization and error correction; The text content is extracted from the preprocessed local document through a document loader.

8. The digital employee assistance method according to claim 1, characterized in that: The digital employee assistance method further includes: In response to the user's preview operation on the displayed similar text block, determining source document data associated with the target similar text block to be previewed, and completely displaying the source document data; and / or, The digital employee assistance method is applied in an enterprise local area network, the vector library is isolated from the Internet, and the digital employee assistance method further includes: The authority of the user is verified, and when the verification is passed, the user is allowed to access the vector library.

9. A digital employee assistance system, characterized in that: The digital employee assistance system comprises: an acquisition module, a similarity calculation module and a prompt display module; The acquisition module is used to acquire the user's query vector data; The similarity calculation module is used to perform vector similarity calculation on the query vector data and the document data stored in the vector library to obtain a number of similar text blocks and the similarity of each of the similar text blocks; The prompt display module is used to display the similar text blocks whose similarity is greater than a preset threshold through a trained large language model according to the preset prompt words.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and used to run on the processor, characterized in that: When the processor executes the computer program, the digital employee assistance method according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Management and evaluation of machine-learned models based on locally logged data

    CN109983483A

  • Knowledge base construction method and question and answer dialogue method and system based on generative large language model

    CN117056471A

  • Question and answer method and system based on LangChain and large language model

    CN117313859A

  • Multipath text recall method and system, electronic equipment and storage medium

    CN117421404A

Cited By

  • Personnel work assisting method and system based on AI assistant

    CN119962701A

  • Personnel work assistance method and system based on ai assistant

    CN119962701B

  • Retrieval system and method based on retrieval enhancement generation

    CN121434374A