A knowledge engineering method and device based on collaborative processing

By collecting and classifying materials, using machine learning and knowledge graphs to extract knowledge automatically, and using Retro-Reader model to convert implicit knowledge, the problem of difficult classification of knowledge sources and difficulty in multi-site collaboration in knowledge engineering is solved, and the problem of improving the problem of "unparalleled" knowledge sources in knowledge engineering is improved. Knowledge quality and management efficiency.

CN112712177BActive Publication Date: 2024-12-03SHANGHAI YONGXIAO INTELLIGENT TECH CO LTD
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202011598175.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-29
Publication Date
2024-12-03
Estimated Expiration
2040-12-29

AI Technical Summary

Technical Problem

In the existing knowledge engineering methods based on collaborative processing, the knowledge sources are difficult to classify, and the knowledge engineers have different backgrounds, resulting in poor knowledge quality, difficulty in collaboration in many places, lack of verification mechanisms and annotation processes, which leads to difficulty in knowledge management and implicit knowledge transformation into There is a lack of quantitative indicators when explicit knowledge.

Method used

By collecting, managing and classifying materials, using machine learning algorithms and knowledge graphs to automatically extract knowledge, assign tasks according to the background of knowledge engineers, and using the Retro-Reader model to automatically transform and verify implicit knowledge.

Benefits of technology

It realizes efficient management of knowledge sources, improves knowledge quality, simplifies collaboration in multiple places, ensures the accuracy and quantifiable monitoring of knowledge, and solves the problem of implicit to explicit transformation in the process of knowledge transformation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112712177B_ABST
    Figure CN112712177B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of collaborative processing technology, and discloses a knowledge engineering method and device based on collaborative processing, including the following steps: S1: collect and manage materials and classify them, so as to obtain quantifiable knowledge sources and efficiently manage materials; S2: automatically assign tasks of the same level according to the background of knowledge engineers, and knowledge engineers of the same level collaborate to complete the corresponding tasks in task management; S3: automatically extract knowledge and convert implicit knowledge into explicit knowledge. In the present invention, by adding material management, task management implementation and knowledge extraction implementation, it effectively solves the problems that the source of knowledge is difficult to classify, the background of knowledge engineers is different, resulting in poor knowledge quality, and the difficulty of collaboration in multiple places. There is no verification mechanism, which also leads to poor knowledge quality, difficult knowledge management, and the problem of converting knowledge from implicit to explicit in knowledge conversion, and there is no labeled process and quantitative indicators to monitor.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of collaborative processing, and in particular to a knowledge engineering method and device based on collaborative processing. Background Art

[0002] Coprocessing occurs in distributed computing systems where two or more computers share the processing of a program or computing task. Coprocessing requires a complex program that can handle distributed loads, shared data files, and memory contention over a network while maintaining information synchronization security and accuracy.

[0003] The existing knowledge engineering methods based on collaborative processing have the following problems in use:

[0004] 1. The sources of knowledge are difficult to categorize.

[0005] 2. Knowledge engineers have different backgrounds, resulting in poor knowledge quality.

[0006] 3. Collaboration among multiple locations is fraught with difficulties.

[0007] 4. No verification mechanism will also lead to poor knowledge quality and difficult knowledge management.

[0008] 5. There is a problem of knowledge transformation from implicit to explicit, and there is no annotated process and quantitative indicators to monitor. To this end, we propose a knowledge engineering method and device based on collaborative processing. Summary of the invention

[0009] 1. Technical issues to be resolved

[0010] In view of the deficiencies of the prior art, the present invention provides a knowledge engineering method and device based on collaborative processing.

[0011] (II) Technical solution

[0012] To achieve the above object, the present invention provides the following technical solution: a knowledge engineering method based on collaborative processing, comprising the following steps:

[0013] S1: Collect, manage and classify materials, and obtain quantifiable knowledge sources to efficiently manage materials;

[0014] S2: Automatically assign tasks of the same level based on the background of knowledge engineers, and knowledge engineers of the same level collaborate to complete the corresponding tasks within task management;

[0015] S3: Automatically extract knowledge and convert implicit knowledge into explicit knowledge.

[0016] As a preference, the S1: collecting, managing and classifying materials to obtain quantifiable knowledge sources, the specific steps of efficiently managing materials are:

[0017] a). Material collection;

[0018] Sources of material collection: government systems, books, government documents, public accounts, and government official websites;

[0019] b). Material import;

[0020] Import relevant original materials into the material library;

[0021] c). Material library;

[0022] The material library is a collection for storing original materials and splitting materials;

[0023] d). Material classification;

[0024] e). Element extraction;

[0025] i. Business word extraction;

[0026] 1). Entity extraction, provided by the knowledge graph;

[0027] 2). Business word extraction, obtained by matching the accumulated knowledge word library;

[0028] 3). Relationship extraction, provided by the knowledge graph.

[0029] Preferably, the material classification in d) is as follows:

[0030] i. Use machine learning algorithm models to classify materials;

[0031] ii. Model 1 used is: a pre-trained model based on Bert;

[0032] 1). Model 1 is for existing classifications, but there are still new classifications and unknown classifications. In this case, Model 2 is used;

[0033] 2). Input: material text;

[0034] 3). Output: material classification;

[0035] iii. Model 2: Bert-based clustering model;

[0036] 1). Input: material text;

[0037] 2). Output: clustering results;

[0038] 3). It is necessary to manually analyze the clustering results and make a type judgment based on the results;

[0039] 4). When the confidence exceeds the threshold, it can be converted to model 1.

[0040] Preferably, the step S2: automatically assigning tasks of the same level according to the background of the knowledge engineers, and the knowledge engineers of the same level collaborate to complete the corresponding tasks in the task management, specifically includes:

[0041] a) Allocation

[0042] i. When multiple people and multiple locations collaborate, the entire task is performed;

[0043] ii. Created by the knowledge manager and assigned corresponding tasks;

[0044] iii. Principles for creating tasks;

[0045] 1). The creation of the same material category is considered as one task;

[0046] 2). The creation of the same knowledge engineering processing material is a task;

[0047] 3). The creation of materials of the same reviewer is considered as one task;

[0048] 4. The creation of a co-located knowledge project is a task;

[0049] b) Collaboration

[0050] i. Multiple people work together on the same task;

[0051] ii. Standard setting;

[0052] 1).Material name;

[0053] 2). Material answers;

[0054] 3). Source of knowledge;

[0055] 4). Source description;

[0056] 5). Links;

[0057] 6). Belonging knowledge graph, knowledge base;

[0058] 7).Instance name;

[0059] 8). Is it universal?

[0060] 9). Material status;

[0061] 10). Editor;

[0062] The above needs to be filled in by the knowledge engineer and submitted accurately for review by the knowledge supervisor. If it passes the review, it will enter the next process. If it fails the review, it will be redone until it passes the review;

[0063] c) Progress

[0064] i. Focus on the knowledge engineer and the entire task progress;

[0065] ii. When the progress of the knowledge engineer seriously slows down the entire task, the knowledge manager will confirm the specific problem and split the related work among other members of the team;

[0066] iii. Knowledge managers are responsible for their respective tasks;

[0067] iv. Knowledge engineers are responsible for the knowledge they create;

[0068] v.KPI is borne by both parties.

[0069] Preferably, the specific steps of S3: automatically extracting knowledge and converting implicit knowledge into explicit knowledge are as follows:

[0070] a). Knowledge management is the core of this system, and its main function is to automatically extract knowledge and verify knowledge;

[0071] b). Knowledge extraction;

[0072] i. Extract relevant questions and answers;

[0073] ii. Model: Retro-Reader model (retrospective reader);

[0074] iii. Principle;

[0075] iv. Retro-Reader·The model includes a sketchy-reading module and an intensive-reading-module to simulate human reading. The sketchy-reading module is used to read the article and questions and get a preliminary judgment; the intensive-reading module verifies the answerability and gives candidates. The outputs of the two modules are aggregated to make the final decision.

[0076] A knowledge engineering device based on collaborative processing includes a task storage module, a knowledge storage module and a knowledge application module.

[0077] The task storage module is specifically:

[0078] Mainly stored in relational databases, it stores all task-related information, and all stored information can be traced back to its source.

[0079] The knowledge storage module is specifically:

[0080] It is mainly stored in the knowledge graph. The final form of knowledge, symbolic representation, and vectorized representation are all stored in the knowledge graph and processed uniformly by the knowledge graph.

[0081] The knowledge application module is specifically:

[0082] a). Knowledge retrieval

[0083] i. Users can be divided into:

[0084] 1). External users: such as government officials, looking for related content;

[0085] 2). Actual users: business processing users, intelligent consultation;

[0086] 3). Knowledge Engineer: Find related and similar content.

[0087] ii. The main function of knowledge retrieval is also

[0088] b). Knowledge quiz.

[0089] i. Questions and answers can be provided for the relevant knowledge in the knowledge base, as well as intelligent consultation;

[0090] c). Data analysis,

[0091] i. Mainly used for statistical analysis of knowledge base related status;

[0092] ii. Quantify the efficiency of knowledge engineers;

[0093] iii. Quantify knowledge accuracy.

[0094] (III) Beneficial effects

[0095] Compared with the prior art, the present invention provides a knowledge engineering method and device based on collaborative processing, which has the following beneficial effects:

[0096] The knowledge engineering method and device based on collaborative processing, by adding material management, task management implementation and knowledge extraction implementation, effectively solves the problems of difficulty in classifying knowledge sources, poor knowledge quality caused by different backgrounds of knowledge engineers, difficulties in multi-site collaboration, poor knowledge quality caused by the lack of verification mechanism, difficulty in knowledge management, problems in knowledge transformation from implicit to explicit, and no labeled processes and quantitative indicators to monitor. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is the software framework diagram of the present invention. DETAILED DESCRIPTION

[0098] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0099] See also Figure 1 , a knowledge engineering method based on collaborative processing, comprising the following steps:

[0100] S1: Collect, manage and classify materials, and obtain quantifiable knowledge sources to efficiently manage materials;

[0101] S2: Automatically assign tasks of the same level based on the background of knowledge engineers, and knowledge engineers of the same level collaborate to complete the corresponding tasks within task management;

[0102] S3: Automatically extract knowledge and convert implicit knowledge into explicit knowledge.

[0103] S1: Collecting and managing materials and classifying them, obtaining quantifiable knowledge sources. The specific steps for efficient material management are:

[0104] a). Material collection;

[0105] Sources of material collection: government systems, books, government documents, public accounts, and government official websites;

[0106] b). Material import;

[0107] Import relevant original materials into the material library;

[0108] c). Material library;

[0109] The material library is a collection for storing original materials and splitting materials;

[0110] d). Material classification;

[0111] e). Element extraction;

[0112] i. Business word extraction;

[0113] 1). Entity extraction, provided by the knowledge graph;

[0114] 2). Business word extraction, obtained by matching the accumulated knowledge word library;

[0115] 3). Relationship extraction, provided by the knowledge graph.

[0116] The material classification d) is specifically:

[0117] i. Use machine learning algorithm models to classify materials;

[0118] ii. Model 1 used is: a pre-trained model based on Bert;

[0119] 1). Model 1 is for existing classifications, but there are still new classifications and unknown classifications. In this case, Model 2 is used;

[0120] 2). Input: material text;

[0121] 3). Output: material classification;

[0122] iii. Model 2: Bert-based clustering model;

[0123] 1). Input: material text;

[0124] 2). Output: clustering results;

[0125] 3). It is necessary to manually analyze the clustering results and make a type judgment based on the results;

[0126] 4). When the confidence exceeds the threshold, it can be converted to model 1.

[0127] S2: automatically assigning tasks of the same level according to the background of knowledge engineers, and knowledge engineers of the same level collaborate to complete the corresponding tasks in task management, specifically:

[0128] a) Allocation

[0129] i. When multiple people and multiple locations collaborate, the entire task is performed;

[0130] ii. Created by the knowledge manager and assigned corresponding tasks;

[0131] iii. Principles for creating tasks;

[0132] 1). The creation of the same material category is considered as one task;

[0133] 2). The creation of the same knowledge engineering processing material is a task;

[0134] 3). The creation of materials of the same reviewer is considered as one task;

[0135] 4. The creation of a co-located knowledge project is a task;

[0136] b) Collaboration

[0137] i. Multiple people work together on the same task;

[0138] ii. Standard setting;

[0139] 1).Material name;

[0140] 2). Material answers;

[0141] 3). Source of knowledge;

[0142] 4). Source description;

[0143] 5). Links;

[0144] 6). Belonging knowledge graph, knowledge base;

[0145] 7).Instance name;

[0146] 8). Is it universal?

[0147] 9). Material status;

[0148] 10). Editor;

[0149] The above needs to be filled in by the knowledge engineer and submitted accurately for review by the knowledge supervisor. If it passes the review, it will enter the next process. If it fails the review, it will be redone until it passes the review;

[0150] c) Progress

[0151] i. Focus on the knowledge engineer and the entire task progress;

[0152] ii. When the progress of the knowledge engineer seriously slows down the entire task, the knowledge manager will confirm the specific problem and split the related work among other members of the team;

[0153] iii. Knowledge managers are responsible for their respective tasks;

[0154] iv. Knowledge engineers are responsible for the knowledge they create;

[0155] v.KPI is borne by both parties.

[0156] The specific steps of S3: automatic knowledge extraction and conversion of implicit knowledge into explicit knowledge are as follows:

[0157] a). Knowledge management is the core of this system, and its main function is to automatically extract knowledge and verify knowledge;

[0158] b). Knowledge extraction;

[0159] i. Extract relevant questions and answers;

[0160] ii. Model: Retro-Reader model (retrospective reader);

[0161] iii. Principle;

[0162] iv. Retro-Reader·The model includes a sketchy-reading module and an intensive-reading-module to simulate human reading. The sketchy-reading module is used to read the article and questions and get a preliminary judgment; the intensive-reading module verifies the answerability and gives candidates. The outputs of the two modules are aggregated to make the final decision.

[0163] A knowledge engineering device based on collaborative processing includes a task storage module, a knowledge storage module and a knowledge application module.

[0164] The task storage module is specifically:

[0165] Mainly stored in relational databases, it stores all task-related information, and all stored information can be traced back to its source.

[0166] The knowledge storage module is specifically:

[0167] It is mainly stored in the knowledge graph. The final form of knowledge, symbolic representation, and vectorized representation are all stored in the knowledge graph and processed uniformly by the knowledge graph.

[0168] The knowledge application module is specifically:

[0169] a). Knowledge retrieval

[0170] i. Users can be divided into:

[0171] 1). External users: such as government officials, looking for related content;

[0172] 2). Actual users: business processing users, intelligent consultation;

[0173] 3). Knowledge Engineer: Find related and similar content.

[0174] ii. The main function of knowledge retrieval is also

[0175] b). Knowledge quiz.

[0176] i. Questions and answers can be provided for the relevant knowledge in the knowledge base, as well as intelligent consultation;

[0177] c). Data analysis,

[0178] i. Mainly used for statistical analysis of knowledge base related status;

[0179] ii. Quantify the efficiency of knowledge engineers;

[0180] iii. Quantify knowledge accuracy.

[0181] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A knowledge engineering method based on collaborative processing, characterized in that: The following steps are involved: S1: Collect, manage and classify materials, and obtain quantifiable knowledge sources to efficiently manage materials; the specific steps are: a). Material collection; b). Material import; c). Material library; d). Material classification; e). Element extraction; S2: Automatically assign tasks based on the knowledge engineer’s background, and the knowledge engineers collaborate to complete the corresponding tasks within task management; The tasks described include: a) Allocation i. When multiple people and multiple locations collaborate, the entire task is performed; ii. Created by the knowledge manager and assigned corresponding tasks; b) Collaboration i. Multiple people work together on the same task; ii. Standard setting; 1).Material name; 2). Material answers; 3). Source of knowledge; 4). Source description; 5). Links; 6). Belonging knowledge graph, knowledge base; 7).Instance name; 8). Is it universal? 9). Material status; 10). Editor; The material name, material answer, knowledge source, source description, link, knowledge graph, knowledge base, example name, whether it is universal, material status, and editor all need to be filled in by the knowledge engineer and submitted accurately for review by the knowledge supervisor. If the review is passed, it will enter the next process. If the review fails, it will be redone until it passes. c) Progress i. Focus on the knowledge engineer and the entire task progress; ii. When the progress of the knowledge engineer seriously slows down the entire task, the knowledge manager will confirm the specific problem and split the related work among other members of the team; iii. Knowledge managers are responsible for their respective tasks; iv. Knowledge engineers are responsible for the knowledge they create; v. KPI is undertaken by both parties; S3: Automatically extract knowledge and convert implicit knowledge into explicit knowledge. The specific steps are: a). Knowledge management is the core of this system, and its function is to automatically extract knowledge and verify knowledge; b). Knowledge extraction; i. Extract relevant questions and answers; ii. Model: Retro-Reader model; The Retro-Reader model includes a skimming module and an intensive reading module to simulate human reading. The skimming module is used to read articles and questions and get a preliminary judgment; the intensive reading module verifies the answerability and gives candidates. The outputs of the two modules are aggregated to make the final decision.

2. The knowledge engineering method based on collaborative processing according to claim 1, characterized in that: S1: Collecting and managing materials and classifying them, obtaining quantifiable knowledge sources. The specific steps for efficient material management are: a). Material collection; Sources of material collection: government systems, books, government documents, public accounts, and government official websites; b). Material import; Import relevant original materials into the material library; c). Material library; The material library is a collection for storing original materials and splitting materials; d). Material classification; e). Element extraction; i. Business word extraction; 1). Entity extraction, provided by the knowledge graph; 2). Business word extraction, obtained by matching the accumulated knowledge word library; 3). Relationship extraction, provided by the knowledge graph.

3. The knowledge engineering method based on collaborative processing according to claim 1, characterized in that: The material classification d) is specifically: i. Use machine learning algorithm models to classify materials; ii. Model 1 used is: a pre-trained model based on Bert; 1). Model 1 is for existing classifications, but there are still new classifications and unknown classifications. In this case, Model 2 is used; 2). Input: material text; 3). Output: material classification; iii. Model 2: Bert-based clustering model; 1). Input: material text; 2). Output: clustering results; 3). It is necessary to manually analyze the clustering results and make a type judgment based on the results; 4). When the confidence exceeds the threshold, it switches to model 1.

4. The knowledge engineering method based on collaborative processing according to claim 1, characterized in that: S2: automatically assigning tasks of the same level according to the background of knowledge engineers, and knowledge engineers collaborate to complete the corresponding tasks in task management. The principles for creating tasks include: 1). The creation of the same material category is a task; 2). The creation of the same knowledge engineering processing material is a task; 3). The creation of materials of the same reviewer is considered as one task; 4). The creation of co-located knowledge engineering is a task.

5. A knowledge engineering device based on collaborative processing, comprising a task storage module, a knowledge storage module and a knowledge application module, characterized in that: The task storage module is specifically: Stored in a relational database, all task-related information is stored, and all stored information can be traced back to its source; The tasks described include: a) Allocation i. When multiple people and multiple locations collaborate, the entire task is performed; ii. Created by the knowledge manager and assigned corresponding tasks; b) Collaboration i. Multiple people work together on the same task; ii. Standard setting; 1).Material name; 2). Material answers; 3). Source of knowledge; 4). Source description; 5). Links; 6). Belonging knowledge graph, knowledge base; 7).Instance name; 8). Is it universal? 9). Material status; 10). Editor; The material name, material answer, knowledge source, source description, link, knowledge graph, knowledge base, example name, whether it is universal, material status, and editor all need to be filled in by the knowledge engineer and submitted accurately for review by the knowledge supervisor. If the review is passed, it will enter the next process. If the review fails, it will be redone until it passes. c) Progress i. Focus on the knowledge engineer and the entire task progress; ii. When the progress of the knowledge engineer seriously slows down the entire task, the knowledge manager will confirm the specific problem and split the related work among other members of the team; iii. Knowledge managers are responsible for their respective tasks; iv. Knowledge engineers are responsible for the knowledge they create; v.KPI is borne by both parties.

6. The knowledge engineering device based on collaborative processing according to claim 5, characterized in that: The knowledge storage module is specifically: Stored in the knowledge graph, the final form of knowledge, symbolic representation, and vectorized representation are all stored in the knowledge graph and processed uniformly by the knowledge graph.

7. The knowledge engineering device based on collaborative processing according to claim 5, characterized in that: The knowledge application module is specifically: a). Knowledge retrieval i. Users are divided into: 1). External users: government officials, looking for related content; 2). Actual users: business processing users, intelligent consultation; 3). Knowledge Engineer: Search for related and similar content; ii. The role of knowledge retrieval is: b). Knowledge quiz; i. Do some questions and answers on the relevant knowledge in the knowledge base, and provide intelligent consultation; c). Data analysis, i. Used for statistical analysis of knowledge base related status; ii. Quantify the efficiency of knowledge engineers; iii. Quantify knowledge accuracy.

Citation Information

Patent Citations

  • Knowledge-engineering-based automatic scheme generation and evaluation system and method

    CN103646149A

  • Collaborative visual data recommendation method and device

    CN110489613A

  • Business process optimization for intelligent office automation system

    CN110489749A

  • NLP-based knowledge graph construction method

    CN111177411A

  • Motor fault knowledge extraction system and method

    CN112000791A