JP2020129385A5 - - Google Patents

Description

Knowledge management system

The present invention relates to a knowledge management system that efficiently accumulates and searches knowledge required for inference, judgment, recognition, and the like.

If a computer can make judgments and inferences that human beings make on a daily basis, it will be possible to process or simplify daily work in large quantities and at high speed. Development for this has been carried out conventionally.
Initially, judgment rules such as "if A is B" and "if B is C" are accumulated, and when the proposition "D is A" is given, in advance. An expert system has been developed that infers that "D is C" by sequentially applying the accumulated judgment rules. "MYCIN" is known to infer the optimal antibiotic for infectious diseases.

In the age of the Internet, more flexible search has been required. For example, using web records and declarations such as "E is a violin player", "a violin player is a musician", "a conductor is a musician", etc., "Is E a musician?" It answers the question (semantic web).
Here, under the superordinate concept of "musician", subordinate concepts such as "violin player", "conductor", and "composer" are developed. In this way, when the concept is described structurally (ontology), even if there is no direct description that "E is a musician" on the web, "E is a violin player, so he is also a musician." Is inferred, and the above question can be answered correctly.
Here, it is necessary to describe individual concepts and relationships in a unified format using a format such as XML so that concepts in multiple fields can be searched across. RDF (Resource Description Framework) has been proposed as an example of this unified format. The method of describing concepts and relationships in individual fragments works well with programming languages such as Prolog, and is easy to get used to relational databases.

In the medical field as well, attempts are being made to construct a disease ontology as shown in Non-Patent Document 1 below. In addition, Patent Document 1 and the like have also been proposed in the general industrial field.

Japanese Patent Application Laid-Open No. 2012-119004

Ontology expression and utilization of clinical medical knowledge (http://www.sanken.osaka-u.ac.jp/labs/nano/contents/meeting/pdf/S12_ohe.pdf)

In the early expert system, it is difficult to accumulate a large amount of judgment rules necessary for practical level operation. Further, since the condition part A of "if A then B" is applied only when it exactly matches, it is applicable when the search for "musician" is performed for "violin player" as in the above example. It will be none.
In the Semantic Web, it is ambiguous to some extent by utilizing the common sense that "violin player", "conductor", "composer", etc. are also included in the subconcept of "musician" by structural description such as ontology. It has become possible to support searches. However, when actually constructing a concrete ontology, the degree of freedom of the description format is so great that multiple description formats are possible even with the same content, and the structure described by the builder fluctuates. Therefore, even if the contents are the same, the description styles are often different. For this reason, when a plurality of ontology is constructed by a plurality of people in parallel work, confusion occurs due to inconsistency in description style. In addition, since various description methods are mixed, it is often difficult to understand even if the expanded ontology is seen, and the readability is often poor.

On the other hand, when knowledge fragments such as individual knowledge areas, knowledge items, attribute descriptions, and parent-child relationship descriptions are described in a format such as RDF, the same fluctuation of the description format is unavoidable, and in the first place, a large number of knowledge fragments are listed. Yes, the whole picture of knowledge cannot be listed, and the readability is extremely poor.
In order to efficiently describe a large amount of knowledge, it goes without saying that it is in a format that can be understood by a computer, but it also needs readability that can be easily understood by humans. As recommended by WEB 2.0, it is possible to easily describe knowledge without any special knowledge or skills, and to be able to easily point out contradictions, etc., even if you have knowledge in a certain field. It is an indispensable characteristic for many non-experts to build ontology etc. in parallel while cooperating via the web.

The present invention has been made to solve such a conventional problem, and an object thereof is to clearly separate and organize knowledge in a certain area into a plurality of knowledge trees, and each knowledge tree. Then, the knowledge entries that make up the tree are expanded in a hierarchical structure of parent-child relationship, and each knowledge entry is a set of entry attribute descriptions related to the knowledge entry, and the entry attribute description is to another knowledge entry or entry attribute description. By including a reference link, it is possible to describe systematic knowledge without fluctuation of notation.
Furthermore, the child knowledge entry inherits the entry attribute description of the parent knowledge entry, eliminating the need to duplicate the attribute description.
By enabling management of another area as another independent knowledge tree group separated by namespace, it is possible to create an area of interest without worrying about duplication of knowledge trees and knowledge entry names in other areas. You can concentrate on it.
By establishing such a knowledge management framework, we have built a knowledge management system that not only enables machine processing, but also has good readability that allows people to view the contents and that no matter who creates it, the description does not fluctuate. It is to provide the basis of knowledge processing that enables flexible and advanced search and inference.
In addition, by making it possible to reconfigure the knowledge management system by extracting a part or all of the knowledge management system and integrating it into another knowledge management system, it is possible to flexibly operate the knowledge management system. Is.
In addition to being able to browse the contents of the constructed knowledge management system, it is also possible to make inquiries to the knowledge management system and respond from the knowledge management system, enabling effective utilization in the real world. Furthermore, by managing the execution authority of the user and making it possible to set the scope of the search, security against destruction and confusion is ensured.

As a means for achieving the above object, in the knowledge management system according to claim 1, in knowledge recording and management,
It is provided with a knowledge tree group management means for managing at least one knowledge tree and a knowledge entry management means for managing at least one knowledge entry existing in each knowledge tree.
Each knowledge entry consists of a knowledge entry attribute description that describes the attributes related to the knowledge entry and a knowledge entry parent-child relationship link that describes the parent-child relationship with other knowledge entries in the knowledge tree.
The knowledge entry attribute description includes a vocabulary having a reference link to a knowledge entry belonging to a different or the same knowledge tree, or a vocabulary having a reference link to the entry attribute description of the knowledge entry.
In the parent-child relationship, it has a knowledge entry attribute description inheritance means that inherits the entry attribute description of the parent knowledge entry to the entry attribute description of the child knowledge entry.
It is provided with an access means for accessing the required knowledge by using a data structure composed of the knowledge entry having the knowledge entry attribute description and the knowledge entry parent-child relationship link.
The knowledge entry attribute category management means for managing the knowledge entry attribute description by dividing it into categories is provided.

The knowledge management system according to claim 2 is characterized in that, in the knowledge management system according to claim 1, the reference link has a reference characteristic management means for managing the strength of the reference and / or the script to be executed in reference. And.

The knowledge management system according to claim 3 is characterized in that, in the knowledge management system according to claim 2, the strength of the reference is made variable according to the observation frequency in the reference characteristic management means.

The knowledge management system according to claim 4 is provided with an external document reference means for referencing an external document or a group of external documents related to the entry attribute description in the knowledge management system according to any one of claims 1 to 3. It is characterized by that.

In the knowledge management system according to claim 5, in the knowledge management system according to any one of claims 1 to 4, the knowledge tree group management means manages a plurality of knowledge tree groups by dividing them into namespaces. It is characterized by having.

In the knowledge management system according to claim 6, in the knowledge management system according to any one of claims 1 to 5, any of the name space, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link constituting the knowledge management system. It is characterized by having a knowledge export means for extracting parts, creating a knowledge management subset, and exporting it to another knowledge management system.

In the knowledge management system according to claim 7, in the knowledge management system according to claim 6, the knowledge management subset extracted by the knowledge export means, or the knowledge management subset from a separately constructed knowledge management system. It is characterized by having a knowledge import means for reconstructing a namespace, a knowledge tree, a knowledge entry, an entry attribute description, and a parent-child relationship link.

In the knowledge management system according to claim 8, in the knowledge management system according to claim 5, each of the namespace, knowledge tree, knowledge entry, knowledge entry attribute category, knowledge entry attribute description, and parent-child relationship link is created and edited. The delete and reference functions are characterized by being provided with a user authority management means for managing the execution authority of the function for each user.

In the knowledge management system according to claim 9, in the knowledge management system according to any one of claims 1 to 8, in the knowledge management system, the parent-child relationship between the knowledge entries and / or the content of the knowledge attribute description of the knowledge entry. It is characterized by having a knowledge browsing means for browsing.

In the knowledge management system according to claim 10, in the knowledge management system according to any one of claims 1 to 9, the knowledge inquiry receiving means and the contents of inquiries that receive inquiries about the contents recorded and managed in the knowledge management system. It is characterized by having a knowledge inquiry response means for responding to.

In the knowledge management system according to claim 11, in the knowledge management system according to any one of claims 5 to 7, in the knowledge management system, the parent-child relationship between the knowledge entries and / or the content of the knowledge attribute description of the knowledge entry. It is equipped with a knowledge browsing means to browse the above, manages users who create each of the above namespace, knowledge tree, knowledge entry, entry attribute description, parent-child relationship link, and manages the part of the knowledge management system created by a specific user or user group. , Excluded from the search target range of the browsing or knowledge inquiry, or conversely, provided with a search scope management means that excludes only the part of the knowledge management system created by a specific user or user group as the search target range of the browsing or knowledge inquiry. It is characterized by that.

The knowledge management system according to claim 1 is provided with a knowledge tree group management means for managing at least one knowledge tree and a knowledge entry management means for managing at least one knowledge entry existing in each knowledge tree. Knowledge in a specific area is clearly separated and managed in each knowledge tree, and in each knowledge tree, the knowledge entries that make up the tree are expanded and managed in a hierarchical structure.
Each knowledge entry consists of a knowledge entry attribute description that describes the attributes related to the knowledge entry and a knowledge entry parent-child relationship link that describes the parent-child relationship with other knowledge entries in the knowledge tree, so the notation does not fluctuate systematically. It is possible to describe knowledge.
Since the knowledge entry attribute description includes a knowledge entry belonging to a different or the same knowledge tree, or a reference link to the entry attribute description of the knowledge entry, it is possible to refer to appropriate information at the link destination.

Since the knowledge entry attribute category management means for managing the knowledge entry attribute description by dividing it into categories is provided, the attribute description of the knowledge entry is appropriately managed for each category.

In the parent-child relationship, since the knowledge entry attribute description inheritance means for inheriting the entry attribute description of the parent knowledge entry to the entry attribute description of the child knowledge entry is provided, the common attribute description is bound to the parent knowledge entry, and the child knowledge entry is. , Minimal attribute description is required.
In addition, all the attribute descriptions of the knowledge entry, which is the parent of the knowledge entry, are inherited and displayed, which facilitates a comprehensive understanding of the attribute descriptions.

The knowledge management system according to claim 2 has a reference characteristic management means for managing the strength of the reference and the script executed for the reference in the reference link, so that it is possible to perform Bayesian inference according to the situation and check mistakes. Prevents and improves medical safety.

In the knowledge management system according to claim 3, since the strength of the reference is made variable according to the observation frequency in the reference characteristic management means, it is possible to perform Bayesian estimation more in line with the situation.

Since the knowledge management system according to claim 4 is provided with an external document reference means for referencing an external document or a group of external documents related to the entry attribute description in the reference link, the related books, files, and the web are provided. Documents can be referred to.

In the knowledge management system according to claim 5, since the knowledge tree group management means has a namespace management means for managing a plurality of knowledge tree groups by dividing them into namespaces, a namespace is set for each area. As a result, it is managed separately from the knowledge tree names in other areas, and large-scale knowledge management becomes possible without confusion.

In the knowledge management system according to claim 6, an arbitrary part of the knowledge space, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link constituting the knowledge management system is extracted to create a knowledge management subset, and another Since it is equipped with a knowledge export means for exporting to a knowledge management system, it is possible to exchange contents between knowledge management systems and to build a knowledge management system on a cloud separate from the company.

The knowledge management system according to claim 7 imports the knowledge management subset extracted by the knowledge export means or the knowledge management subset from a separately constructed knowledge management system, and imports a namespace, a knowledge tree, and knowledge. Since it is equipped with knowledge import means for reconstructing entries, entry attribute descriptions, and parent-child relationship links, it is possible to incorporate some or all of the knowledge management system developed on the cloud into its own knowledge management system.
In addition, some or all of the knowledge management system can be ported to other networks as needed.

In the knowledge management system according to claim 8, the functions of creating, editing, deleting, and referencing each of the namespace, knowledge tree, knowledge entry, knowledge entry attribute category, knowledge entry attribute description, and parent-child relationship link are described for each user. Since the user authority management means for managing the execution authority of the function is provided, it is possible to prevent the system from being damaged by an unfamiliar user. In addition, by limiting the reference range by the user, it is possible to retain knowledge that should be kept secret.

The knowledge management system according to claim 9 is provided with a knowledge browsing means for browsing the parent-child relationship between knowledge entries and the content of the knowledge attribute description of the knowledge entry, so that the knowledge can be browsed according to the hierarchical structure.

The knowledge management system according to claim 10 is provided with a knowledge inquiry receiving means for receiving inquiries about recorded and managed contents and a knowledge inquiry replying means for replying to the contents of the inquiry in the knowledge management system. Accessibility is improved by automating searches for necessary information.

The knowledge management system according to claim 11 manages each user who creates each of the namespace, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link, and manages the part of the knowledge management system created by a specific user or user group. , Excluded from the search target range of the browsing or knowledge inquiry, or conversely, provided with a search scope management means that limits only the part of the knowledge management system created by a specific user or user group to the search target range of the browsing or knowledge inquiry. Therefore, it is possible to exclude inappropriate targets from the search range, or to limit the search target to a specific area and acquire appropriate information.

It is an example of a knowledge tree group, and is a schematic diagram of a medical knowledge tree group. It consists of a knowledge tree such as disease name, symptom / finding, drug name, etc. This is an example of knowledge tree group management means. The knowledge tree ID of each knowledge tree in the knowledge tree group and the knowledge tree name are managed in the form of a master table. As an example of the knowledge tree group, a part of the knowledge entry group constituting the disease name knowledge tree is illustrated. The knowledge tree ID, the knowledge entry ID, and the knowledge entry name are managed in the form of a master table for the knowledge entry group constituting the disease name knowledge tree, which is an example of the knowledge entry management means. The parent-child relationship link and knowledge entry attribute description are shown using diabetes, which is one of the knowledge entries that make up the disease name knowledge tree, as an example. This is an example of knowledge entry attribute category management means. The knowledge tree ID, knowledge entry attribute category ID, and knowledge entry attribute category name are managed in the form of a master table. It is explanatory drawing of the reference characteristic management means of the reference link which constitutes the entry attribute description. (a) Example of reference link to label and definition knowledge attribute category in knowledge entry item or knowledge entry item (b) Reference link example that manages reference strength such as frequency of occurrence in reference link (c) When referring Reference link example that defines a series of scripts to be executed It is explanatory drawing of namespace management. It is explanatory drawing of the namespace management means. This is an example of the screen configuration of the knowledge viewing means.

In FIG. 1, the knowledge recording system of the present invention will be outlined by taking medical knowledge as an example.
As a knowledge tree, "disease", "symptoms, laboratory findings", "drugs", etc. are illustrated.
In addition, "treatment", "insurance claim", etc. can be considered.
Each knowledge tree consists of a set of knowledge entries connected by a parent-child relationship.
The knowledge tree group management means manages the knowledge tree in the knowledge tree group.

FIG. 2 is an example of the knowledge tree group management means, and manages the knowledge tree ID and the knowledge tree name in the group in the form of a master table.
FIG. 3 illustrates a group of knowledge entries constituting the “disease name” knowledge tree as an example of the knowledge tree.
The knowledge entry group of "disease name" is first divided into major categories such as "metabolic system", "digestive system", "musculoskeletal system", and "circulatory system". Each major classification is divided into middle categories such as "sugar metabolism system", "lipid metabolism system", and "amino acid metabolism system" in "metabolic system". The middle classification "sugar metabolism system" is further divided into sub-categories such as "diabetes" and "glycogen storage disease".

The subclass "diabetes" includes "type I diabetes" and "type II diabetes".
Here, "type I diabetes" and "type II diabetes", which correspond to the leaves at the end of the knowledge tree, are specific disease names. Knowledge entries such as minor classification, middle classification, and major classification, which correspond to branches by grouping out these common attributes, are constructed, and these function as container-type knowledge entries containing the lower classification and leaf disease names.
Here, the classification hierarchy is large, medium, and small, but depending on the area, a deeper hierarchy may be used, or a shallow hierarchy may be sufficient. Here, the classification criteria based on the pathological condition were used, but in addition, the criteria for classification by site such as "neck", "neck", "upper limb", "inflammatory system", "tumor system", There are also criteria for classification by etiology such as "infectious system" and "genetic system". In this way, there may be fluctuations in the expression due to the classification criteria, but in any case, once the administrator of the knowledge management system decides the classification criteria, the fluctuations in the subsequent expressions disappear. In short, the administrator of the knowledge management system may set the classification criteria according to the purpose. In some cases, a plurality of knowledge trees having different classification criteria may coexist.

FIG. 4 is an example of knowledge entry management means.
In the form of a master table, the IDs of knowledge entries in the knowledge tree are managed in the knowledge tree group.
The knowledge entry name must be unique within the knowledge tree to which it belongs, but if the knowledge tree to which it belongs is different, it is distinguished from each other by the knowledge tree ID defined as shown in Fig. 2, so the same knowledge Entry names are also allowed.
In this figure, the knowledge entries in the knowledge tree group are centrally managed regardless of the knowledge tree to which they belong, but in some cases, a separate master table is created for each knowledge tree to manage the knowledge entries. Is also good.
If the knowledge entry name is unique throughout the knowledge tree group regardless of the knowledge tree to which it belongs, the knowledge tree ID in the first column of FIG. 4 is unnecessary.

FIG. 5 illustrates the parent-child relationship link and attribute description, taking the knowledge entry of "diabetes", which is one of the container-type knowledge entries, and the leaves of "type I diabetes" and "type II diabetes" as examples. is there.
"Diabetes" has a link to "metabolic system" as a parent knowledge entry, and further has a link to "type I diabetes" and "type II diabetes" as a child knowledge entry.

The attribute description category of "diabetes" includes <pathological condition> and <complication>.
<Pathological conditions> include decreased insulin secretion, hyperglycemia, and excretion of sugar in urine.
Each <pathology> item consists of a reference link to the relevant knowledge entry in another knowledge tree, the "symptoms / findings" knowledge tree.
By doing so, the attribute description is performed using the controlled vocabulary like a thesaurus, and the fluctuation of the expression can be suppressed.
Of course, here, attribute description may be performed directly using a character string as in the past, or a link to a document may be provided, but fluctuations cannot be suppressed and the reference link function described later cannot be utilized. , Not preferable.
<Complications> include "renal failure" and "arteriosclerosis obliterans". Reference links to these complications will be reference links to the knowledge entry in the same "disease name" knowledge tree.

The cause of "type I diabetes" is the rapid necrosis of beta cells in the pancreas, so the only treatment is insulin injection supplementation.
On the other hand, the cause of "type II diabetes" is obesity and binge eating, and treatment is dietary restriction, oral administration of hypoglycemic agents, and finally injection of insulin.
In this way, "type I diabetes" and "type II diabetes" have different <cause> and <treatment>, but other <pathological conditions> and <complications> are common to both, and are common to both. It is described in a certain "diabetes".
In "type I diabetes" and "type II diabetes", which correspond to the leaves, even if only <cause> and <treatment> are described, the attribute description of <pathology> and <complication> of "diabetes", which corresponds to the parental knowledge entry, Furthermore, attribute descriptions such as higher metabolic diseases are inherited.
In this way, by grouping the common attribute description into the parent knowledge entry, the child knowledge entry can be completed with the minimum attribute description. The attribute description of the knowledge entry may be individually viewed and configured in the mind while tracing the parent-child relationship, but the knowledge entry attribute description inheritance means can be used to create the attribute description of the knowledge entry that is the parent rather than the knowledge entry. If all are inherited and displayed, it becomes easy to comprehensively understand the attribute description.

The attribute description inherited from the parent for each knowledge entry attribute category may be the knowledge entry attribute description of the child knowledge entry as it is, but what is the attribute description inherited to the knowledge entry attribute category of the child knowledge entry? If a different knowledge entry attribute description exists, the knowledge entry attribute description of the child knowledge entry will overwrite or add to the knowledge entry attribute description inherited from the parent. This overwritten or added knowledge entry attribute description will be further inherited for the knowledge entries below the grandchildren.
The proper use of overwriting or appending may be appropriately set by the knowledge entry attribute description inheritance means.

It is possible to describe attributes without using attribute categories, but by using attribute categories, it is possible to easily grasp the inheritance of each attribute category by the knowledge entry attribute description inheritance means. In addition, it can be said that it is more useful because the attribute description can be easily overwritten and added.

<Case> In the knowledge entry attribute category, a reference link to the medical record of the case having the disease name is recorded.
This makes it possible to directly refer to the patient record of the disease. The reference link to the case chart may be in any format as long as it enables access to the case information, such as the medical institution ID + patient ID, the URL of the patient chart, and the name of the patient chart file. Further, up to the document ID of the patient chart, for example, the document type + creation date may be included. In addition, a knowledge entry attribute category such as <literature> may be provided to describe links to related books, files, and documents on the Web (external document reference means).

FIG. 6 is an example of the knowledge entry attribute category management means.
In the form of a master table, the ID of the knowledge entry attribute category in the knowledge tree is managed in the knowledge tree group.
The knowledge entry attribute category name must be unique within the knowledge tree to which it belongs, but if the knowledge tree to which it belongs is different, it is distinguished from each other by the knowledge tree ID, so the same knowledge entry attribute category name is also allowed. ..
In this figure, the knowledge entry attribute categories in the knowledge tree group are centrally managed regardless of the knowledge tree to which they belong. However, in some cases, a separate master table is created for each knowledge tree to create the knowledge entry attribute category. May be managed.
If a unique knowledge entry attribute category name is used in the knowledge tree group, the knowledge tree ID in the first column is unnecessary. It is also possible to directly write the knowledge entry attribute category in the knowledge entry corresponding to the root of the knowledge tree and inherit it to the following without using the knowledge entry attribute category management means as explained in this figure. This is not preferable, especially in large-scale knowledge management systems, because the procedure is more complicated than the master table method when automating. In addition, although it is possible for the user to freely define the knowledge entry attribute category for each knowledge entry, duplication and fluctuation of the description are likely to occur, which is also not preferable.

FIG. 7 shows various embodiments of the reference link.
FIG. 7A is an example showing the structure of {hyperglycemia}, which is a reference link in the <pathology> of “diabetes” in FIG.
It consists of a link to the label "Hyperglycemia" and the knowledge entry "Hyperglycemia" in the "Symptoms / Findings" knowledge tree.
In some cases, it may be linked to the knowledge entry attribute category <definition> in "hyperglycemia". The label is the content to be displayed when browsing.
As in this example, the knowledge entry name of the link destination may be used as it is, but for easy viewing, "hyperglycemia blood glucose level> 140 mg / dl" may be used.
By doing so, it is possible to easily list and grasp the outline without the process of individually referring to the link destination.

FIG. 7B shows a reference link for the <side effects and incidence> knowledge entry attribute category within the "furosemide" knowledge entry in the "drug" knowledge tree.
The first line consists of a "hyperkalemia" label, a reference intensity of "5%", and a reference link to the "hyperkalemia" knowledge entry in the "symptoms / findings" knowledge tree.
Even with various side effects, the incidence is not uniform. Having information on the frequency of occurrence as the reference intensity clarifies the order of side effects to be considered.
In addition, this frequency of occurrence is a prior probability in terms of Bayesian probability, and is useful for Bayesian estimation to estimate the causative drug when side effects occur due to administration of a plurality of drugs.

FIG. 7 (c) shows the "herniated disk disk" knowledge entry in the disease name knowledge tree, with the "MRI" label in the <examination> knowledge entry attribute category, and the "MRI" knowledge entry in the "examination" knowledge tree. Along with the reference link, the flowchart of the check script at the time of implementation is shown.
In this way, by having a reference characteristic management means that manages the strength of the reference and the script that is executed when the reference is made, Bayesian estimation can be performed to help with the response, and check mistakes can be eliminated to improve medical safety. .. Further, in the reference characteristic management means, by making the strength of the reference variable according to the observation frequency in the case, it is possible to perform Bayesian estimation more in line with the situation.
Further, for example, when describing "fever" as a <symptom> of a patient, the reference link to the "fever" knowledge entry in the "symptoms / findings" knowledge tree uses the reference characteristic management means to "fever". You may manage the polarity of "yes / no". In addition, a reference link to the patient's medical record may be provided at the same time from the "fever" knowledge entry. In this way, what is the proportion of patients with fever in Tokyo? Or, you can easily answer the request for a list of patients with fever.

The knowledge tree group shown in FIG. 1 is related to medical care, but there are innumerable other areas where knowledge should be described, such as art and industry.
It is essential that the knowledge tree name is unique and unique with other knowledge tree names, but if the same knowledge tree name is used in other areas that are not known, it causes troublesome problems.
In order to avoid this problem, a namespace is set for each area as shown in FIG. As a result, the knowledge tree name becomes the namespace + the knowledge tree name, so that the knowledge tree names in other areas can be completely separated. FIG. 9 is an example in which the namespace ID and the namespace name are managed in the form of a master table (namespace management means). By adding the namespace ID to the first column of FIGS. 2, 4 and 6 as needed, large-scale knowledge management becomes possible without confusion.

FIG. 10 is an example of knowledge browsing means.
First, from the left, "Medical" is selected from the list of namespaces in the first column.
As a result, the second column lists the names of knowledge trees that belong to the "medical" namespace. If you select the "disease name" knowledge tree here, a list of major categories such as metabolic system and circulatory system is first shown. When "metabolic system" is selected, a list of "sugar metabolism system", "lipid metabolism system", etc., which are the lower taxa, is displayed.
When "sugar metabolism system" is selected, the subclasses "diabetes", "glycogen storage disease", etc. are listed.
When "diabetes" is selected, knowledge entry attribute categories such as <pathology> and <complications> are displayed.
When <Pathology> is selected, reference links such as "hyperglycemia" and "high HbA1c level", which are attribute descriptions constituting <pathology>, are displayed.
When "hyperglycemia" is selected, the attribute categories <definition>, <test method>, etc. of the "hyperglycemia" knowledge entry in the "symptom / finding" knowledge tree, which is the reference link destination, are displayed.
When <Definition> is selected, the contents of the knowledge entry attribute category are displayed.

As shown in FIG. 10, it is useful to proceed with browsing manually, but it is also useful to make inquiries such as search to the knowledge management system, perform processing such as set operation and logical operation on the search result, and display the result. Downloading to a file is even more useful.
It would be efficient if a series of inquiry procedures could be written in a script or the like and automatically processed in order (knowledge inquiry receiving means and knowledge inquiry replying means).

The knowledge management system of the present invention is assumed to be constructed on the cloud.
However, some companies and hospitals do not connect to the Web for security reasons. In such a case, extract any part of the knowledge space, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link that make up the knowledge management system, create a knowledge management subset, and export it to another knowledge management system. The knowledge export means to be used, the knowledge management subset extracted by the knowledge export means, or the knowledge management subset from a separately constructed knowledge management system is imported, and the namespace, knowledge tree, knowledge entry, etc. By providing a knowledge import means that integrates and reconfigures entry attribute descriptions and parent-child relationship links into the knowledge management system, some or all of the knowledge management system developed on the cloud is incorporated into the company's internal knowledge management system. Can be done.
In addition, if an excellent knowledge management system can be constructed in-house and there are other purchase applicants, a part or all of the knowledge management system can be transplanted to another by the same procedure.

Since the amount of information handled by the knowledge management system is enormous, the cooperation of many people is required to create it.
Set permissions for namespaces, knowledge trees, knowledge entries, knowledge entry attribute categories, knowledge entry attribute descriptions, parent-child relationship link creation, editing, deletion, reference functions, etc., depending on the privileges and skills of the people involved. Is indispensable (user authority management means).
This makes it possible to prevent damage to the knowledge management system by a user with poor understanding. In addition, by limiting the reference range by the user, it is possible to retain knowledge that should be kept secret.

To implement the knowledge management system of the present invention in a database, a graph database such as Neo4j, a relational database (RDB) that has been widely used in the past, and a key value store (KVS) that has recently attracted attention as a processing method for big data. and so on.
It is possible to use any implementation, but each has its own features / disadvantages.
First of all, Neo4J is good at setting and displaying network-like graph relationships, but the processing speed is not always fast in large-scale processing, so it will not be suitable for large-scale knowledge management systems.
In RDB, the knowledge management system of the present invention is implemented by describing parent-child relationship links and knowledge entry attributes for each row of relations under master table management such as namespace, knowledge tree, knowledge entry, and knowledge entry attribute category. Can be done.
Since the degree of freedom of search is high, the inquiry language such as SQL is complete, and there are many software assets from the past, implementation using RDB will be realistic.

In KVS, it consists of one to several items of data (columns) and a set of key data. Therefore, a wide variety of free searches are inferior to RDBs. However, for processing a large amount of data, distributed processing such as MapReduce is possible at high speed.
The case list in <cases> in FIG. 5 and the <cases> of "hyperglycemia" in the "symptoms / findings" knowledge tree will soon become a huge case list if an electronic medical record or the like is actually used. The frequency distribution of one attribute can be easily obtained by any method, but there are multiple cases such as "How often are diabetic patients with HbA1c exceeding 10 and urinary protein 2+ or more?" When determining the frequency of factors from large-scale case data, it sometimes exceeds the processing capacity of RDB. In such cases, it is necessary to perform a set operation between the diabetic patient list, the HbA1c> 10 patient list, and the urinary protein> 2+ patient list. In such a case, the distributed processing by MapReduce of KVS is extremely effective.

Further, in the knowledge management system of the present invention, since the knowledge entries have a hierarchical structure, for example, when it is desired to search the patient list of "fever", "remittent fever" or "continuous fever" which is a subconcept of "fever" It is possible to expand the search range to a patient list such as "" as needed.
As described above, the knowledge management system of the present invention may be most effective in a combination such that the basic part is RDB and the case list and the like are processed by KVS, regardless of the database implementation method.

The user authority management means sets the authority for each user regarding access to the knowledge management system, but this alone is not sufficient. There are examples of followers of a conflicting theory repeating overwriting, denying each other's writing. As a solution in such a case, a peaceful solution can be achieved by using a search scope management means that excludes the writing of another specific user from the search target from the viewpoint of a certain user.

Even if the same "chest pain" occurs, there is a high possibility of myocardial infarction if it occurs at a hospital specializing in heart disease, but there is a high possibility of rib fracture in the orthopedic outpatient department. In this way, the prior distribution of the frequency of occurrence of symptoms / findings often differs depending on the clinical department. In such a case, it is useful to set the search scope in a place that matches similar clinical departments, medical institution types, and regions.

Although the examples have been described above, the specific configuration of the present invention is not limited to the above examples, and even if there is a design change or the like within a range that does not deviate from the gist of the invention, it is included in the present invention.
For example, in this application, medical care is taken as an example, but other similar discussions such as education and human resources are possible. Experienced designers need to carefully define classification criteria such as knowledge trees and knowledge entries, and setting of knowledge entry attribute categories. The classification criteria for this application are just an example. Once the classification criteria have been defined, subsequent fluctuations in the description will be minimal.

Claims (11)

In the recording and management of knowledge
It is provided with a knowledge tree group management means for managing at least one knowledge tree and a knowledge entry management means for managing at least one knowledge entry existing in each knowledge tree.
Each knowledge entry consists of a knowledge entry attribute description that describes the attributes related to the knowledge entry and a knowledge entry parent-child relationship link that describes the parent-child relationship with other knowledge entries in the knowledge tree.
The knowledge entry attribute description includes a vocabulary having a reference link to a knowledge entry belonging to a different or the same knowledge tree, or a vocabulary having a reference link to the entry attribute description of the knowledge entry.
In the parent-child relationship, it has a knowledge entry attribute description inheritance means that inherits the entry attribute description of the parent knowledge entry to the entry attribute description of the child knowledge entry.
It is provided with an access means for accessing the required knowledge by using a data structure composed of the knowledge entry having the knowledge entry attribute description and the knowledge entry parent-child relationship link.
A knowledge management system characterized in that the knowledge entry attribute category management means for managing the knowledge entry attribute description by dividing it into categories is provided. The knowledge management system according to claim 1, wherein the reference link has a reference characteristic management means for managing the strength of the reference and / or the script to be executed in reference. The knowledge management system according to claim 2, wherein in the reference characteristic management means, the strength of the reference is made variable according to the observation frequency. The knowledge management system according to any one of claims 1 to 3, further comprising an external document reference means for referencing an external document or a group of external documents related to the entry attribute description. The knowledge management system according to any one of claims 1 to 4, wherein the knowledge tree group management means includes a namespace management means for dividing and managing a plurality of knowledge tree groups in a namespace. Knowledge export means that extracts arbitrary parts of the namespace, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link that make up the knowledge management system, creates a knowledge management subset, and exports it to another knowledge management system. The knowledge management system according to any one of claims 1 to 5, wherein the knowledge management system is provided. The knowledge management subset extracted by the knowledge export means, or the knowledge management subset from a separately constructed knowledge management system is imported, and the namespace, knowledge tree, knowledge entry, entry attribute description, parent-child relationship link are imported. The knowledge management system according to claim 6, further comprising a knowledge importing means for reconstructing the above. User authority to manage the execution authority of the function for each user for each function of creating, editing, deleting, and referencing the namespace, knowledge tree, knowledge entry, knowledge entry attribute category, knowledge entry attribute description, and parent-child relationship link. The knowledge management system according to claim 5, wherein the knowledge management system is provided with a management means. The knowledge according to any one of claims 1 to 8, wherein the knowledge management system is provided with a knowledge browsing means for browsing the parent-child relationship between the knowledge entries and / or the contents of the knowledge attribute description of the knowledge entry. Management system. Claims 1 to 9 characterized in that the knowledge management system is provided with a knowledge inquiry receiving means for receiving inquiries about recorded and managed contents and a knowledge inquiry replying means for replying to the contents of the inquiry. The knowledge management system described. The knowledge management system includes a knowledge viewing means for viewing the parent-child relationship between the knowledge entries and / or the content of the knowledge attribute description of the knowledge entry.
Each creation user of the namespace, knowledge tree, knowledge entry, entry attribute description, and parent-child relationship link is managed, and the part of the knowledge management system created by a specific user or user group is searched for for browsing or knowledge inquiry. Claim 5 is provided with a search scope management means for excluding from the range, or conversely, limiting only the part of the knowledge management system created by a specific user or user group to the search target range of the browsing or knowledge inquiry. ~ 7 Knowledge management system described in any one.