JPH0616268B2 - Separate knowledge base generation support method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to a knowledge base generation method for storing various data necessary for construction or execution of a knowledge processing system including an expert system, and a knowledge base. The present invention relates to a separation-type knowledge base support method for making a method of extracting data in a base simpler and more effective.

[Conventional technology]

Conventionally, the knowledge base is generated by a knowledge engineer who draws knowledge based on his experience from an expert and systematizes it in addition to highly general knowledge that is widely known in the problem domain. 1 object, 1 object from 1 data obtained through a specific format from an existing data file by expressing the relationship between each object.
A method is used in which the commands or functions prepared by the knowledge processing system are used for generation. The knowledge base created in this way has a mixture of objects for controlling knowledge processing and objects for storing data, and thus becomes a considerably large knowledge base.

On the other hand, there are two methods for fetching data from the data storage object: direct fetch and fetch via demon, which is a procedure that is automatically activated when a specific condition occurs. It is carried out by the Norwich engineer for each data item. This information can be stored in the knowledge base as well as ordinary objects.

[Problems to be solved by the invention]

In the above-mentioned conventional knowledge base generation method, it is necessary to issue a description or command (or function) prepared by the knowledge processing system for generation of all objects. Therefore, some knowledge of knowledge engineers is required even for very plain general data. Furthermore, if you want to use the data in an existing file, such as a database, you must create a tool that reads the contents of the data you want to use and creates an object. In such a case, the data format of the target data is often specified, and there is a problem of poor expandability and versatility. Moreover, in the conventional knowledge base,
In many cases, a knowledge processing control object such as a rule object and a data storage object are closely integrated. Therefore, there is a problem that the entire knowledge base becomes huge. When knowledge processing is performed such that the range of objects to be accessed can be determined in advance of such a knowledge base, the extra objects have little meaning and may cause a problem in data inspection.

On the other hand, in the method of extracting data from the data storage object, the data directly extracted is usually in a format convenient for data input or processing.
In many cases, it cannot be used as embedded information when outputting a processing result message to the user. Therefore, there is a method that uses demon, which is a procedure that is automatically activated when a specific condition occurs, and then converts the data into an arbitrary format and retrieves it. However, in the conventional method, this demon must be intentionally created by all Norwich engineers,
It increased the work burden on Norwich engineers.

[Means for solving problems]

In the distributed knowledge base generation support method of the present invention, a. Input user-specified format data that indicates which part of one input data unit corresponds to which item, and any item name in any one data unit, and the value of that item is in which classification The user's instructions can be returned according to the standard support of this support system and the classification function that the user defines in a simple format. Select and input the data within the range from the existing data file and expand it on the memory. As a result, the range of data to be handled can be limited in advance, and the burden of each process such as inspection can be reduced.

b. In addition, the thus expanded data is converted into data in a knowledge storage format suitable for the knowledge processing system to be used, that is, an object. At this time, the object can be hierarchized based on the classification name derived from the classification function according to the user's instruction, and it is possible to deal with advanced knowledge processing.

c. In order to convert the data in each data storage object generated in b into a format suitable for the processing intended by the user and to retrieve it, according to the instruction of the format data in a, a specific condition A demon, which is a procedure that is automatically activated when it occurs, is registered for each data item. With the demon that this support system supports as standard, the user can retrieve the data through digit conversion of numeric data, automatic conversion of 1-byte characters to 2-byte characters, and addition of units and subscript names. , The user's burden is reduced. Also, user-defined demons can be registered arbitrarily, and more advanced conversion and extraction is possible. In addition, when activating these demons, it is possible to dynamically give control information such as a mode and a pair flag to retrieve a wider range of information.

d. The management information for realizing the function of c can be in the form of an object, and can be saved together with the data storage object generated in b as one knowledge base and saved in a file. The knowledge base created in this way can be easily combined with the knowledge base of the parent who is the subject of knowledge processing.
Since it can be re-separated, it is possible to use and maintain the knowledge base in a distributed manner.

〔Example〕

The present invention will be described with reference to the drawings. FIG. 1 shows the overall flow of the present invention. The central part shows the flow of processing, and the processing of the present invention is the part surrounded by the broken line. The left side abstractly shows how the data group to be processed changes. The right side shows the information that the user should give in each process. As a specific example of the explanation, an example of generating a data storage knowledge base used in the rental housing consultation expert system will be shown below.

First, in the data input / conversion unit 2, the input data file 1
Data is converted into, for example, a LISP list format that can be supported by the support system of the present invention, and passed to the data selection expansion unit 3. In the data input / conversion unit 2, the internal processing differs depending on the format of the input data file such as sequential, random access, and database, and the language used.

Next, the processing of the data selection / decompression unit 3 will be described with reference to FIG. First, the data selection expansion unit 3 uses the format data 35.
Thus, the recognition module 31 recognizes which data item each part of the captured one data unit corresponds to. Then, according to the selection information, the validity of the specified item is checked by the check module. “* <” And “*> =” shown in FIG. 2 are classification functions 36 provided by the data selection expansion unit 3 as a standard for examining the magnitude relationship of the numbers.
"Bath / toilet?" Is a user-defined classification function 37. When the value of the specified item provided by the data selection expansion unit 3 is a list format, use the * GETL function that returns the nth element. , "Item name: When the fourth element of the floor plan is BT (with bath / toilet) or UB (unit bath)," Yes "
Is returned, and in other cases "none" is returned. " Here, the rent of data b is 950,000 yen, so it violates the condition of (<< 7.0 rent) and is excluded from the target data. With respect to the other data, a data name is generated by the function shown in the lower part of the selection information 38 and is expanded on the memory 5 in FIG. Here, by the function * PROP-INC provided by the data selection expansion unit 3, the name is given as "(value of nearest station) + (expansion order with that value)".

Next, the object conversion unit / conversion demon registration unit 4 will be described in detail with reference to FIG. First, the developed data is given a classification name in the classification step 41 in accordance with the instruction of the classification information 45. Here, the user-defined classification function 44 called "along the railway" returns the name of the railway as the classification name from the value of the nearest station. Next, the expanded data is made into an object, that is, an object format, by an object generation command (or function) provided by the knowledge processing system. In the case of Shinjuku 1, the classification name is Yamate, but initially there is no Yamate object, so this is always created under the existing TOP object 49, and Shinjuku 1 is created under it. Other data is similarly made into objects, and the entire hierarchical structure is constructed. Then, the conversion demon registration unit 4 refers to the format data 46 and corresponds to each item (which is the slot of the object),
A demon 48, which is a procedure that is automatically activated when a specific condition occurs, is registered in the TOP object 49. The standard support demon * UNIT is registered for items for which a character string is simply specified. Here, Table 1 shows an example of starting each demon.

When starting the demon, you can specify the mode and pair flag. When the mode is DIRECT, the demon is not activated, and when it is EXEC, it is activated. NEXEC and KEX
In case of EC, the item name and the demon start result are combined and returned as a sentence. When the pair flag is ON, the value when the demon is not activated and the value when it is activated are grouped in a list and returned. * UNIT of the standard support demon works by combining the item value and the specified unit name and returning it as a character string. * FORM
AT is a demon that performs format conversion of numerical values and unit name combination, and here, normally floating point display is done, but unit name "10,000 yen" is combined in the format of one decimal part, character string is returned I'm telling you to. SYUBET
SU and MADORI are user-defined demons.

Finally, the management information object conversion unit 6 has a function provided in the next process of converting the information such as the format data 46 and the classification function 44, which have been used in the processing so far, into an object and saving the file as a knowledge base. .

An example of using the knowledge base created in this way is shown in FIG. The upper part shows the block diagram of the rental house recommendation expert system. The part indicated by the file is the knowledge base created by this support system. Since the amount of rent to be paid for the user can be inevitably decided at first, the corresponding knowledge base 6
You can choose 8. The lower part shows how the object groups 61 are combined at that time. It is assumed that the knowledge processing control object group 61 contains control knowledge for searching and recommending properties in the rental house recommendation system. This knowledge, commonly referred to as rule knowledge, is driven by the inference engine 69. On the other hand, the data storage object group 62 is a knowledge base created by this support system. The knowledge processing control object selects the knowledge base in which the required data storage object group 62 is stored by referring to the management object 6A of the arbitrary knowledge base in which this data storage object group 62 is stored. , Can be combined under the objects it has. For example, the range of the data storage object group 62 of rent can be known by looking at the information of the classification function of the management object. Therefore, at the initial stage, the user inputs the desired rent, the data storage object group 62 is selected and combined, and the search and recommendation processing is performed within that range. Once combined in this way, the user can search for properties that satisfy the desired conditions according to the inspection rule 63. Further, when a plurality of candidates are submitted, the recommendation rule 64 can be used to receive professional advice. If the property data is maintained on a small scale, the knowledge editor 65 is used to correct the contents of each knowledge base. In the case of a large scale change, the support system may regenerate the knowledge base based on the changed data file, and at this time, the information such as the format data and the classification function can be used.

〔The invention's effect〕

As described above, the present invention can generate a knowledge base from an existing data file under arbitrary conditions by giving very simple information such as format data even if not an expert such as a Norwich engineer. it can.

Also, since this knowledge base can be combined with the parent knowledge base for controlling knowledge processing, multiple knowledge bases can be created for each major classification, and selected / combined. Can be improved. The same applies to maintenance.

On the other hand, when data is extracted from the objects in the knowledge base, it can be converted and output by the standard support demon, and by effectively using the unit addition and text conversion, the use of the demon can be simplified and the workload of the knowledge engineer can be reduced. There are effects such as reduction.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration including an embodiment of the present invention, and FIG.
The figure shows the data selection and
FIG. 3 is a diagram showing how the data is expanded, FIG. 3 is a diagram showing how the data finally expanded in FIG. 2 is made into an object, and demon registration is performed, and FIG. It is a figure which shows the example at the time of using a knowledge base in a rental housing consultation expert system, and the mode of combination. In FIG. 1, 1 ... Input data file, 2 ... Data input conversion unit, 3 ... Data selection expansion unit, 4 ... Object conversion unit / conversion demon registration unit, 5 ... Memory, 6 ... Management information Object, 7 ... File save, 8 ... Output file,

Claims (1)

[Claims] 1. A selective expansion means for selecting data satisfying a user's instruction from a data group having a certain format and expanding it on a memory, and this selective expanding means expanded on a memory. Object conversion means for converting data into an object which is data in a knowledge storage format suitable for a knowledge processing system, and at least one of a variable value and an attribute value in the object converted by this object conversion means A conversion demon registration means for registering a demon for conversion into a format suitable for processing and taking it out, and an object converted by the object conversion means so as to be able to be additionally combined with a knowledge processing system such as an expert system. Management information for creating objects in the form of a knowledge base with data storage objects Separated knowledge base generation support system, characterized in that by including the management information objects means that allows saving to.