JP2000200197A - Knowledge storage/selection method, knowledge storage/ selection device and storage medium recording knowledge storage/selection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To present effective knowledge to a user and reduce the effect of the erroneous judgement of the user in a knowledge storage/retrieval device for selecting and presenting knowledge strongly relating to inputted knowledge from a huge amount of knowledge. SOLUTION: This device is provided with a means 21 for extracting the relational knowledge of the inputted knowledge from a data base 30, a means 22 for calculating a relational degree based on a utilization frequency for the extracted relational knowledge, a means 23 for selecting the knowledge whose relational degree is equal to or more than a prescribed threshold from the extracted relational knowledge, also selecting some knowledge at random from the unselected remaining relational knowledge and presenting the selected knowledge to the user and a means 24 for updating the utilization frequency of the pertinent knowledge in the data base in response to the user select the required knowledge from the presented relational knowledge.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to knowledge learning of a knowledge acquisition / utilization device, and more particularly to a knowledge accumulation / selection method and device suitable for improving learning efficiency, and a recording medium on which the program is recorded.

[0002]

2. Description of the Related Art In a knowledge storage / search device that selects and presents knowledge that is strongly related to input knowledge from a vast amount of knowledge, the knowledge of the degree of association is learned by the presence or absence of the selection. When narrowing down to reflect on knowledge selection from, it is necessary that the strength of association with important knowledge be as strong as possible. In such cases, human judgment is reflected in the strength of the relationship to improve the accuracy of learning, and the strength of the relationship is learned so as not to be greatly affected even if the judgment is wrong. It is desirable.

In the conventional knowledge accumulation / selection method, a weighted relation between knowledge and knowledge (degree of relevance) is stored in a database, and when a search is performed, a value on a threshold having the degree of relevance is stored. Had selected. For example, FIG. 3 shows knowledge that is strongly related to certain input knowledge and its use frequency. The relevance in the figure is a value obtained by (frequency / sum of frequencies). The current value of 1 indicates that knowledges A and B having a relevance of 0.2 or more (here, assumed to be 0.2) are selected as knowledge related to the input knowledge Q. Then, in the next step of 2, after increasing the frequency of the knowledge A and B selected in 1 by +5, the relevance is calculated again. Here, even when the same knowledge Q is input, knowledges A and B having a relevance of darkness 0.2 or more are selected. Thus, in response to the user's knowledge input, presenting the knowledge whose relevance calculated from the frequency exceeds the threshold as related knowledge, and updating the frequency of the presented knowledge, the next time the same knowledge is input The related knowledge was selected using the frequency after the update.

[0004]

In the prior art, since the above-described process is performed, there is a very low possibility that knowledge C to F will be selected after n steps, as shown at 3 in FIG. . As described above, in the conventional knowledge accumulation / selection method, an erroneous determination is made in the initial stage of knowledge accumulation, and once the value falls below the threshold value, a situation may occur in which knowledge that should be important is not selected.

The present invention has been made in view of the above-mentioned drawbacks of the conventional method, and not only selects knowledge having a relevance greater than a certain threshold value, but also partially selects the magnitude of the relevance. Based on probabilistic selection, knowledge with a degree of relevance less than the threshold has the possibility of being selected, and accumulation and selection of knowledge that can respond even if erroneous judgments are made. With that purpose.

[0006]

According to the present invention, there is provided a means for extracting relevant knowledge of input knowledge from a database, a means for calculating a degree of relevance of the extracted relevant knowledge based on a use frequency thereof, From the extracted related knowledge, select the knowledge whose relevance is equal to or higher than a predetermined threshold, and select some knowledge stochastically from the remaining related knowledge that has not been selected, and use the selected knowledge. The frequency of selected knowledge is added and the frequency of unselected knowledge is subtracted in response to the means to present to the user and the user selecting necessary knowledge from the presented related knowledge. Means to update the frequency of use of the relevant knowledge in the database not only increases the frequency of frequently selected knowledge, but also absorbs the user's judgment errors at the initial stage, Can be Ri is characterized in that to present a valid knowledge.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to an embodiment with reference to the drawings. FIG. 1 is a diagram showing a configuration example of hardware of the present invention. Reference numeral 10 denotes an input / output device including a keyboard 11 and a display 12 used by a user to input knowledge and output related knowledge, 20 denotes a CPU that controls overall processing, 30 denotes a large number of knowledge and related knowledge, and uses. This is a database that stores frequencies. Here, the CPU 20 is divided into a related knowledge collating unit 21, a relevance calculating unit 22, a knowledge selecting unit 23, and a frequency updating unit 24 as a configuration according to the present invention.

FIG. 2 shows a processing flow of the configuration example of FIG. The related knowledge collating unit 21 determines whether or not the knowledge 201 entered by the user from the keyboard 11 or the like exists in the database 30, and if so, extracts the related knowledge from the database 30 (step 210). . 215 indicates the extracted related knowledge. In order to narrow down the related knowledge to be presented to the user, the relevance calculating unit 220 calculates the relevance of the related knowledge 215 extracted by the related knowledge collating unit 21 from the use frequency attached to the knowledge (step). 220). The knowledge selection unit 230 selects related knowledge to be presented to the user based on the calculation result of the relevance calculated by the relevance calculation unit 220, and displays it on the display 12 (step 230). here,
The knowledge selecting unit 230 first selects, for the related knowledge 215 extracted by the related knowledge matching unit 21, knowledge whose relevance is equal to or greater than a predetermined threshold value, and then selects a number of remaining knowledge items not selected this time. Is probabilistically selected based on the magnitude of use frequency. The user selects necessary knowledge from the related knowledge displayed on the display 12. The frequency updating unit 24 adds the frequency of the knowledge selected by the user and subtracts the frequency of the non-selected knowledge, so that the frequency of use of the relevant knowledge in the database 30 depends on whether or not the user has selected. Update (step 240).

Hereinafter, the processing operations of relevance calculation, knowledge selection and frequency update will be described in detail. FIG. 4 shows an embodiment focusing on the calculation of the degree of association in the present invention. The table is obtained by extracting the knowledge A to F related to the input knowledge α and the use frequency of each of them from the database. The degree of relevance is determined by (each usage frequency) / (sum of usage frequencies)
By applying this, the relevance of each of the knowledge A to F to the keyword α is 0.22, 0.20, 0.1, respectively.
8, 0.16, 0.14, and 0.12.

FIG. 5 shows an embodiment focusing on knowledge selection in the present invention. Table 1 shows the frequency and relevance of knowledge A to F. First, knowledge that is equal to or greater than a predetermined threshold x is selected. If there is no knowledge equal to or greater than the threshold x, the knowledge with the highest relevance is selected. Here, since the threshold is set to 0.2, knowledges A and B are selected as shown in FIG. Next, some of the knowledge C to F not selected this time are selected stochastically based on the magnitude of the frequency. The number of selections is the number obtained by rounding up the decimal part of the result of ((1 / (number of selected knowledge)) * k)). Here k = 4
As shown in FIG. 3, ((1/2)) * 4 = 2 pieces are stochastically selected. No. 4 shows that knowledge D and E are newly selected as a result of stochastic selection, and finally knowledge A, B, D and E
Indicates that is selected.

Next, a stochastic selection method will be described with reference to the flowchart of FIG. 6 and a specific example of FIG. FIG.
9 is a flowchart of a stochastic selection method applied at the time of knowledge selection. The knowledge (relevant knowledge) initially listed as a probabilistic selection candidate is sorted in order of frequency (step 6).
10). Next, variables J and k are initialized (step 62).
0). Next, 1 is added to k (step 630). Next, the k-th frequency is added to the variable J, and is set as J again (step 640). Next, a value obtained by adding 1 to the remainder obtained by dividing the value obtained at random by the total frequency is set as an index. If the variable J is equal to or larger than the index, the k-th knowledge is selected (steps 650 and 660). Variable J is Ind
If it is less than ex, the process returns to the process of adding 1 to k.
After that, it repeats until one knowledge is selected. In selecting n pieces of knowledge, the operation of selecting one piece of knowledge from n pieces of knowledge first, and then selecting one piece of knowledge from n-1 pieces of knowledge is repeated. Through this series of processes, it is possible to stochastically select from knowledge having a relevance lower than the threshold based on the magnitude of the relevance.

FIG. 7 shows an embodiment of a stochastic selection method applied at the time of knowledge selection. Consider a case where two out of four knowledges C, D, E, and F that are below the threshold described in FIG. 5 are selected. To select two pieces of knowledge, two stochastic selections must be made. First, the first stochastic selection method will be described. In FIG. 7, 2 is a label in which the knowledge name and the frequency of 1 are rearranged in ascending order, and is a label indicating a value J obtained by adding the k-th, knowledge name, frequency and frequency in order from the left. First, a random number Rand is obtained. Here, it is assumed that Rand = 1234. Index = (1
234% (30 + 35 + 40 + 45)) is a random number of 128
4 means the remainder after dividing by the sum of the four frequencies, and the result is Index = 34. The value J is sequentially traced from the top, and knowledge having the first Index = 34 or more is selected. Here, knowledge E is selected. The second probabilistic selection is performed in the same manner as the first. However, three knowledge F, D and C excluding the knowledge E selected at the first time are selected. Here, Index = 45 and knowledge D with J = 70 are selected.

FIG. 8 shows an embodiment in which attention is paid to frequency updating in the present invention. In FIG. 8, 1 is the input keyword α
5 shows a case where the user selects A and E from the knowledge A, B, D, and E selected through the processes 2, 3, and 4 in FIG. Here, 5 is added to the frequency for knowledge A and E selected by the user, and 5 is subtracted from the frequency for knowledge B and D not selected. Thereby, in the next selection, A and E are easily selected, and B and D are hardly selected.

FIG. 9 shows an embodiment focusing on the effect of the present invention by comparing the present invention with a conventional method. FIG.
In the conventional method shown in (1), there is a very low possibility that knowledge not selected in the initial step of knowledge accumulation will be selected in a later step. Therefore, even if knowledge E is actually important knowledge, As shown, the relevance after n steps remains low. However, applying the present invention based on the threshold, the judgment of the user, and the probabilistic selection, there is a possibility that the knowledge not exceeding the threshold is also selected. It can be seen that the degree has increased and the possibility of being selected by the user has increased.

Although the embodiment of the present invention has been described above, a computer-readable recording medium such as a program for causing a computer to execute the processing procedures and algorithms of FIGS. It can be provided by recording it on a floppy (registered trademark) disk, a memory card, an MO, a CD-ROM, or the like.

[0016]

As described above, according to the present invention,
In a knowledge accumulation / search device that selects and presents knowledge that is strongly related to the input knowledge from a vast amount of knowledge,
In addition to presenting effective knowledge for users,
It is possible to perform knowledge accumulation / selection that is not easily affected by a user's erroneous judgment.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an embodiment of a knowledge accumulation / selection device of the present invention.

FIG. 2 is a processing flowchart showing an embodiment of a knowledge accumulation / selection method of the present invention.

FIG. 3 is a diagram illustrating a conventional general knowledge accumulation / selection method.

FIG. 4 is a diagram for explaining relevance calculation according to the embodiment of the present invention.

FIG. 5 is a diagram illustrating knowledge selection according to the embodiment of this invention.

FIG. 6 is a flowchart of a stochastic selection method applied at the time of knowledge selection according to the embodiment of the present invention.

FIG. 7 is a diagram illustrating a specific example of stochastic selection according to an embodiment of the present invention.

FIG. 8 is a diagram for explaining frequency updating in the embodiment of the present invention.

FIG. 9 is a diagram showing the effect of knowledge accumulation / selection of the present invention.

[Explanation of symbols]

 Reference Signs List 10 input / output device 20 CPU 21 related knowledge collating unit 22 relevance calculating unit 23 knowledge selecting unit 24 frequency updating unit 30 database

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Mitsuo Nagaoka 3-19-2 Nishishinjuku, Shinjuku-ku, Tokyo F-term within Nippon Telegraph and Telephone Corporation (reference) 5B075 ND03 PP02 PP03 PQ02 PQ38 PQ46 PR04 PR08 QM08 QP01

Claims (3)

[Claims] 1. A knowledge storage / retrieval device comprising a database for storing knowledge and its use frequency, selecting knowledge strongly related to the input knowledge (hereinafter referred to as related knowledge) and presenting it to a user. A step of extracting relevant knowledge of input knowledge from a database, a step of calculating a degree of relevance of the extracted relevant knowledge based on a use frequency, and a step of Selecting, from among the knowledge, knowledge whose relevance is equal to or greater than a predetermined threshold, stochastically selecting some knowledge from the remaining unselected related knowledge, and presenting the selected knowledge to the user; Updating the frequency of use of the relevant knowledge in the database in response to the user selecting necessary knowledge from the presented related knowledge. Knowledge accumulated and selection method characterized and. 2. A database for storing knowledge and its use frequency, means for extracting relevant knowledge of the input knowledge from the database, and means for calculating the degree of relevance of the extracted relevant knowledge based on the use frequency And, from among the extracted related knowledge, select the knowledge whose relevance is equal to or greater than a predetermined threshold, and probabilistically select some knowledge from the remaining unselected related knowledge, and use the selected knowledge. Means for presenting to the user, and means for updating the use frequency of the relevant knowledge in the database in response to the user selecting necessary knowledge from the presented related knowledge. Knowledge storage and selection device. 3. A knowledge storage / retrieval device comprising a database for storing knowledge and its use frequency, selecting knowledge strongly related to the input knowledge (hereinafter referred to as related knowledge) and presenting it to a user. Is a computer-readable storage medium that stores a knowledge accumulation / selection program for the process of extracting related knowledge of input knowledge from a database, and relating the extracted related knowledge based on the frequency of use. A process of calculating the degree, and selecting knowledge having a degree of relevance equal to or greater than a predetermined threshold from the extracted related knowledge, and selecting some knowledge stochastically from the remaining related knowledge not selected. In the process of presenting the selected knowledge to the user, and in response to the user selecting necessary knowledge from the presented related knowledge, the database Recording medium, characterized in that the process of updating the use frequency of the corresponding knowledge, is recorded.