Artificial Intelligence System

Abstract

To provide an artificial brain system which can be used as a brain without taking an enormous amount of time in learning, and even during the learning.SOLUTION: An artificial brain system is configured to uses a concept structure table which associates representative concepts with subordinate concepts, a concept knowledge table which associates the concepts with concept values, and a knowledge table which associates the representative concept values with the subordinate concept values, so as to confirm that the concept (flower) of a knowledge (flower species name) asked by question information (e.g., asking what pink spring flowers are) is identified by the concept (color, season) corresponding to the concept values (pink, spring) equivalent to each of one or more element values (pink, spring) included in the question information (e.g., asking what pink spring flowers are); generate answer information in response to the question information (e.g., asking what pink spring flowers are) on the basis of the representative concept values (cherry trees, lotus flower grasses) corresponding to the subordinate concept values (pink, spring) equivalent to the element values (pink, spring); and output the answer information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to artificial intelligence systems. [Background technology]

[0002] A neural network has been proposed as a computational processing model that mimics the information processing mode of the brain (see Patent Document 1). This neural network is a model that mimics the neurons and synapses of the brain, and is composed of two stages of processing: learning and discrimination. In the learning stage, the features of a large number of inputs are learned, and a neural network for discrimination processing is constructed. In the discrimination stage, the neural network is used to discriminate what new inputs are.

[0003] In recent years, technology at the learning stage has made great advances; for example, deep learning is making it possible to build multi-layer neural networks with high expressive power. The knowledge structure of a neural network has a fast processing speed when inference is made as unchanging knowledge of conceptual knowledge, and can be said to be an optimal knowledge structure. In addition, the more parent data that is learned, the deeper the knowledge of characteristic concepts becomes, and it can be said to be an excellent knowledge inference system when evaluating the entire conceptual knowledge. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2010-146514 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, in a neural network, which can be said to be an artificial brain that mimics the information processing method of the human brain, it is necessary to extract features from a large amount of input information for learning, which takes a lot of time.In addition, during this time-consuming learning process, it is not possible to perform discrimination operations (inference operations), and the neural network cannot be used as a brain during learning.

[0006] The present invention has been made in view of the above circumstances, and provides an artificial intelligence system which does not require a large amount of time for learning and which can be used as a brain even during learning. [Means for solving the problem]

[0007] The artificial intelligence system according to the present invention includes a storage unit for storing a concept structure table that associates a concept determined as a representative concept with a subordinate concept determined as a concept that specifies the representative concept, a concept knowledge table that associates a concept with a concept value as information representing knowledge about the concept, and a knowledge table that associates a representative concept value as information representing knowledge about the representative concept with a subordinate concept value as information representing knowledge of a subordinate concept determined as a concept that specifies the representative concept, and an information input unit for inputting question information (what is a pink spring flower) to be asked by presenting knowledge (flower species name) about a certain concept (flower) with element values ​​(pink, spring) representing each of one or more elements that specify the knowledge (flower species name). the conceptual structure table, the conceptual knowledge table, and the knowledge table stored in the memory unit to confirm that the concept (flower) of knowledge (flower species name) asked in the question information (what are the pink spring flowers) is identified by the concept (color, season) corresponding to the concept value (pink, spring) equivalent to each of one or more element values ​​(pink, spring) included in the question information (what are the pink spring flowers), and generate answer information to the question information (what are the pink spring flowers) based on the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) equivalent to the element value (pink, spring); and an information output unit to output the answer information obtained by the information processing unit. Note that the descriptions in parentheses are merely examples and do not limit the configuration of the present invention (the same applies to the section on Means for Solving the Problems).

[0008] According to this configuration, when a question information (what is a pink spring flower) is input to ask about knowledge (for example, a flower species name) by presenting element values ​​(pink, spring) representing one or more elements that identify the knowledge (flower species name), it is confirmed that the concept (flower) of knowledge (flower species name) asked in the question information (what is a pink spring flower) is specified by the concept (color, season) corresponding to the concept value (pink, spring) corresponding to one or more element values ​​(pink, spring) included in the question information (what is a pink spring flower) using three tables, the concept structure table, the concept knowledge table, and the knowledge table, and answer information is generated based on the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) corresponding to the element value (pink, spring). Then, the answer information (cherry blossom, lotus flower) is output as an answer to the question (what is a pink spring flower) represented by the question information.

[0009] In the artificial intelligence system according to the present invention, the information processing unit may have a concept knowledge determination means for determining whether or not there is a correspondence between a concept value (pink, spring) corresponding to each of the one or more element values ​​(pink, spring) and a concept (color, season) in the concept knowledge table; a concept structure confirmation means for confirming, when the concept knowledge determination means determines that there is the correspondence in the concept knowledge table, that the concept (color, season) corresponding to the concept value (pink, spring) in the concept knowledge table and the concept (flower) of the knowledge (name of flower species) asked in the question information (What are the pink spring flowers?) are corresponding to each other in the concept structure table; and an answer generation means for generating the answer information based on a representative concept value (cherry blossom, lotus flower) corresponding to a subordinate concept value (pink, spring) corresponding to the one or more element values ​​(pink, spring) in the knowledge table after confirmation by the concept structure confirmation means.

[0010] According to this configuration, when it is determined that there is a correspondence in the concept knowledge table between a concept value (pink, spring) corresponding to each of one or more element values ​​(pink, spring) and a concept (color, season), it is confirmed that the concept (color, season) corresponding to the concept value (pink, spring) in the concept knowledge table and the concept (flower) of the knowledge (flower species name) asked in the question information (What are the pink spring flowers?) are corresponding in the concept structure table, and after this confirmation, the answer information is generated based on the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) corresponding to the one or more element values ​​(pink, spring) in the knowledge table.

[0011] In the artificial intelligence system of the present invention, the answer generation means can be configured to have an answer determination means for determining a concept value (cherry blossom, lotus flower) corresponding to the representative concept value (cherry blossom, lotus flower) in the concept knowledge table as an answer concept value, and to generate the answer information based on the answer concept value determined by the answer determination means.

[0012] With this configuration, the concept value (cherry blossom, lotus flower) corresponding to the representative concept value (cherry blossom, lotus flower) in the concept knowledge table is determined as the answer concept value, and the answer information is generated based on the answer concept value.

[0013] In the artificial intelligence system according to the present invention, the information processing unit, when it is determined by the concept knowledge determination means that there is a correspondence between a concept value (pink, spring) corresponding to each of the one or more element values ​​(pink, spring) of the question information (what are pink spring flowers?) and a concept in the concept knowledge table, manages the concept value (pink, season) as an arousal concept value of the corresponding concept, and manages a subordinate concept in the knowledge table that is the same as the concept value (pink, spring) managed as an arousal concept value by the first arousal concept value management means. The system can be configured to have a second arousal concept value management means for managing a representative concept value (cherry blossom, lotus flower) in the concept knowledge table as an arousal concept value, and a third arousal concept value management means for managing a concept value (cherry blossom, lotus flower) in the concept knowledge table that is the same as a representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) managed as an arousal concept value by the second arousal concept value management means, and the answer determination means determines the concept value (cherry blossom, lotus flower) managed as an arousal concept value by the third arousal concept value management means in the concept knowledge table as the answer concept value.

[0014] With this configuration, when it is determined that there is a correspondence between a concept value (pink, spring) corresponding to one or more element values ​​(pink, spring) of question information (what are pink spring flowers) and a concept (color, season) in the concept knowledge table, the concept value (pink, season) of the concept (color, season) is managed as an arousal concept value (first arousal concept value management means), and the subordinate concept value (pink, spring) in the knowledge table that is the same as the concept value (pink, spring) managed as the arousal concept value is managed as an arousal concept value (second arousal concept value management means). Furthermore, the concept value (cherry blossom, lotus flower) in the concept knowledge table that is the same as the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) managed as an arousal concept value in the knowledge table is managed as an arousal concept value (third arousal concept value management means). Then, the concept value (cherry blossom, lotus flower) managed as the awakening concept value in the concept knowledge table (third awakening concept value managing means) is determined as the answer concept value, and the answer information is generated based on the answer concept value.

[0015] Each of the first arousal concept value management means and the third arousal concept value management means can include an arousal flag that is associated with each concept value in the concept knowledge table and has a different state depending on whether the concept value is the arousal concept value or not. Also, the second arousal concept value management means can include an arousal flag that is associated with each subordinate concept value in the knowledge table and has a different state depending on whether the subordinate concept value is the arousal concept value or not.

[0016] Depending on the state of the arousal flag, it is possible to manage whether each concept value (subordinate concept value) is an arousal concept value or not.

[0017] In the artificial intelligence system according to the present invention, the information processing unit may be configured to have an answer control means (decision-making flag) for controlling whether or not a representative concept value is used in generating the answer information in the knowledge table.

[0018] With this configuration, it is possible to control whether or not to use the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) equivalent to the element value (pink, spring) included in the question information in generating the answer information.

[0019] In the artificial intelligence system according to the present invention, Answer system The control means may be configured to have an answer control flag (decision-making flag) that is associated with a representative concept value in the knowledge table and has a different state depending on whether or not the representative concept value is used to generate the answer information.

[0020] With this configuration, it is possible to control, depending on the state of the answer control flag, whether or not to use, as answer information, the representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value (pink, spring) equivalent to the element value (pink, spring) contained in the question information.

[0021] In the artificial intelligence system according to the present invention, a related answer generating means is provided for generating related answer information related to the answer information (cherry blossom) based on a representative concept value (entrance ceremony) corresponding in the knowledge table to a subordinate concept value (cherry blossom) that is the same as a representative concept value (cherry blossom) used in generating answer information to the question information by the answer generating means, and the information output unit is connection The related answer information generated by the answer generating means may be output.

[0022] With this configuration, knowledge (entrance ceremony) related to the answer information (cherry blossoms) determined for the question information can be provided as a related answer (related answer information).

[0023] In the artificial intelligence system of the present invention, the information processing unit, when it is unable to obtain the answer information because there is a lack of a concept value representing a concept (color) for identifying the concept (flower) of the knowledge (name of a flower species) asked about in the question information (What are spring flowers?), generates concept query information for inquiring about the concept value of the missing concept (color) and has concept query control means for causing the concept query information to be output from the information output unit, and after the concept query information is output from the information output unit, generates the answer information by taking into account information on the concept value (pink) representing the missing concept (color) input to the information input unit.

[0024] With this configuration, when the answer information cannot be obtained because there is a shortage of concept values ​​representing the concept (color) for identifying the concept (flower) of knowledge (name of flower species) in question in the question information (what is a spring flower), concept inquiry information for inquiring about the concept value of the missing concept (color) is generated and the concept inquiry information is output. Then, after the concept inquiry information is output, the answer information is generated by taking into account the element value (pink) information representing the missing concept (color) that is input. In this way, when there is a shortage of concept values ​​representing the concept (color) for identifying the concept (flower) of knowledge (name of flower species) in question in the question information (what is a spring flower), the AI ​​system itself can learn the missing information (subject-based learning: task-dependent type) and provide answer information that takes into account the results of the learning.

[0025] In the artificial intelligence system according to the present invention, the information processing unit includes a concept structure updating unit which updates the concept structure table based on new concept structure information when new concept structure information indicating that an existing concept (flower) already registered in the concept structure table is specified by one or more new concepts (country) is input to the information input unit, and a concept knowledge query unit which generates concept knowledge query information for querying the concept value of the new concept (country) in the new concept structure information corresponding to each concept value (cherry blossom, wisteria, lotus flower) of the existing concept (flower) included in the concept knowledge table, and outputs the concept knowledge query information to the information processing unit. The information processing device can be configured to include a concept knowledge query control means for causing the information output unit to output the information, a concept knowledge update means for updating the concept knowledge table based on a concept value (Japan) of one or more of the new concepts (countries) in the new concept structure information input to the information input unit after the concept knowledge query information is output from the information output unit, and a knowledge update means for updating the knowledge table so as to correspond the concept values ​​(cherry blossom, wisteria, lotus flower) of the existing concepts (flowers) to representative concept values ​​and the respective concept values ​​(Japan) of the one or more new concepts (countries) to subordinate concept values.

[0026] With this configuration, when new concept structure information indicating that an existing concept (flower) already registered in the concept structure table is specified by one or more new concepts (countries) is input, the concept structure table is updated based on the new concept structure information, and concept knowledge inquiry information is generated for inquiring about the concept values ​​of the one or more new concepts (countries) corresponding to the concept values ​​(cherry blossom, wisteria, lotus flower) of the existing concept (flower) included in the concept knowledge table, and the concept knowledge inquiry information is output. After that, the concept knowledge table is updated based on the concept values ​​(Japan) of the one or more new concepts (countries) input, and the knowledge table is updated so that the concept values ​​(cherry blossom, wisteria, lotus flower) of the existing concept (flower) are set as representative concept values, and the concept values ​​(Japan) of the one or more new concepts (countries) are set as subordinate concept values, and the knowledge table is updated so as to correspond to them. In this way, when information (new concept structure information) indicating knowledge about a new concept is provided, the artificial intelligence system itself can learn (subject learning: self-regulation type) the new concept structure as well as knowledge about the new concept structure. Effect of the Invention

[0027] According to the artificial intelligence system of the present invention, answer information is generated for question information using three tables, namely, a conceptual structure table, a conceptual knowledge table, and a knowledge table, so that the amount of knowledge information about concepts can be increased by updating (learning) the information in these three tables without spending a lot of time on learning. Also, by using these three tables in parallel with updating (learning) the three tables, answer information for question information can be determined, so that the system can be used as a brain even during learning. [Brief description of the drawings]

[0028] [Figure 1] FIG. 1 is a block diagram showing an artificial intelligence system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a conceptual structure table. [Diagram 3]FIG. 3 is a diagram showing an example of a concept knowledge table. [Figure 4] FIG. 4 is a diagram illustrating an example of the knowledge table. [Diagram 5] FIG. 5 is a diagram showing the relationship between the conceptual structure table, the conceptual knowledge table, and the knowledge table. [Figure 6] FIG. 6 is a sequence showing a sequence of using the conceptual structure table, conceptual knowledge table and knowledge table in the processing from question input to answer output in the information processing section of the artificial intelligence system. [Figure 7] FIG. 7 is a flowchart showing the flow of processing in the information processing unit of the artificial intelligence system. [Figure 8] FIG. 8 is a flowchart showing the flow of the concept knowledge determination process. [Figure 9] FIG. 9 is a diagram showing an example of the concept knowledge table updated by the concept knowledge determination process. [Figure 10] FIG. 10 is a diagram showing an example of a knowledge table updated by the concept knowledge determination process. [Figure 11] FIG. 11 is a flowchart showing the flow of the conceptual structure confirmation process. [Figure 12] FIG. 12 is a flowchart showing a flow (part 1) of the answering process. [Figure 13] FIG. 13 is a diagram showing an example of a concept knowledge table updated by the answering process according to the procedure shown in FIG. [Figure 14] FIG. 14 is a diagram showing an example of a knowledge table updated by the response process according to the procedure (cooperative awakening) shown in FIG. [Figure 15] FIG. 15 is a diagram showing an example of a concept knowledge table updated by the answering process according to the procedure (joint awakening) shown in FIG. [Figure 16] FIG. 16 is a flowchart showing the flow of the answering process (part 2). [Figure 17] FIG. 17 is a flowchart showing the flow of the related response process. [Figure 18]FIG. 18 is a sequence showing a sequence of using the conceptual structure table, conceptual knowledge table, and knowledge table during learning processing (task-dependent type) in the information processing unit of the artificial intelligence system. [Figure 19] FIG. 19 is a diagram showing an example of a state in which concepts are insufficient in the concept knowledge table. [Figure 20] FIG. 20 is a diagram showing an example of a state in which concepts are insufficient in the knowledge table. [Figure 21] FIG. 21 is a flowchart showing the flow of the learning process (task-dependent type). [Figure 22] FIG. 22 is a sequence showing a sequence of using the conceptual structure table, conceptual knowledge table, and knowledge table during learning processing (self-adjusting type) in the information processing section of the artificial intelligence system. [Diagram 23] FIG. 23 is a flowchart showing a flow (part 1) of the learning process (self-adjusting type). [Figure 24] FIG. 24 is a flowchart showing a flow (part 2) of the learning process (self-adjusting type). [Diagram 25] FIG. 25 is a diagram showing an example of a conceptual structure table updated by the learning process (self-adjusting type). [Figure 26] FIG. 26 is a diagram showing an example (part 1) of a knowledge table updated by the learning process (self-adjusting type). [Figure 27] FIG. 27 is a diagram showing an example (part 2) of the knowledge table updated by the learning process (self-adjusting type). [Figure 28] FIG. 28 is a diagram showing an example of a concept knowledge table updated by the learning process (self-adjusting type). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0029] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0030] An application system (for example, a Japanese flower encyclopedia) to which an artificial intelligence system according to an embodiment of the present invention is applied is configured as shown in FIG.

[0031] In FIG. 1, the application system 100 (Japanese flower encyclopedia) includes an artificial intelligence system 10 and an interface device 20 (PC, tablet terminal, smartphone, etc.) that provides an interface (UI) between the artificial intelligence system 10 and a user (person). The interface device 20 and the artificial intelligence system 10 are connected to be able to send and receive information directly by wire, directly by wireless, by wire and wireless, or via a predetermined network (Internet, etc.). The interface device 20 has an input sentence analysis function that performs sentence analysis of text information input by a user and extracts words from the text information, and an output sentence generation function that generates sentences from output information from the artificial intelligence system 10. The interface device 20 displays and / or outputs by voice the input information (text information) and the output information (text information) from the artificial intelligence system 10 in response to the input information, for example, in a format similar to the output format of a chatpot. The interface device 20 can also extract words from input sentence information input by voice from the user. Also, the functionality of the interface device 20 may be included in the artificial intelligence system 10 .

[0032] The artificial intelligence system 10 has an information processing unit 11, a memory unit 12, an input unit 13 (information input unit), and an output unit 14 (information output unit). The information processing unit 11 is configured by a computer operated by hardware resources such as a CPU and software resources. The memory unit 12 stores various data required for processing in the information processing unit 11, together with three tables (described in detail below): a conceptual structure table T1, a conceptual knowledge table T2, and a knowledge table T3.

[0033] The concept structure table T1 is a table in which concepts defined as representative concepts are registered and which associates the representative concepts with subordinate concepts defined as concepts that specify the representative concepts, and is configured, for example, as shown in Fig. 2. In the concept structure table T1 shown in Fig. 2, the concept "color", the concept "season", the concept "flower", and the concept "event" are each registered as representative concepts, and the representative concept "flower" is associated with two subordinate concepts "color" and "season" that are defined as concepts that specify the concept "flower", and the representative concept "event" is associated with one subordinate concept "flower" that is defined as a concept that specifies the concept "event".

[0034] The concept knowledge table T2 is a table that associates concepts with concept values ​​as information that represents knowledge about the concepts, and is configured, for example, as shown in Fig. 3. In the concept knowledge table T2 shown in Fig. 3, the concept "color" is associated with a concept value "pink" as information that represents knowledge about the concept "color". Similarly, in the concept knowledge table T2, the concept "color" is associated with the concept value "purple", the concept "season" is associated with the concept value "spring", the concept "flower" is associated with the concept value "cherry blossom", the concept "flower" is associated with the concept value "wisteria", the concept "flower" is associated with the concept value "lotus flower", and the concept "event" is associated with the concept value "entrance ceremony".

[0035] The knowledge table T3 is a table that associates representative concept values ​​as information representing knowledge about the representative concept described above with subordinate concept values ​​as information representing knowledge of subordinate concepts defined as concepts that identify the representative concept, and is configured, for example, as shown in Fig. 4. In the knowledge table T3 shown in Fig. 4, a representative concept value "cherry blossom" as information representing knowledge about the representative concept "flower" is associated with two subordinate concept values ​​"pink" and "spring" as information representing knowledge of two subordinate concepts "color" and "season" (see Fig. 4: concept structure table) that are defined as concepts that identify the representative concept "flower". Similarly, in the knowledge table T3, the representative concept value "wisteria" is associated with two subordinate concept values ​​"purple" and "spring", the representative concept value "lotus flower" is associated with two subordinate concept values ​​"pink" and "spring", and the representative concept value "entrance ceremony" is associated with one subordinate concept value "cherry blossom".

[0036] In the above-mentioned concept knowledge table T2 (see FIG. 3), an awakening flag (first awakening concept value management means, third awakening concept value management means) is associated with each concept value, which indicates whether the concept value is managed as an awakening concept value described later, specifically, has different states (1, 0) depending on whether the concept value is an awakening concept value or not. In addition, in the above-mentioned knowledge table T3 (see FIG. 4), an awakening flag (second awakening concept value management means) is associated with each subordinate concept value, which indicates whether the subordinate concept value is managed as an awakening concept value described later, specifically, has different states (1, 0) depending on whether the subordinate concept value is an awakening concept value or not. Furthermore, in the knowledge table T3 (see FIG. 4), an answer control flag is associated with each representative concept value, which manages whether the representative concept value is used to generate answer information described later, specifically, has different states (1, 0) depending on whether the representative concept value is used to generate the answer information or not.

[0037] In the artificial intelligence system 10, the three tables mentioned above (conceptual structure table T1, conceptual knowledge table T2, and knowledge table T3) are used to manage various concepts related to "flowers" in the "Japanese Flower Encyclopedia" (application system 100) and knowledge (conceptual values) about those concepts.

[0038] The above-mentioned three tables (concept structure table T1, concept knowledge table T2, and knowledge table T3) have the relationship shown in FIG.

[0039] As shown in Fig. 5, in the concept knowledge table T2, each concept is identified by a concept ID. In the concept structure table T1, each representative concept (e.g., "flower" in Fig. 2) is identified by a representative concept ID having the same value as the concept ID of the corresponding concept (e.g., "flower" in Fig. 3) in the concept knowledge table T2, and each subordinate concept (e.g., "color" in Fig. 2) is identified by a subordinate concept ID having the same value as the concept ID of the corresponding concept (e.g., "color" in Fig. 3) in the concept knowledge table T2.

[0040] In the knowledge table T3, the representative concept (e.g., "flower") corresponding to each representative concept value (e.g., "cherry blossom") and the subordinate concept (e.g., "color") corresponding to each subordinate concept value (e.g., "pink"), which are omitted in Fig. 4, are identified by the representative concept ID and subordinate concept ID, respectively. Each of the representative concept ID and subordinate concept ID in the knowledge table T3 is the same value as the concept ID of the corresponding concept in the concept knowledge table T2. In addition, each of the representative concept value (e.g., "cherry blossom") and subordinate concept value (e.g., "pink") in the knowledge table T3 is linked to the concept value in the knowledge structure table T2 (e.g., "cherry blossom" and "pink" in Fig. 3).

[0041] 5, the awakening flag in the concept knowledge table T2 and the awakening flag in the knowledge table T3 are updated to the state (1, 0) in cooperation with each other. However, when the answer control flag in the knowledge table T3 (see FIG. 4) is in the state (0) where the corresponding representative concept is not used for generating answer information, the awakening flag of the subordinate concept value corresponding to the representative concept value is maintained in the state (0) regardless of the state of the awakening flag of the corresponding concept value in the concept knowledge table T2.

[0042] Next, the operation of the information processing unit 11 of the artificial intelligence system 10 will be described.

[0043] The information processing unit 11 executes processing according to the procedure shown in FIG. 7 while using three tables (conceptual structure table T1, concept knowledge table T2, and knowledge table T3) stored in the storage unit 12 according to the sequence shown in FIG.

[0044] When a user inputs information related to a question (e.g., "What is a pink spring flower?") to the interface device 20, the interface device 20 displays and outputs the question by voice, and performs a sentence analysis on the question to extract the wording used in the question. For example, the words "pink", "spring", and "flower" are extracted from the question. Then, the interface device 20 outputs question information in a predetermined format for asking about knowledge (flower species name) of the concept "flower" by presenting element values ​​"pink" and "spring" that represent elements that identify the knowledge (flower species name) based on the extracted wording. The question information output from the interface device 20 is input to the input unit 13 of the artificial intelligence system 10, and the question information is provided to the information processing unit 11 of the artificial intelligence system 10. Hereinafter, the operation of the information processing unit 11 will be described using an example in which question information (corresponding to the question "What is a pink spring flower?") for asking about knowledge (flower species name) of the concept "flower" by presenting two element values ​​"pink" and "spring" that identify the knowledge (flower species name) is provided.

[0045] As shown in Fig. 7, the information processing unit 11 checks whether question information is acquired (S11). When the question information is acquired via the input unit 13 (YES in S11), the information processing unit 11 extracts the concept "flower" as a query concept and element values ​​"pink" and "spring" from the question information (S12). Then, the information processing unit 11 performs a concept knowledge determination process S13 (concept knowledge determination means) using a concept knowledge table T2 (see Fig. 3) and a knowledge table T3 (see Fig. 4) stored in the storage unit 12.

[0046] The concept knowledge determination process S13 is performed according to FIG.

[0047] In Fig. 8, the information processing unit 11 designates a concept value (S132) while checking that all concept values ​​have not been designated in the concept knowledge table T2 (see Fig. 3) (NO in S131). Then, the information processing unit 11 judges whether the designated concept value corresponds to any of the element values ​​("pink" or "spring") included in the question information (S133: step (1) in Fig. 6). If the designated concept value does not correspond to any of the element values ​​(NO in S133), the information processing unit 11 returns to step S131, checks that all concept values ​​have not been designated in the concept knowledge table T2 (NO in S131), designates the next concept value (S132), and performs the judgment again (S133). On the other hand, if the specified concept value corresponds to the element value (YES in S133), the information processing unit 11 updates the awakening flag corresponding to the specified concept value in the concept knowledge table T2 (see FIG. 9) to the state “1” (ON) (S134: step (2) in FIG. 6), and updates the awakening flag corresponding to the same subordinate concept value as the specified concept value in the knowledge table T3 (see FIG. 10) to the state “1” (S135: step (3) in FIG. 6).

[0048] After that, the information processing unit 11 repeatedly executes the same process (S131 to S135). Then, when the process for all the concept values ​​in the concept knowledge table T2 is completed (YES in S131), the information processing unit 11 further judges whether or not all the element values ​​("pink" and "spring") extracted from the question information have been processed (S136). If all the element values ​​have been processed (YES in S136), the information processing unit 11 ends the concept knowledge judgment process S13 and proceeds to the next concept structure confirmation process (S14: see FIG. 7).

[0049] As a result of the above-mentioned process, in the concept knowledge table T2, as shown in FIG. 9, each of the arousal flags corresponding to the concept values ​​"pink" and "spring" corresponding to the extracted element values ​​"pink" and "spring" is updated to the state "1" (ON). As a result, the information processing unit 11 recognizes that there is a correspondence between the concept values ​​"pink" and "spring" corresponding to the element values ​​("pink", "spring"), respectively, and manages each of the concept values ​​"pink" and "spring" as the arousal concept value of the corresponding concept "color" or "season" (first arousal concept value management means). Also, in the knowledge table T3, as shown in FIG. 10, each of the arousal flags corresponding to the subordinate concept values ​​"pink" and "spring" corresponding to the extracted element values ​​"pink" and "spring" is updated to the state "1" (ON). As a result, the information processing unit 11 manages each of the subordinate concept values ​​"pink" and "spring" as the arousal concept value in the knowledge table T3 (second arousal concept value management means).

[0050] If processing cannot be completed for all element values ​​(NO in S136 in FIG. 8), it means that there is no correspondence between a concept value corresponding to a certain element value and a concept in the concept knowledge table T2, and the information processing unit 11 determines that the artificial intelligence system 10 does not have enough knowledge to answer the input question ("What is that pink spring flower?") (S137), returns to FIG. 7, resets the awakening flag (S19), and ends processing.

[0051] Returning to FIG. 7, when the above-mentioned concept knowledge determination process (S13) is completed, the information processing unit 11 performs a concept structure confirmation process S14 (concept structure confirmation means) using the concept knowledge table T2 (see FIG. 9) and the concept structure table T1 (see FIG. 2) stored in the memory unit 12.

[0052] The conceptual structure confirmation process S14 is performed according to FIG.

[0053] In FIG. 11, the information processing unit 11 acquires the concepts "color" and "season" corresponding to the concept values ​​"pink" and "spring" managed as arousal concept values ​​(the arousal flag is in the state "1" (ON)) in the concept knowledge table T2 (see FIG. 9) (S141). Then, the information processing unit 11 judges whether or not there is a correspondence between the representative concept and the subordinate concept in the concept structure table T1 (see FIG. 2), in which the query concept "flower" is a representative concept and the concepts "color" and "season" corresponding to the concept values ​​"pink" and "spring" managed as the arousal concept values ​​are subordinate concepts (S142: concept structure confirmation means: step (4) in FIG. 6). Here, if there is a correspondence between the representative concept "flower" and the subordinate concepts "color" and "season" (YES in S142), it is determined that the artificial intelligence system 10 can answer the query, and the information processing unit 11 proceeds to the next answer process S15 (see FIG. 7: answer generation means).

[0054] On the other hand, if the correspondence between the representative concept (query concept "flower") and the subordinate concept does not exist in the conceptual structure table T1 (NO in S142), it is determined that the concept information is insufficient to answer the question, and the information processing unit 11 judges whether learning, which will be described later, is possible according to a predetermined condition (e.g., a condition regarding the number of insufficient concepts, etc.) (S143). If learning is possible (YES in S143), the information processing unit 11 proceeds to learning process S20 (see FIG. 7). Details of the learning process S20 will be described later.

[0055] If it is determined that learning is not possible (NO at S143), the information processing unit 11 determines that the artificial intelligence system 10 does not have enough conceptual information to answer the input question (S144), and returns to FIG. 7 to reset the awakening flag (S19) and terminate the process.

[0056] The reply process S15 (see FIG. 7) is performed according to the procedures shown in FIG. 12 and FIG.

[0057] In Fig. 12, the information processing unit 11 designates a representative concept value (S152) while confirming that the designation of all representative concept values ​​is not completed in the knowledge table T3 (see Fig. 10) (NO in S151). Then, the information processing unit 11 judges whether all subordinate concept values ​​corresponding to the designated representative concept value are managed as arousal concept values ​​(arousal flag state "1" (ON)) in the knowledge table T3 (see Fig. 10) (S153). If any subordinate concept value corresponding to the designated representative concept value is not managed as an arousal concept value in the knowledge table T3 (NO in S153: for example, the subordinate concept value "purple" corresponding to the designated representative concept value "wisteria" is not managed as an arousal concept value), the information processing unit 11 returns to step S151, confirms that the designation of all representative concept values ​​is not completed in the knowledge table T3 (NO in S151), designates the next representative concept value (S152), and performs the judgment again (S153).

[0058] On the other hand, when all subordinate concept values ​​(e.g., "pink", "spring") corresponding to the designated representative concept value (e.g., "cherry blossom") in the knowledge table T3 (see FIG. 10) are managed as arousal concept values ​​(awakening flag state "1" (ON)) (YES in S153), the information processing unit 11 confirms that the answer control flag (answer control means) corresponding to the designated representative concept value in the knowledge table T3 (see FIG. 10) is in the state "1" (ON) (YES in S154), and updates the awakening flag corresponding to the designated representative concept value (e.g., "cherry blossom") and the same concept value (e.g., "cherry blossom") in the concept knowledge table T2 (see FIG. 13) to the state "1" (ON) (S155: step (5) in FIG. 6).

[0059] In addition, if the answer control flag corresponding to the designated representative concept value in the knowledge table T3 is not in the state "1" (is in the state "0") (NO in S154), the information processing unit 11 assumes that the designated representative concept value is not used for generating answer information, and performs similar processing (S151 to S154) for the next designated representative concept value without updating the awakening flag corresponding to the same concept value as the designated representative concept value in the concept knowledge table T2 (S155).

[0060] When the awakening flag corresponding to the same concept value (e.g., "cherry blossom") as the designated representative concept value in the concept knowledge table T2 (see FIG. 13) is updated to the state "1" (ON) (S155), the information processing unit 11 performs a process (co-awakening process: S156-S158) to detect a concept value that will be awakened (co-awakened) in conjunction with the awakening (updating of the awakening flag to the state "1" (ON)) of that concept value (e.g., "cherry blossom").

[0061] In this cooperative awakening process, the information processing unit 11 first judges whether or not the concept value (e.g., "cherry blossom") whose corresponding awakening flag is set to the state "1" (ON) in the concept knowledge table T2 (see FIG. 13) and the knowledge table T3 (see FIG. 10) have the same subordinate concept value (S156). If there is no such subordinate concept value in the knowledge table T3 (NO in S156), the information processing unit 11 interrupts the cooperative awakening process, returns to step 151, and performs the same process (S151 to S156) for the next representative concept value. On the other hand, if there is a concept value (e.g., "cherry blossom") whose corresponding awakening flag is set to the state "1" (ON) in the concept knowledge table T2 (see FIG. 13) and the same subordinate concept value (e.g., "cherry blossom") in the knowledge table T3 (see FIG. 10) (YES in S156), the information processing unit 11 updates the awakening flag corresponding to the subordinate concept value (e.g., "cherry blossom") in the knowledge table T3 to the state "1" (ON) (S157). Then, the information processing unit 11 updates the awakening flag corresponding to the representative concept value (e.g., "entrance ceremony") corresponding to the subordinate concept value (e.g., "cherry blossom") whose awakening flag is updated to the state "1" (ON) in the knowledge table T3 and the same concept value (e.g., "entrance ceremony") in the concept knowledge table T2 (see FIG. 13) to the state "1" (ON) (S158, see FIG. 15).

[0062] After that, the information processing unit 11 repeatedly executes the above-mentioned processes (S151 to S158). Then, when the processes for all the representative concept values ​​in the knowledge table T2 are completed (YES in S151), the information processing unit 11 proceeds to the process of FIG.

[0063] 12, the information processing unit 11 processes the concept value "cherry blossom" (corresponding to the representative concept value "cherry blossom" in the knowledge table T3) corresponding to the concept "flower" and the concept value "lotus grass" (corresponding to the representative concept value "lotus grass" in the knowledge table T3) corresponding to the concept "flower" in the concept knowledge table T2, and updates the awakening flags to "1" (ON) as shown in Fig. 13. As a result, the information processing unit 11 manages the concept values ​​"cherry blossom" and "lotus grass" as the awakening concept values ​​of the corresponding concept "flower" in the concept knowledge table T2 (third awakening concept value management means).

[0064] In addition, in association with the update of the awakening flag corresponding to the concept value "cherry blossom" in the concept knowledge table T2 to the state "1" (ON) (awakening), in the knowledge table T3, as shown in FIG. 14, the awakening flag corresponding to the subordinate concept value "cherry blossom" which is the same as the concept value "cherry blossom" is updated to the state "1" (ON) (awakening). And, in association with this, the awakening flag corresponding to the representative concept value "entrance ceremony" corresponding to the awakened subordinate concept value "cherry blossom" in the knowledge table T3 and the same concept value "entrance ceremony" in the concept knowledge table T2 is updated to the state "1" (ON) as shown in FIG. 15 (co-awakening). As a result, the information processing unit 11 further manages the concept value "entrance ceremony" as the awakening concept value of the corresponding concept "event" (concept value of co-awakening) in the concept knowledge table T2.

[0065] In the state where the concept knowledge table T2 and the knowledge table T3 are updated as described above, the information processing unit 11, which has moved to the process of Fig. 16, specifies the query concept "flower" (S160) while confirming that all the query concepts "flower" have not been specified in the concept knowledge table T2 (see Fig. 15) (NO in S159). Then, the information processing unit 11 determines whether the awakening flag of the concept value (e.g., "cherry blossom") corresponding to the specified query concept "flower" is in the state "1" (ON) in the concept knowledge table T2 (see Fig. 15), that is, whether the concept value (e.g., "cherry blossom") corresponding to the specified query concept "flower" is managed as an awakening concept value (S161).

[0066] If the awakening flag corresponding to the designated query concept "flower" is not in the state "1" (NO) (is in the state "0") (NO in S161: for example, see the awakening flag corresponding to the concept value "wisteria"), the information processing unit 11 returns to step S159, confirms that the designation of all query concepts "flower" has not been completed in the concept knowledge table T2 (see FIG. 15) (NO in S159), designates the next query concept "flower" (S160), and performs the determination again (S161). On the other hand, if the awakening flag corresponding to the designated query concept "flower" is in the state "1" (ON) in the concept knowledge table T2 (YES in S161: for example, see the awakening flag corresponding to the concept value "cherry blossom"), the information processing unit 11 adds the concept value (for example, "cherry blossom") corresponding to the designated query concept "flower" to the list as an answer concept value (S162: answer determination means).

[0067] Thereafter, the information processing unit 11 repeatedly executes the same process (S159 to S162) until all query concepts "flower" in the concept knowledge table T2 are specified. As a result, the concept values ​​"cherry blossom" and "lotus flower" in the concept knowledge table T2 (see FIG. 15) are listed as answer concept values. Then, when the process for all query concepts "flower" in the concept knowledge table T2 is completed (YES in S159), the information processing unit 11 generates answer information to the question "What is a pink spring flower?" based on the concept values ​​("cherry blossom" and "lotus flower") listed as answer concept values ​​(S163: step (6) in FIG. 6).

[0068] Returning to FIG. 7, following the answer process S15 described above, the information processing unit 11 performs a related answer process S17 (related answer generating means). The related answer process S17 is a process for generating related answer information based on a concept value (related concept value) related to the answer concept value obtained by the answer process S15 described above. In this related answer process S17, a concept value (e.g., "entrance ceremony") awakened in conjunction with the awakening of a concept value (e.g., "cherry blossom") obtained as the answer concept value is obtained as the related concept value. This related answer process S17 is performed according to the procedure shown in FIG.

[0069] In Fig. 17, the information processing unit 11 specifies an answer concept value (S172) while confirming that all answer concept values ​​(e.g., "cherry blossom" and "lotus flower") obtained in the answer process S15 have not been specified (NO in S171). Then, the information processing unit 11 judges whether or not there is a subordinate concept value that is the same as the designated answer concept value among the subordinate concept values ​​whose awakening flags are in the state "1" (ON) in the knowledge table T3 (see Fig. 14) (S173). If there is no subordinate concept value that is the same as the designated answer concept value (e.g., "lotus flower") in the knowledge table T3 (NO in S173), the information processing unit 11 returns to step 171, confirms that all answer concept values ​​have not been specified (NO in S171), specifies the next answer concept value (S172), and performs the judgment (S173) again.

[0070] On the other hand, if a subordinate concept value (e.g., "cherry blossom") that is the same as the specified answer concept value (e.g., "cherry blossom") exists in the knowledge table T3 (see FIG. 14) (YES in S173), the information processing unit 11 confirms that the answer control flag corresponding to the subordinate concept value is in the state "1" (ON) (YES in S174), and then determines whether or not there is a correspondence between the subordinate concept (e.g., "flower") and the representative concept (e.g., "event") identified by the subordinate concept ID and representative concept ID corresponding to the subordinate concept value (e.g., "cherry blossom") and its corresponding representative concept value (e.g., "entrance ceremony") in the knowledge table T3 (see FIG. 1) in the concept structure table T1 (see FIG. 2) (S175).

[0071] If there is a correspondence between the representative concept and the subordinate concept (e.g., a correspondence between "event" and "flower") in the conceptual structure table T1 (see FIG. 2) (YES in S175), the information processing unit 11 specifies a concept (e.g., "event") that is the same as the representative concept (e.g., "event") in the conceptual knowledge table T2 (see FIG. 15) (S176), and determines whether the awakening flag of the conceptual value (e.g., "entrance ceremony") corresponding to that concept (e.g., "event") is in the "1" (ON) state (S177). Here, if the awakening flag of the concept value (e.g., "entrance ceremony") corresponding to the specified concept (e.g., "event") in the concept knowledge table T2 (see Figure 15) is in the state "1" (ON) (YES in S177), that is, if the concept value (e.g., "entrance ceremony") is managed as the awakening concept value (concept value of linked awakening) of the corresponding concept (e.g., "event"), the information processing unit 11 adds the concept value (e.g., "entrance ceremony") to the list as a related concept value in correspondence with the answer concept value (e.g., "cherry blossoms") (S178).

[0072] In addition, in step 174, if the answer control flag in the knowledge table T3 is not in the state "1" (state "0": NO in S174), in step S175, if there is no correspondence between the representative concept and the subordinate concept in the conceptual structure table T1 (NO in S175), and in step 177, if the awakening flag is not in the state "1" (state "0": NO in step S177), the information processing unit 11 returns to step S171 and continues processing, since it is not possible to obtain a related concept value.

[0073] The information processing unit 11 repeatedly executes the same processing (S171 to S178) until the designation of all the listed answer concept values ​​is completed (YES in S1712). As a result, the concept value "entrance ceremony" in the concept knowledge table T2 (see FIG. 15) is listed in a state in which it is associated with the answer concept value "cherry blossom" as a related concept value. Then, when the processing for all the answer concept values ​​("cherry blossom", "lotus flower") is completed (YES in S171), the information processing unit 11 judges whether or not there is a listed related concept value (S179). Then, when there is a listed related concept value (YES in S179), the information processing unit 11 generates related answer information (for example, information representing "cherry blossoms are associated with entrance ceremonies") based on the concept value "entrance ceremony" listed in a state in which it is associated with the answer concept value "cherry blossom" as a related concept value (S180). After that, the information processing unit 11 proceeds to the answer output processing S18 (see FIG. 7).

[0074] On the other hand, in the above-mentioned related answer processing S17 (see FIG. 7 and FIG. 17), if a related concept value cannot be obtained (NO in S179), the information processing unit 11 proceeds to the answer output processing S18 (see FIG. 7) without generating related answer information.

[0075] Returning to Fig. 7, in the answer output process S18, the information processing unit 11 outputs the answer information ("cherry blossom", "lotus flower") generated as described above (see Figs. 12 and 16) from the output unit 14 (information output unit) to the interface device 20. The interface device 20, which has acquired the answer information ("cherry blossom", "lotus flower"), outputs information "cherry blossom and lotus flower" representing the answer to the question "What are the pink spring flowers?" by displaying text, sound, etc. This allows the user to obtain the information "cherry blossom and lotus flower" as the answer to the question "What are the pink spring flowers?" from the application system 100 (Japanese flower encyclopedia) including the artificial intelligence system 10.

[0076] Furthermore, if the related answer information exists (related answer process S17: see FIG. 17), the information processing unit 11 causes the output unit 14 to output the related answer information ("Entrance ceremony") associated with the answer information ("Cherry blossoms") together with the answer information ("Cherry blossoms"), to the interface device 20 by the answer output process S18. In this case, the interface device 20 outputs information "Cherry blossoms and lotus flowers" representing the answer to the question "What pink spring flowers are there?" by displaying characters, sound, etc., as described above, and also outputs information related to the answer "Cherry blossoms" by displaying characters, sound, etc., for example, "Cherry blossoms are associated with entrance ceremonies". As a result, the user can obtain information "Cherry blossoms are associated with entrance ceremonies" related to the answer "Cherry blossoms" from the application system 100 (Japanese flower encyclopedia) including the artificial intelligence system 10, together with information "Cherry blossoms and lotus flowers" as the answer to the question "What pink spring flowers are there?".

[0077] As described above, when it is determined in the conceptual structure confirmation process S14 after the concept knowledge determination process S13 that the conceptual information is insufficient to answer the question (NO in S142 in FIG. 11), the information processing unit 11 executes the learning process S20. This learning process S20 corresponds to the "task-dependent" subjective learning of "tackling the task with interest and concern."

[0078] In the learning process S20, the information processing unit 11 executes the process according to the procedure shown in Fig. 21 while using three tables (conceptual structure table T1, concept knowledge table T2, and knowledge table T3) stored in the storage unit 12 according to the sequence shown in Fig. 18. Then, after acquiring information on the missing concept by the learning process S20, the information processing unit 11 executes the conceptual structure confirmation process S14 (see Fig. 11) and the answer process S15 (see Figs. 12 and 16).

[0079] For example, the following describes a case where a question ("What are spring flowers?") is input, which asks about knowledge (flower species names) about the concept "flower" by presenting an element value "spring" that indicates an element that identifies that knowledge. In this case, it is determined that information about the concept "color" is insufficient to obtain an answer to the query concept "flower" (NO in S142 in FIG. 11), and the learning process S30 is executed.

[0080] By the above-mentioned concept knowledge judgment process S13 (see Fig. 7 and Fig. 8), the awakening flag corresponding to the concept value "Spring" corresponding to the element value "Spring" included in the question information is updated to the state "1" (ON) in the concept knowledge table T2 as shown in Fig. 19, and the awakening flag corresponding to the subordinate concept value "Spring" which is the same as the concept value "Spring" is updated to the state "1" (ON) in the knowledge table T3 as shown in Fig. 20. In this state, it is judged that the concept ("color") is insufficient (S142 in Fig. 11) (steps (1) to (4) in Fig. 18).

[0081] In FIG. 21, the information processing unit 11 identifies the concept "color" of the concept value (knowledge) that is missing to obtain an answer (knowledge) about the query concept "flower" (S201), and generates concept query information (e.g., "What is the color?") for inquiring about the concept value of the concept "color" (S202: step (5) in FIG. 18). Then, the information processing unit 11 causes the output unit 14 to output the concept query information (concept query control means). The concept query information output from the output unit 14 is provided to the interface device 20. Thereafter, the information processing unit 11 enters a waiting state for a response to the concept query from the interface device 20 (S204).

[0082] The interface device 20, which has acquired the concept inquiry information, outputs information representing the inquiry, "What color is it?", by displaying characters, by voice, or the like. The user who has asked the question, "What are the spring flowers?" inputs a concept value representing knowledge of the concept "color", for example, the concept value "pink", in response to the inquiry, "What color is it?", output from the interface device 20. The interface device 20 then performs a sentence analysis on the answer and extracts the wording used in the answer. For example, the wording "pink" is extracted from the answer. The interface device 20 then outputs the wording "pink" as information on the answer concept value. The answer concept value "pink" output from the interface device 20 is provided to the information processing unit 11 via the input unit 13.

[0083] As described above, when the information processing unit 11, which is in a state of waiting for an answer (S204), acquires the answer concept value "pink" from the interface device 20 (YES in S204), it judges whether or not a concept value identical to the answer concept value "pink" exists in the concept knowledge table T2 (see FIG. 19) (S205). If a concept value identical to the answer concept value "pink" does not exist (NO in S205), the information processing unit 11 determines that the application system 100 (Japanese flower encyclopedia) including the artificial intelligence system 10 does not have enough knowledge to answer the initial question ("What are the spring flowers?") (S208), returns to FIG. 7, resets the awakening flag (S19), and ends the process. On the other hand, if a concept value "pink" that is the same as the answer concept value "pink" exists in the concept knowledge table T2 (see Figure 19) (YES in S205), the awakening flag corresponding to the concept value "pink" that is the same as the answer concept value in the concept knowledge table T2 is updated to the state "1" (ON) (S206: see Figure 9), and the awakening flag corresponding to the subordinate concept value that is the same as the concept value "pink" in the knowledge table T3 is updated to the state "1" (ON) (S207: see Figure 10).

[0084] Thereafter, the information processing unit 11 executes the above-mentioned concept structure confirmation process S14 (see Figs. 7 and 11) by taking into consideration the concept value "pink" of the concept "color" newly managed as an arousal concept value (awakening flag ON) in the concept knowledge table T2 (see Fig. 9). Thereafter, the information processing unit 11 executes the above-mentioned answer process S15 (see Figs. 12 and 16) by taking into consideration the subordinate concept value "pink" newly managed as an arousal subordinate concept value (awakening flag ON) in the knowledge table T3 (see Fig. 10). As a result, the information processing unit 11 generates answer information based on the concept value "cherry blossom" and the concept value "lotus flower" as described above, and also generates related answer information based on the concept value "entrance ceremony" related to the concept value "cherry blossom" included in the answer information.

[0085] According to the above-mentioned artificial intelligence system 10, answer information (e.g., "It is 'cherry blossom' and 'lotus flower'") is generated for question information (e.g., "What is a pink spring flower?") using three tables, namely, the concept structure table T1 (see FIG. 2), the concept knowledge table T2 (see FIG. 3), and the knowledge table T3 (see FIG. 4), so that the amount of knowledge information about concepts can be increased by updating (learning) the three tables T1, T2, and T3 without spending a lot of time on learning. Also, according to the above-mentioned artificial intelligence system 10, answer information for question information can be determined by using the three tables in parallel with updating (learning) the three tables T1, T2, and T3, so that the system can be used as a brain without stopping its function during the learning process.

[0086] Furthermore, related answer information related to the answer information is generated based on the conceptual value ("Entrance Ceremony") that is awakened (awakening flag is updated to state "1" (ON)) in conjunction with the awakening of the conceptual value ("Cherry Blossoms") that is the basis of the answer information (awakening flag is updated to state "1" (ON)), so that knowledge (entrance ceremony) related to the answer information (cherry blossoms) determined in response to the question information can be provided as an answer (related answer information).

[0087] Furthermore, if the answer information cannot be obtained because there is a shortage of concept values ​​("color") representing the concept (flower) of the knowledge (flower species name) asked in the question information ("What are spring flowers?"), the learning process S20 (corresponding to task-dependent subjective learning) can output inquiry information about the shortage concept values ​​and then receive the shortage concept values ​​as a response to the inquiry. Therefore, appropriate answer information can be generated by taking the shortage concept values ​​into account.

[0088] The updating (learning) of the three tables (concept structure table T1, concept knowledge table T2, knowledge table T3) can be performed by the information processing unit 11 or an external processing unit accessing the storage unit 12. The updating of these three tables is not limited to this, and the updating of the three tables (concept structure table T1, concept knowledge table T2, knowledge table T3) can be performed by other learning processes (corresponding to self-adjusting subjective learning).

[0089] The learning process (corresponding to self-adjusting subjective learning) is performed, for example, in accordance with the procedures shown in FIGS. 23 and 24 while using the three tables in accordance with the sequence shown in FIG.

[0090] 23, the information processing unit 11 acquires new concept structure information together with a learning request from the interface device 20 via the input unit 13 (S31). This new concept structure information indicates that an existing concept (e.g., "flower") already registered in the concept structure table T3 is specified by one or more new concepts (e.g., "country"). Below, an example will be described in which new concept structure information "Flowers have countries" is acquired, indicating that the existing concept "flower" is specified by the new concept "country."

[0091] The information processing unit 11, which has acquired the new concept structure information, extracts the new concept "country" and the existing concept "flower" from the new concept structure information (S32). Then, as shown in FIG. 25, the information processing unit 11 registers the new concept "country" as a representative concept in the concept structure table T1, and registers the correspondence between the representative concept and the subordinate concept, with the new concept "country" as a subordinate concept and the existing concept "flower" as the representative concept (S33: concept structure update means: see step (1) in FIG. 22). Also, the information processing unit 11 acquires the concept values ​​"cherry blossom", "wisteria", and "lotus flower" of the existing concept "flower" by referring to the concept knowledge table T2, and registers the concept values ​​"cherry blossom", "wisteria", and "lotus flower" as new representative concept values ​​in the knowledge table T3, as shown in FIG. 26 (S34: see step (2) in FIG. 22). In addition, in the knowledge table T3 (see FIG. 26), the awakening flag is set to the state "0" and the answer control flag is set to the state "1" as default.

[0092] The information processing unit 11 then proceeds to step S35 shown in Fig. 24, and while checking that all existing concepts "flower" corresponding to the new concept "country" have not been specified in the concept knowledge table T2 (see Fig. 3) (NO in S35), specifies a concept value (e.g., "cherry blossom") corresponding to the existing concept "flower" (S36: see step (3) in Fig. 22). Then, the information processing unit 11 generates concept knowledge inquiry information (e.g., "What is the country of cherry blossoms?") for inquiring about the concept value of the new concept "country" for the specified concept value (e.g., "cherry blossom") (S37). Then, the information processing unit 11 causes the output unit 14 to output the concept knowledge inquiry information (S39: concept knowledge inquiry control means: see step (4) in Fig. 22). After that, the information processing unit 11 enters a waiting state for a response concept value for the concept knowledge inquiry information (S40).

[0093] The interface device 20, which has acquired the concept knowledge inquiry information, outputs information representing the inquiry (e.g., "What is the country of cherry blossoms?") by displaying characters, speaking, or the like. The user, who has acquired the information representing the inquiry output from the interface device 20, inputs an answer to the inquiry (e.g., "Japan") to the interface device 20. The answer is then provided from the interface device 20 to the information processing unit 11 via the input unit 13 as an answer concept value (e.g., "Japan").

[0094] As described above, when the information processing unit 11, which is in a waiting state for an answer concept value (S40), acquires an answer concept value (e.g., "Japan") from the interface device 20 (YES in S40), it associates the answer concept value (e.g., "Japan") as a subordinate concept value with the same representative concept value as the designated concept value (e.g., "cherry blossom") in the knowledge table T3 as shown in Fig. 27 (S41: see step (5) in Fig. 22). Then, the information processing unit 11 associates the answer concept value ("Japan") with the new concept "Country" in the concept knowledge table T2 as shown in Fig. 28 (S42: concept knowledge update means: see step (6) in Fig. 22).

[0095] After that, the information processing unit 11 repeatedly executes the same processes (S35 to S42) as described above until all existing concepts "flower" corresponding to the new concept "country" are specified. Then, when all existing concepts "flower" in the concept knowledge table T2 (see FIG. 28) are specified (YES in S35), the information processing unit 11 ends this learning process.

[0096] Through the learning process described above, the three tables T1, T2, and T3 are updated so that in the conceptual structure table T1, a correspondence between the representative concept "flower" and the subordinate concept "country" is added along with the representative concept "country", as shown in FIG. 25; in the knowledge table T3, a correspondence between each of the representative concept values ​​"cherry blossom", "wisteria", and "lotus flower" and the subordinate concept value "Japan" is added, as shown in FIG. 27; and in the conceptual knowledge table T2, a correspondence between the new concept "country" and the conceptual value "Japan" is added, as shown in FIG. 28.

[0097] According to this type of learning process, when information expressing knowledge about the new concept of "country" (new conceptual structure information: "flowers" have countries") is provided, the artificial intelligence system 10 can learn (subject-driven learning: self-regulating type) the new conceptual structure as well as knowledge about the new conceptual structure (for example, cherry blossoms are Japan, wisteria is Japan, lotus flower is Japan).

[0098] The application system 100 including the artificial intelligence system 10 described above is related to an "illustrated book of Japanese flowers," but is not limited to this. Any application system may be used as long as it is capable of inputting question information for asking about knowledge of a certain concept by presenting element values ​​representing one or more elements that specify the knowledge, and answering questions based on the question information.

[0099] Furthermore, the specific process of generating answer information (cherry blossoms) to question information (What is the pink spring flower?) using the three tables (concept structure table T1, concept knowledge table T2, and knowledge table T3) is not limited to the above.

[0100] Although the embodiment of the present invention has been described above, this embodiment is presented as an example and is not intended to limit the scope of the invention. These novel embodiments described above can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims. [Industrial Applicability]

[0101] The artificial intelligence system according to the present invention is useful because it does not require a large amount of time for learning and can be used as a brain even during learning. [Explanation of symbols]

[0102] 10 Artificial Intelligence Systems 11 Information Processing Section 12 Storage section 13 Input section 14 Output section 20 Interface device 100 Application System (Japanese Flower Encyclopedia)

Claims

1. a concept structure table that associates a concept defined as a representative concept with a subordinate concept defined as a concept that specifies the representative concept; a concept knowledge table that associates concepts with concept values ​​as information representing knowledge about the concepts; a storage unit for storing a knowledge table that associates a representative concept value as information representing knowledge about the representative concept with a dependent concept value as information representing knowledge of a dependent concept defined as a concept that identifies the representative concept; an information input unit for inputting question information for inquiring about knowledge of a certain concept by presenting element values ​​representing one or more elements that identify the knowledge; an information processing unit that uses the conceptual structure table, the conceptual knowledge table, and the knowledge table stored in the storage unit to confirm that the concept of knowledge asked in the question information is specified by a concept corresponding to a conceptual value corresponding to each of one or more element values ​​included in the question information, and generates answer information to the question information based on a representative conceptual value corresponding to a subordinate conceptual value corresponding to the element value; and an information output unit that outputs the answer information obtained by the information processing unit.

2. The information processing unit includes: a concept knowledge determination means for determining whether or not there is a correspondence between a concept value and a concept corresponding to each of the one or more element values ​​in the concept knowledge table; a concept structure confirmation means for confirming, when the concept knowledge determination means determines that the correspondence exists in the concept knowledge table, that the concept corresponding to the concept value in the concept knowledge table and the concept of knowledge asked in the question information are associated in the concept structure table; and an answer generating means for generating the answer information based on a representative concept value associated with a subordinate concept value corresponding to the one or more element values ​​in the knowledge table after confirmation by the conceptual structure confirmation means.

3. The answer generating means an answer determination means for determining a concept value corresponding to the representative concept value in the concept knowledge table as an answer concept value; 3. The artificial intelligence system according to claim 2, wherein said answer information is generated based on an answer concept value determined by said answer determination means.

4. The information processing unit includes: a first arousal concept value management means for managing, when the concept knowledge determination means determines that there is a correspondence between a concept value and a concept corresponding to each of the one or more element values ​​of the question information in the concept knowledge table, the concept value as an arousal concept value of the corresponding concept; a second arousal concept value management means for managing, as an arousal concept value, a subordinate concept value in the knowledge table which is the same as the concept value managed as the arousal concept value by the first arousal concept value management means; a third arousal concept value management means for managing, as an arousal concept value, a concept value in the concept knowledge table which is the same as a representative concept value (cherry blossom, lotus flower) corresponding to the subordinate concept value managed as an arousal concept value by the second arousal concept value management means; 4. The artificial intelligence system according to claim 3, wherein said answer determining means determines, as said answer concept value, a concept value managed as an arousal concept value by said third arousal concept value managing means in said concept knowledge table.

5. 5. The artificial intelligence system according to claim 4, wherein each of said first awakening concept value management means and said third awakening concept value management means includes an awakening flag which is associated with each concept value in said concept knowledge table and has a different state depending on whether said concept value is said awakening concept value or not.

6. 6. The artificial intelligence system according to claim 4, wherein the second awakening concept value management means includes an awakening flag that is associated with each subordinate concept value in the knowledge table and has a different state depending on whether the subordinate concept value is the awakening concept value or not.

7. The information processing unit includes:

7. The artificial intelligence system according to claim 2, further comprising answer control means for controlling whether or not a representative concept value is used in generating said answer information in said knowledge table.

8. 8. The artificial intelligence system according to claim 7, wherein said answer control means has an answer control flag which is associated with a representative concept value in said knowledge table and has a different state depending on whether or not said representative concept value is used in generating said answer information.

9. a related answer generating means for generating related answer information related to the answer information based on a representative concept value in the knowledge table that corresponds to the same dependent concept value as a representative concept value used in generating answer information to the question information by the answer generating means, 3. The artificial intelligence system according to claim 2, wherein said information output section outputs said related answer information generated by said related answer generating means.

10. The information processing unit includes: a concept query control means for generating concept query information for inquiring about the concept value of the missing concept when the answer information cannot be obtained due to a lack of a concept value representing the concept for identifying the concept of knowledge in question in the question information, and for causing the information output unit to output the concept query information; 10. The artificial intelligence system according to claim 1, wherein after the concept inquiry information is output from the information output section, the answer information is generated by taking into account information on a concept value representing the missing concept that is input to the information input section.

11. The information processing unit includes: a concept structure updating means for updating said concept structure table based on new concept structure information when new concept structure information indicating that an existing concept already registered in said concept structure table is specified by one or more new concepts is input to said information input section; a concept knowledge inquiry control means for generating concept knowledge inquiry information for inquiring about a concept value of the new concept in the new concept structure information corresponding to each concept value of the existing concept included in the concept knowledge table, and outputting the concept knowledge inquiry information from the information output unit; a concept knowledge update means for updating the concept knowledge table based on a concept value of one or more of the new concepts in the new concept structure information input to the information input unit after the concept knowledge inquiry information is output from the information output unit; and a knowledge updating means for updating the knowledge table so as to correspond the conceptual value of the existing concept to a representative conceptual value and the conceptual value of each of the one or more new concepts to a subordinate conceptual value.

Images