Information processing device, information processing system, information processing method, and program
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2026-01-20
- Publication Date
- 2026-08-13
Smart Images
Figure JP2026001574_13082026_PF_FP_ABST
Abstract
Description
Information Processing Apparatus, Information Processing System, Information Processing Method, and Program
[0001] The present invention relates to an information processing apparatus, an information processing system, an information processing method, and a program.
[0002] Patent Document 1 below discloses a knowledge information creation support device for reducing the creation burden of knowledge information such as FAQs created from past response histories with customers. This support device extracts only utterance sentences having information as questions from a large number of past response logs, groups similar utterance sentences (questions) together, and creates FAQs.
[0003] Japanese Patent Application Laid-Open No. 2021-157534
[0004] By the way, the problems faced by users such as customers are diverse, and the countermeasures against those problems are also diverse. Therefore, simply proposing a countermeasure method uniformly using FAQs created based on the questions in past response logs is insufficient as a mechanism for proposing countermeasure methods against problems.
[0005] An object of the present invention is to provide an information processing apparatus, an information processing system, an information processing method, and a program that can propose a suitable countermeasure method for information input by a user.
[0006] An information processing unit according to one aspect of the present invention includes: a storage unit that stores a first tree structure configured with a hierarchical relationship between a plurality of first nodes that store data relating to a situation, and a second tree structure configured with a hierarchical relationship between a plurality of second nodes that store data relating to a situation; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information regarding a method of dealing with the first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0007] Furthermore, an information processing system according to another aspect of the present invention includes: a storage unit that stores a first tree structure configured with a hierarchical relationship between a plurality of first nodes that store data relating to a situation, and a second tree structure configured with a hierarchical relationship between a plurality of second nodes that store data relating to a situation; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information regarding a method of dealing with the first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0008] Furthermore, another aspect of the present invention relates to an information processing method, which is executed by a processor and includes: a storage step of storing a first tree structure comprising a plurality of first nodes that store data relating to a situation in a hierarchical relationship, and a second tree structure comprising a plurality of second nodes that store data relating to a situation in a hierarchical relationship; a reception step of receiving first input information from a user; a first assignment step of comparing the received first input information with each of the first nodes included in the first tree structure, calculating a first weight coefficient for each first node based on the similarity with the first input information, and assigning the calculated first weight coefficient to the first node; a second assignment step of assigning a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal step of generating proposal information regarding a method of dealing with the first input information based on the second weight coefficients assigned to the second nodes and proposing it to the user.
[0009] Furthermore, a program according to another aspect of the present invention causes the computer to function as follows: a storage unit that stores a first tree structure configured with a hierarchical relationship between a plurality of first nodes that store data relating to a situation, and a second tree structure configured with a hierarchical relationship between a plurality of second nodes that store data relating to a situation; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient based on the similarity with the first input information for each first node, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information regarding how to deal with the first input information based on the second weight coefficients assigned to the second nodes and proposes it to the user.
[0010] According to these embodiments, a first tree structure is configured with a hierarchical relationship between multiple first nodes that store data about a situation, and a second tree structure is configured with a hierarchical relationship between multiple second nodes that store data about how to deal with the situation. A first weight coefficient is assigned to the first node based on its similarity to the first input information received from the user. The first weight coefficient is propagated to the second node using the degree of association between the first and second nodes, and a second weight coefficient is assigned to the second node. Based on the second weight coefficient, the system generates proposed information for the first input information and proposes it to the user.
[0011] This allows for the identification of a second node associated with a first node that stores situation data similar to the first input information, and the generation and proposal of suggested information for the first input information using the handling method data of that second node.
[0012] In each of the embodiments described above, the system further includes a response unit that generates follow-up information to supplement the user's situation based on a first node to which a first weight coefficient has been assigned, and responds to the user; the reception unit further receives second input information from the user in response to the follow-up information; the first assignment unit compares the first input information and the second input information with each of the first nodes included in the first tree structure, calculates a weight coefficient for each first node based on the similarity between the first input information and the second input information, recalculates the first weight coefficient, and assigns the recalculated first weight coefficient to the first node; the second assignment unit assigns a second weight coefficient to each of the second nodes included in the second tree structure based on the first weight coefficient recalculated by the first assignment unit; and the proposal unit generates proposal information for the first input information and the second input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0013] According to this embodiment, based on a first node to which a weighting coefficient based on similarity with the first input information has been assigned, a follow-up question information is generated to ask the user to supplement their situation and is sent back to the user. Second input information corresponding to the follow-up question information is added to the first input information. A second node associated with the first node that stores situation data similar to the first and second input information is identified. Using the handling method data of that second node, suggestion information for the first and second input information is generated and proposed to the user.
[0014] According to the present invention, it is possible to provide an information processing device, an information processing system, an information processing method, and a program that can propose purposeful ways of handling information input by a user.
[0015] This is a block diagram illustrating the configuration of the information processing device according to the embodiment. This is a conceptual diagram for explaining the node groups divided into upper and lower hierarchies. This is a conceptual diagram for explaining the node groups classified by the nodes placed in a predetermined hierarchical level. This is a flowchart illustrating the operation of the information processing device according to the embodiment.
[0016] A preferred embodiment of the present invention will be described with reference to the attached drawings. In each drawing, components denoted by the same reference numerals have the same or similar configuration.
[0017] [Configuration of the Information Processing Device] Referring to Figure 1, an example of the configuration of the information processing device 1 according to the embodiment will be described. The information processing device 1 includes, for example, a processor 11, a storage device 12, and a communication interface 13 as its physical configuration.
[0018] The processor 11 is, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). The processor 11 executes the program 121 stored in the memory device 12 to realize various functions described later.
[0019] The storage device 12 is, for example, a computer-readable recording medium such as a disk drive or semiconductor memory. The storage device 12 functions as a storage unit that stores programs 121 for realizing various functions of the information processing device 1 and various data 122 used in those programs 121.
[0020] The various data sets 122 include, for example, data on the problematic situation, data on how to address the problematic situation, and environmental data for setting up the environment for consulting. Each data set is described below.
[0021] The data regarding the situation and the data regarding the countermeasures will be related to a common, specific event. Examples of such specific events include FA (Factory Automation) and personnel matters.
[0022] For example, if the data regarding the situation relates to FA (Financial Advisory), then the data regarding the countermeasures will also relate to FA. In this case, it is preferable to separate the data by tagging or classifying it to prevent other matters unrelated to FA (e.g., personnel matters) from being mixed in with the data regarding the situation and the data regarding the countermeasures.
[0023] Specific examples of data regarding the situation and data regarding countermeasures include the process status and countermeasures when equipment trouble occurs, the customer's situation and proposed improvement countermeasures during FA consulting, the employee's challenges and the content of the proposed curriculum when proposing a human resource development curriculum, and the physical and lifestyle conditions and countermeasures when illness occurs. It should be noted that the situation data is not limited to situations that present problems. Any situation in which a countermeasure can be proposed is acceptable.
[0024] Data regarding the situation and data regarding countermeasures can be managed using a tree structure composed of multiple nodes. For example, the tree structure may include a first tree structure consisting of multiple first nodes that store data regarding the situation arranged in a hierarchical relationship, and a second tree structure consisting of multiple second nodes that store data regarding countermeasures arranged in a hierarchical relationship.
[0025] Nodes in a hierarchical relationship function similarly to clustering labels, with nodes in higher levels having similar content attached to them.
[0026] It is preferable to configure the nodes at each level of the tree structure so that the content gradually changes from abstract to concrete as one descends the hierarchy. Specifically, the most abstract content is stored in the top-level nodes, intermediate content in the middle-level nodes, and the most concrete content in the lowest-level nodes.
[0027] Each of the first and second nodes may further include attribute data representing the attributes of the data to be stored. For example, the user's job title and affiliation can be registered as attribute data. Specifically, a department head might be registered in the top-level node, a section chief in the middle-level node, and a regular employee in the lowest-level node.
[0028] Environmental data includes data for setting the environment for consulting, such as the status of users using the information processing device 1 and the usage status of the system on the information processing device 1. Specifically, the environmental data can include the user's job title, the number of times interviews have been conducted with the user, the number of questions entered by the user, and the industries to which the proposed solutions will be applied.
[0029] The communication interface 13 connects to a network and functions as a communication unit that communicates with other devices on the network.
[0030] The information processing device 1 includes, for example, a receiving unit 111, a vector conversion unit 112, an allocation unit 113, a response unit 114, and a proposal unit 115, as functional configurations for realizing various functions. Each unit will be described below.
[0031] [Reception Unit 111] The reception unit 111 receives question texts entered by the user of the information processing device 1 by operating an input device. The input device may be provided on the information processing device 1 or may be connected to the information processing device 1 via a network.
[0032] The reception unit 111 also accepts input of attribute data such as job title and affiliation, as well as environmental data. Attribute data and environmental data are entered, for example, by a user of the information processing device 1 operating an input device. In addition, some of the environmental data may be updated as appropriate by the information processing device 1 that is performing the processing.
[0033] [Vector Transformation Unit 112] The vector transformation unit 112 has two functions: (1-1) a function as a first vector transformation unit and (1-2) a function as a second vector transformation unit. Each function is described below.
[0034] (1-1) Function as the first vector transformation unit: The function of the first vector transformation unit is to transform the question sentence into a first vector, which is a sequence of vectors.
[0035] The vector transformation unit 112 divides the question sentence using, for example, morphological analysis or a rule-based method, inputs the divided question sentence into a natural language processing (NLP) model such as Bert (Bidirectional Encoder Representations from Transformers), and uses the sequence of vectors output from that model as the first vector.
[0036] (1-2) Function as a second vector transformation unit: The function of the second vector transformation unit is to transform the data about the situation stored in each of the first nodes constituting the first tree structure into a second vector, which is a sequence of vectors.
[0037] The vector transformation unit 112 inputs the data stored in each of the first nodes into a natural language processing model such as Bert, and the sequence of vectors output from that model becomes the second vector.
[0038] [Allocation Unit 113] The allocation unit 113 has two functions: (2-1) as a first allocation unit and (2-2) as a second allocation unit. Each function is described below.
[0039] (2-1) Function as the first assignment unit: The function of the first assignment unit is to compare each of the first nodes included in the first tree structure with the question text, calculate a first weight coefficient for each first node based on its similarity to the question text, and assign the calculated first weight coefficient to the first node used in the calculation. This will be explained in detail below.
[0040] First, the allocation unit 113 calculates the similarity using the first vector and each of the second vectors. This similarity can be represented as a third vector, which is a sequence of vectors, and is preferably calculated using, for example, cosine similarity.
[0041] Cosine similarity is a method for evaluating the similarity between two texts. Specifically, cosine similarity evaluates how similar two vectors are based on a value calculated by dividing the dot product of two vectors by the magnitudes of the two vectors. The calculated value is normalized to a range of -1 to 1, where 1 means they are similar, -1 means they are not similar, and 0 means neither.
[0042] Next, the allocation unit 113 calculates the first weight coefficient for each first node constituting the first tree structure using the vectors contained in each third vector. The first weight coefficient may be obtained, for example, by calculating the sum of the top n vectors with the largest values among the vectors contained in each of the third vectors. n can be arbitrarily determined.
[0043] Next, the allocation unit 113 allocates each of the calculated first weight coefficients to the first node used in their respective calculations.
[0044] Here, in the hierarchical relationship of the first tree structure, the lower node having the highest degree of relevance (the strongest association) with the upper first node may be defined as the first node having the largest first weight coefficient among the lower nodes connected directly below the upper first node.
[0045] (2-2) Function as the second allocation unit: The function as the second allocation unit is, for example, a function of assigning a second weight coefficient based on the first weight coefficient of the first lowest-level hierarchical node to each of the second nodes included in the second tree structure, using the degree of relevance representing the strength of the association defined between each of the nodes arranged in the lowest level of the first tree structure (hereinafter referred to as "first lowest-level hierarchical node") and each of the nodes arranged in the lowest level of the second tree structure (hereinafter referred to as "second lowest-level hierarchical node"). As the degree of relevance, for example, it is preferable to use a propagation matrix described later. This will be specifically described below.
[0046] First, the allocation unit 113 calculates the second weight coefficient of each of the second lowest-level hierarchical nodes using the first weight coefficient of each of the first lowest-level hierarchical nodes and the propagation matrix.
[0047] The propagation matrix is a matrix having the number of nodes of the first lowest-level hierarchical nodes as the number of rows and the number of nodes of the second lowest-level hierarchical nodes as the number of columns. For each component of the propagation matrix, it is preferable to set, for example, the strength of the association (degree of relevance) between the corresponding first lowest-level hierarchical node and the second lowest-level hierarchical node in each component.
[0048] As this strength of the association, for example, a numerical value from 0 to 1 may be set, and in that case, 1 may be the strongest and 0 may be the weakest. By using such a propagation matrix, the second weight coefficient can be determined by the fully connected layer between the first lowest-level hierarchical node and the second lowest-level hierarchical node. Note that the second weight coefficient may be determined using a structure other than the fully connected layer.
[0049] Next, the allocation unit 113 uses the calculated second weight coefficients for each of the second lowest-level nodes to calculate the second weight coefficients for each of the second nodes located above the second lowest-level nodes, starting from the lowest level. For example, the second weight coefficient for a second node located one level above the lowest level may be the sum of the second weight coefficients of one or more second nodes located directly below that second node.
[0050] In the above description, the second assignment unit assigns the second weight coefficient to each of the second nodes based on the degree of association defined between each of the first lowest-level nodes and each of the second lowest-level nodes. However, it is not limited to using the lowest-level nodes. For example, the second weight coefficient may be assigned based on the degree of association defined between identical hierarchical nodes located at the same hierarchical level in each tree structure.
[0051] Specifically, the second weight coefficient of the second node may be assigned using the degree of association defined between each first node of the nth level of the first tree structure and each second node of the nth level of the second tree structure. Here, n in the nth level is a natural number. The relationship between the levels and n is such that the highest level is the first level, and the levels proceed downwards to the nth level as the second, third, fourth, ... levels.
[0052] [Response Unit 114] The response unit 114 generates a follow-up question to supplement the user's problematic situation based on the first node to which the first weight coefficient is assigned, and responds to the user. Preferably, the follow-up question is generated based on data relating to the user's question and having a relatively small first weight coefficient, providing information about the situation where the user's question alone does not provide sufficient information. An example of a method for generating a follow-up question is described in (a) to (c) below.
[0053] (a) The response unit 114 prioritizes selecting one or more first nodes with relatively small first weight coefficients in a predetermined hierarchical group to which the first node to which the first weight coefficient has been assigned belongs, and generates a follow-up question to ask the user to provide further information about the selected first node, and sends this response to the user.
[0054] As the predetermined hierarchical groups described above, for example, (a-1) a group consisting of node groups divided into upper and lower hierarchies in the first tree structure, and (a-2) a group consisting of node groups classified according to the node placed at a predetermined hierarchical level in the first tree structure, can be provided. Each group will be explained in detail below.
[0055] (a-1) A group consisting of node groups divided into upper and lower levels in the first tree structure: In this group, for example, as shown in Figure 2, the first node of the first tree structure Ta is divided in half into the first node group of the upper level (hereinafter also called the "upper node group") Ga and the first node group of the lower level (hereinafter also called the "lower node group") Gb, and it is determined which of the upper and lower levels has a bias in the first weight coefficient (which first weight coefficient is larger).
[0056] If the first tree structure consists of, for example, five levels, the first group of nodes in the second and third levels will be considered the upper-level node group, and the first group of nodes in the fourth and fifth levels will be considered the lower-level node group.
[0057] For example, an index based on the first weight coefficient, described later, is calculated for both the upper node group Ga and the lower node group Gb. The node with the smaller calculated index is determined to have a small first weight coefficient and insufficient information, that is, insufficient information about the situation. In this case, the response unit 114 generates a follow-up question to request additional information about the situation of one or more first nodes belonging to the node group with the smaller index based on the first weight coefficient.
[0058] Let's explain in detail. For example, let's consider the case where the sum of the first weight coefficients of the upper node group Ga is A, the number of nodes in the upper node group Ga is Na, the sum of the first weight coefficients of the lower node group Gb is B, and the number of nodes in the lower node group Gb is Nb. The index Ia based on the first weight coefficients of the upper node group Ga, and the index Ib based on the first weight coefficients of the lower node group Gb, are calculated, for example, by the following equations (1) and (2).
[0059] Ia = A / log (Na) … (1) Ib = B / log (Nb) … (2)
[0060] Here, for example, we define the sine of the vector from Ia to Ib as a bias index indicating the bias of the distribution of the first weight coefficients. This bias index will take values from -1 to 1.
[0061] In this case, the larger the bias index is in the positive direction (Ia < Ib), the more it can be determined that the first weight coefficient is biased towards the lower node group Gb. Therefore, it is preferable to generate a follow-up question to request additional information about the status of one or more first nodes belonging to the upper node group Ga.
[0062] On the other hand, the larger the bias index is in the negative direction (Ia > Ib), the more it can be determined that the first weight coefficient is biased towards the upper node group Ga. Therefore, it is preferable to generate a response sentence to ask for clarification on the status of one or more first nodes belonging to the lower node group Gb. The first nodes whose status is to be clarified may be specified as follows.
[0063] If the first weight coefficient is biased towards the lower node group Gb, the first node belonging to the upper node group Ga that has the maximum first weight coefficient is identified as an important node, and a follow-up question is generated to request further information about the identified important node. At this time, the first node located one level above the first node with the maximum first weight coefficient may be further identified as an important node.
[0064] If the first weight coefficient is biased towards the upper node group Ga, the first node belonging to the lower node group Gb that has the maximum first weight coefficient is identified as an important node, and a follow-up question is generated to request further information about the identified important node. At this time, the first node located one level below the first node with the maximum first weight coefficient may be further identified as an important node.
[0065] Here, we will explain with a specific example how to use follow-up questions to ask for clarification on the status of key nodes. For example, if the question is "Production volume is insufficient," and the key nodes identified are "Improving productivity," "Equipment utilization rate," and "Worker efficiency," then the following follow-up question will be generated and sent to the user: "You said 'Production volume is insufficient,' but could you please be a little more specific? Is it a problem with 'Equipment utilization rate'? Or is it a problem with 'Worker efficiency'?"
[0066] Suppose the user responds to this question with an additional question: "I want to prevent the equipment from stopping and improve the skills of the workers." In this case, it becomes possible to propose more specific solutions to the user regarding the insufficient production volume, such as ways to prevent the equipment from stopping and ways to improve the skills of the workers.
[0067] (a-2) A group consisting of node groups classified for each node located at a predetermined level in the first tree structure: In this group, if the predetermined level is, for example, the second level, then, as shown in Figure 3, for each first node located at the second level, node groups Gc and Gd are determined, including the first nodes that are subordinate to that first node, and it is determined which node groups Gc and Gd the first weight coefficient is biased towards. Note that if there are three first nodes located at the second level, then there will be three node groups to be determined.
[0068] For example, an index based on the first weight coefficient, described later, is calculated for each node group Gc and Gd. The smaller the calculated index, the smaller the first weight coefficient and the less information is available, meaning there is insufficient information about the situation. In this case, the response unit 114 generates a follow-up question to request additional information about the situation of one or more first nodes belonging to the node group with a small index based on the first weight coefficient.
[0069] Let me explain in detail. For example, if the first node located in the second layer consists of two nodes, node a and node b, then node a and all the nodes below node a are denoted as node group Gc, and node b and all the nodes below node b are denoted as node group Gd. Then, let C be the sum of the first weight coefficients of node group Gc, and let Nc be the number of nodes in node group Gc, and let D be the sum of the first weight coefficients of node group Gd, and let Nd be the number of nodes in node group Gd. In this case, the index Ic based on the first weight coefficients of node group Gc and the index Id based on the first weight coefficients of node group Gd are calculated, for example, by the following equations (3) and (4).
[0070] Ic = C / log (Nc) … (3) Id = D / log (Nd) … (4)
[0071] Here, for example, we define {|Ic - Id| / (the larger of Ic and Id)} as a bias index indicating the bias of the distribution of the first weight coefficients. This bias index will take values from 0 to 1. If there are three or more node groups, the maximum and minimum values of the index based on three or more first weight coefficients should be applied to Ic and Id, and equations (3) and (4) should be calculated.
[0072] The larger the bias index, the greater the bias between the first weight coefficient of node group Gc and the first weight coefficient of node group Gd. In this case, it is preferable to generate a follow-up question to request clarification on the status of one or more first nodes belonging to the node group with the smaller index based on the first weight coefficient. The first nodes whose status is to be clarified may be identified as follows.
[0073] If the node group Gc has a smaller index based on the first weight coefficient, the first node belonging to node group Gc with the highest first weight coefficient is identified as an important node, and a follow-up question is generated to request further information about the identified important node. At this time, a first node adjacent to the first node with the highest first weight coefficient (located in the next higher or next lower hierarchical level) may be further identified as an important node.
[0074] If the index based on the first weight coefficient is smaller for node group Gd, the first node belonging to node group Gd with the maximum first weight coefficient is identified as an important node, and a follow-up question is generated to request further information about the identified important node. At this time, the first node adjacent to the first node with the maximum first weight coefficient may be further identified as an important node.
[0075] (b) The response unit 114 identifies the second node with the maximum second weight coefficient (hereinafter also referred to as the "maximum second node"), determines one or more first nodes with relatively large influence indicators that show the degree to which they affect the fluctuation of the second weight coefficient of the identified maximum second node, and responds to the user with a question to ask for further information on the status of the determined first nodes.
[0076] The second weight coefficient of the second node can be calculated, for example, using the first weight coefficient of the first node located at the same hierarchical level as the second node, and the propagation matrix described above. In this case, first nodes with a high degree of similarity to the question will have a large first weight coefficient, and therefore the second weight coefficient of second nodes that are strongly associated with such first nodes will also tend to be large. Consequently, it can be inferred that second nodes with a large second weight coefficient have a high degree of similarity to the question. Based on this, the second weight coefficient of the second node is easily influenced by the first weight coefficient of the first node that is strongly associated with that second node. Therefore, by supplementing the situation of first nodes, which have a relatively large influence index indicating the degree to which they affect the fluctuation of the second weight coefficient of the second node, it is possible to increase the accuracy of the approach to the second node.
[0077] The first node with a relatively large influence index, which indicates the degree to which it affects the variation of the second weight coefficient of the largest second node, can be determined, for example, as follows:
[0078] When calculating the second weight coefficient of the maximum second node, a single-column matrix used for calculating the second weight coefficient of the maximum second node is extracted from the propagation matrix. The values of each component of this extracted single-column matrix are added to the first weight coefficient of the corresponding component in the first node matrix, which is multiplied by the propagation matrix when calculating the second weight coefficient of the maximum second node. Then, the added first weight coefficient is divided by the first weight coefficient before the addition. This allows us to calculate the degree of change in the first weight coefficient before and after the addition for each component of the first node matrix, which is multiplied by the propagation matrix. The larger this degree of change, the larger the influence index, which indicates the degree to which it affects the fluctuation of the second weight coefficient of the maximum second node. Therefore, first nodes with a relatively large degree of change can be determined as first nodes with a relatively large influence index, which indicates the degree to which they affect the fluctuation of the weight coefficient of the maximum second node.
[0079] Here, the first node from which to obtain additional information may be identified as follows: Among the first nodes from which the impact index has been calculated, the first node with the highest impact index is identified as the important node, and a follow-up question is generated to obtain additional information about the situation of the identified important node. At this time, the first node adjacent to the first node with the highest impact index may be further identified as an important node.
[0080] (c) The response unit 114 generates a follow-up question to the user to ask for further information on the status of other first nodes adjacent to the first node to which the largest first weight coefficient has been assigned.
[0081] Here, when generating a reply sentence, the reply sentence may be generated by employing any of the methods (a) to (c) above, or by employing all of the methods (a) to (c) above.
[0082] When generating a response using all methods, for example, the absolute value of the bias index in (a-1) above, the absolute value of the bias index in (a-2) above, and the absolute value of the influence index in (b) above may be calculated, and the response may be generated based on the method that maximizes the absolute value. In this case, if the maximum absolute value is smaller than a predetermined lower limit, the response may be generated by method (c) above. The predetermined lower limit can be appropriately determined according to the operational situation, etc.
[0083] Furthermore, when generating response sentences using all methods, bias indicators and influence indices may be adjusted using adjustment values that are linked to environmental data. This makes it possible to dynamically change bias indicators and influence indices based on environmental data, and consequently, to dynamically change the content of the response sentences based on environmental data.
[0084] Specifically, in (a-1) above, if the bias index is positive, it means that the first weight coefficient is biased towards the group of lower nodes where specific situations are stored. Therefore, it is preferable to set adjustment values according to the job title such that the higher the job title of the user included in the environmental data, the larger the value of the bias index in (a-1) after adjusting with the adjustment value. This makes it possible to generate a response message for high-ranking users to supplement the situation of one or more first nodes belonging to the group of higher-ranking nodes where abstract situations are stored.
[0085] In (a-1) above, if the bias index is negative, it means that the first weight coefficient is biased towards the higher-level node group where abstract situations are stored. Therefore, it is preferable to set adjustment values according to the job title such that the lower the job title of the user included in the environmental data, the larger the value of the bias index in (a-1) after adjusting with the adjustment value. This makes it possible to generate a response message for users with low job titles to ask them to supplement the situation of one or more first nodes belonging to the lower-level node group where specific situations are stored.
[0086] In (a-2) above, for example, when the number of interviews conducted is small, such as in the initial stages of interviews, it is possible to improve the quality of consulting by understanding the situation from a wide range of perspectives. Therefore, it is preferable to set the adjustment value according to the number of interviews conducted, such that the value of the bias index in (a-2) above after adjustment increases as the number of interviews included in the environmental data decreases. This makes it possible to propose countermeasures to users who have had a small number of interviews after understanding the situation from a wide range of perspectives.
[0087] In (b) above, for example, when proposing countermeasures for abnormalities in the manufacturing industry, it may be preferable to prioritize the identification of a second node that stores data on countermeasures, depending on the industry to which it is applied. Therefore, it is preferable to set adjustment values according to the industry so that the value of the impact index in (b) above after adjustment increases, depending on the industry to which the countermeasures included in the environmental data are applied.
[0088] In addition, a method other than the methods (a) to (c) above may be used to generate a response sentence. As an example of a method other than the methods (a) to (c) above, the following method (d) may be used.
[0089] (d) The response unit 114, when there are multiple second nodes associated with the first node to which the first weight coefficient has been assigned, generates a follow-up question to ask the user to provide information on the status of other first nodes associated with any of the multiple second nodes. For example, the response unit 114 may generate a follow-up question to ask the user to provide information on the status of other first nodes associated with the second node with the smaller second weight coefficient among the multiple second nodes.
[0090] If a follow-up question is sent to the user, the system will then receive a further question from the user regarding that follow-up question (hereinafter also referred to as the "second question"). In this case, the receiving unit 111 and the assignment unit 113 function, for example, as follows.
[0091] The reception unit 111 receives a second question from the user in response to the first question. After receiving the second question, the allocation unit 113 compares the first question (hereinafter also referred to as the "first question") and the second question with each of the first nodes included in the first tree structure, and calculates a weight coefficient for each first node based on the similarity between the first question and the second question. In other words, the allocation unit 113 recalculates the first weight coefficient based on the first question and the second question. Then, the allocation unit 113 allocates the recalculated first weight coefficient to the first node used in its calculation.
[0092] [Proposal Unit 115] Based on the second weight coefficient assigned to the second node, the proposal unit 115 generates a first question statement, or a statement of how to deal with the first and second question statements, and proposes it to the user.
[0093] For example, a second node with a relatively large second weight coefficient may be selected, and based on data regarding how to deal with the selected second node, a first question statement, or a statement of how to deal with both the first and second questions, may be generated and proposed.
[0094] Furthermore, the proposal unit 115 may generate and propose a solution statement if the sum of the first weight coefficients assigned to the first node included in the first tree structure exceeds a predetermined proposal criterion value.
[0095] The predetermined reference values for proposals may be adjusted using adjustment values set in conjunction with environmental data. For example, in the initial stages of interviews, when the number of interviews is small, increasing the amount of information gathered from a wide range of perspectives can improve the quality of the consulting. Therefore, it is preferable to set adjustment values according to the number of interviews conducted, such that the value after adjusting the predetermined reference values for proposals increases as the number of interviews included in the environmental data decreases. This makes it possible to propose solutions to users who have had few interviews by increasing the number of interviews and gaining a thorough understanding of the situation from a wide range of perspectives.
[0096] Here, the suggestions provided to the user are not limited to just a description of a solution. For example, the suggestions could include natural language, code, numbers, images, graphs, and structured data in JSON format.
[0097] [Operation of the Information Processing Device] An example of the operation of the information processing device 1 according to the embodiment will be described with reference to Figure 4.
[0098] First, the reception unit 111 of the information processing device 1 receives a question sentence entered by the user by operating the input device (step S101).
[0099] Next, the vector conversion unit 112 of the information processing device 1 converts the question text into a first vector and converts the data related to the situation stored in each of the first nodes constituting the first tree structure into second vectors (step S102).
[0100] Next, the allocation unit 113 of the information processing device 1 calculates a first weight coefficient for each first node based on its similarity to the question text, and allocates the calculated first weight coefficient to the first node used in its calculation (step S103).
[0101] Next, the allocation unit 113 of the information processing device 1 assigns a second weight coefficient to each of the second nodes based on the first weight coefficient, using the degree of association between the first node and the second node (step S104).
[0102] Next, the response unit 114 of the information processing device 1 calculates various indicators (step S105). These various indicators may include, for example, the bias indicator (a-1) above, the bias indicator (a-2) above, and the influence indicator (b) above.
[0103] Next, the response unit 114 of the information processing device 1 adjusts the various indicators calculated in step S105 using adjustment values that are set in correspondence with the environmental data (step S106).
[0104] Next, the proposal unit 115 of the information processing device 1 determines whether the sum of the first weight coefficients is equal to or greater than a predetermined proposal standard value (step S107).
[0105] If this determination is NO (step S107; NO), the response unit 114 of the information processing device 1 identifies important nodes based on the indicator with the largest absolute value among the indicators adjusted in step S106 (step S108).
[0106] Next, the response unit 114 of the information processing device 1 generates a follow-up question to the user to request further information on the status of the important node identified in step S108, and sends it back to the user (step S109). Then, the process proceeds to step S101.
[0107] On the other hand, if the determination in step S107 determines that the sum of the first weight coefficients is equal to or greater than a predetermined reference value for proposals (step S107; YES), the proposal unit 115 generates proposal information for the question received in step S101 based on the second weight coefficients assigned to the second node and proposes it to the user (step S110). Then, this operation ends.
[0108] As described above, the information processing device 1 according to the embodiment stores a first tree structure configured with a hierarchical relationship between a plurality of first nodes that store data relating to the problematic situation, and a second tree structure configured with a hierarchical relationship between a plurality of second nodes that store data relating to methods for dealing with the problematic situation. It assigns a first weight coefficient to the first node based on the similarity with the question sentence received from the user, propagates the first weight coefficient to the second node using the degree of relationship between the first node and the second node to assign a second weight coefficient to the second node, and generates suggestion information for the question sentence using the dealing method data of the second node identified based on the second weight coefficient, and can propose it to the user.
[0109] This allows us to identify a second node associated with a first node that stores situational data similar to the question, and then use the response method data of that second node to generate and propose suggestion information for the question.
[0110] Furthermore, according to the information processing device 1 of the embodiment, based on a first node to which a weighting coefficient based on the similarity to the question sentence has been assigned, a follow-up question sentence is generated to ask the user to supplement the situation that is the user's problem, and a follow-up question sentence is added to the previous question sentence in response to that follow-up question sentence. A second node associated with the first node that stores situation data similar to the added question sentence is identified, and using the handling method data of the second node, suggestion information for the added question sentence is generated and suggested to the user.
[0111] Therefore, according to the information processing device 1 of this embodiment, it is possible to propose a purposeful response method to the question entered by the user.
[0112] It should be noted that the present invention is not limited to the embodiments described above, and can be implemented in various other forms without departing from the spirit of the invention. For this reason, the above embodiments are merely illustrative in all respects and should not be interpreted restrictively. For example, the order of each processing step described above can be arbitrarily changed or executed in parallel, as long as there is no inconsistency in the processing content.
[0113] Furthermore, in the embodiment described above, the information processing device 1 includes a receiving unit 111, a vector conversion unit 112, an allocation unit 113, a response unit 114, and a proposal unit 115. However, some of these units may be provided by other devices other than the information processing device 1. In this case, the present invention can be applied to an information processing system comprising the information processing device 1 and other devices.
[0114] Furthermore, in the embodiment described above, the reception unit 111 receives a question statement including a first question statement and a second question statement from the user, but the information received is not limited to a question statement. The information received can be any information (input information) that is input to the information processing device 1 by the user, and may be, for example, natural language, code, numerical values, images, graphs, structured data in JSON format, etc. Also, the user is not limited to a human, but may be, for example, a computer or robot equipped with AI.
[0115] Furthermore, in the embodiment described above, the response unit 114 generates a follow-up question and responds to the user, but the information to be responded to is not limited to a follow-up question. The information to be responded to can be any information that is presented to the user as a follow-up question (follow-up question information), and may include, for example, natural language, code, numerical values, images, graphs, or structured data in JSON format.
[0116] [Note] The embodiments of this embodiment include the following disclosures.
[0117] (Note 1) Information processing device (1) comprising: a storage unit that stores a first tree structure comprising a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship; a second tree structure comprising a plurality of second nodes that store data relating to a method of dealing with the situation, arranged in a hierarchical relationship; a reception unit that receives a first question from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of relevance defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information relating to a method of dealing with the first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0118] (Note 2) The information processing device (1) according to Note 1, further comprising: a response unit that generates a follow-up question to supplement the user's situation based on the first node to which the first weight coefficient has been assigned, and responds to the user; the receiving unit further receives a second input information from the user in response to the follow-up question; the first assignment unit compares the first input information and the second input information with each of the first nodes included in the first tree structure, calculates a weight coefficient for each first node based on the similarity between the first input information and the second input information, recalculates the first weight coefficient, and assigns the recalculated first weight coefficient to the first node; the second assignment unit assigns the second weight coefficient based on the first weight coefficient recalculated by the first assignment unit to each of the second nodes included in the second tree structure; and the proposal unit generates the proposal information for the first input information and the second input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0119] (Note 3) The response unit, in a predetermined hierarchical group to which the first node to which the first weight coefficient is assigned belongs, preferentially selects one or more first nodes to which the first weight coefficient is relatively small, generates follow-up information to supplement the status of the selected first nodes, and responds to the user, as described in Note 2 (1).
[0120] (Note 4) The information processing device (1) according to Note 3, wherein the predetermined hierarchical group is composed of a first group of nodes divided into upper and lower hierarchies in the first tree structure, or is composed of a group of nodes classified according to the nodes located at a predetermined hierarchical level in the first tree structure.
[0121] (Note 5) The information processing device (1) described in Note 2, wherein the response unit identifies the weight coefficient of the second node that has the maximum second weight coefficient, determines one or more first nodes that have a relatively large influence on the variation of the weight coefficient of the identified second node, and generates follow-up information to supplement the status of the determined first nodes and responds to the user.
[0122] (Note 6) The information processing device (1) described in Note 2, wherein the response unit generates a follow-up question to the user to supplement the status of other first nodes associated with any of the multiple second nodes, when there are multiple second nodes associated with the first node to which the first weight coefficient has been assigned.
[0123] (Note 7) The information processing device (1) according to any one of Notes 1 to 6, wherein the proposal unit selects the second node in which the second weight coefficient is relatively large, and generates the proposal information for the first input information based on the data regarding the handling method of the selected second node.
[0124] (Note 8) The proposal unit generates the proposal information for the first input information and the second input information and proposes it to the user when the sum of the first weight coefficients assigned to the first node included in the first tree structure is equal to or greater than a predetermined proposal reference value, as described in any of Notes 2 to 7 (1).
[0125] (Note 9) The response unit calculates a bias index indicating the bias in the distribution of the first weight coefficients for each predetermined hierarchical group to which the first node to which the first weight coefficient is assigned belongs, and calculates an influence index indicating the degree to which the fluctuation of the weight coefficient of the second node to which the second weight coefficient is maximum affects each first node associated with the second node to which the second weight coefficient is maximum, and determines the first node based on the index with the largest absolute value among the calculated bias index and influence index, and generates follow-up information to ask the user to supplement the status of the determined first node and responds to the user, as described in any of Notes 2 to 8.
[0126] (Note 10) The information processing device (1) described in Note 9, wherein the multiple bias indicators and influence indicators are adjusted using adjustment values that change dynamically based on environmental data related to the consulting.
[0127] (Note 11) An information processing system comprising: a storage unit that stores a first tree structure comprising a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship; a second tree structure comprising a plurality of second nodes that store data relating to a situation, arranged in a hierarchical relationship; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information relating to a first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
[0128] (Note 12) An information processing method performed by a processor, comprising: a storage step of storing a first tree structure comprising a plurality of first nodes for storing data relating to a situation, arranged in a hierarchical relationship, and a second tree structure comprising a plurality of second nodes for storing data relating to the situation, arranged in a hierarchical relationship; a receiving step of receiving first input information from a user; a first assignment step of comparing the received first input information with each of the first nodes included in the first tree structure, calculating a first weight coefficient for each first node based on the similarity with the first input information, and assigning the calculated first weight coefficient to the first node; a second assignment step of assigning a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal step of generating proposal information regarding a method for dealing with the first input information based on the second weight coefficients assigned to the second nodes and proposing it to the user.
[0129] (Note 13) A program that causes a computer to function as follows: a storage unit that stores a first tree structure comprising a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship, and a second tree structure comprising a plurality of second nodes that store data relating to a method of dealing with the situation, arranged in a hierarchical relationship; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information relating to a method of dealing with the first input information based on the second weight coefficients assigned to the second nodes and proposes it to the user.
[0130] 1... Information processing device, 11... Processor, 12... Storage device, 13... Communication interface, 111... Reception unit, 112... Vector conversion unit, 113... Allocation unit, 114... Response unit, 115... Proposal unit, 121... Program, 122... Various data
Claims
1. An information processing device comprising: a storage unit that stores a first tree structure comprising a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship; a second tree structure comprising a plurality of second nodes that store data relating to a method of dealing with the situation, arranged in a hierarchical relationship; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information relating to a method of dealing with the first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
2. The information processing apparatus according to claim 1, further comprising: a response unit that generates a follow-up question to supplement the user's situation based on the first node to which the first weight coefficient has been assigned, and responds to the user; the reception unit further receives a second input information from the user regarding the follow-up question information; the first assignment unit compares the first input information and the second input information with each of the first nodes included in the first tree structure, calculates a weight coefficient for each first node based on the similarity between the first input information and the second input information, recalculates the first weight coefficient, and assigns the recalculated first weight coefficient to the first node; the second assignment unit assigns the second weight coefficient based on the first weight coefficient recalculated by the first assignment unit to each of the second nodes included in the second tree structure; and the proposal unit generates the proposal information for the first input information and the second input information based on the second weight coefficient assigned to the second node and proposes it to the user.
3. The information processing apparatus according to claim 2, wherein the response unit preferentially selects one or more first nodes with relatively small first weight coefficients in a predetermined hierarchical group to which the first node to which the first weight coefficient is assigned belongs, and generates follow-up information to ask the user to supplement the status of the selected first nodes and responds to the user.
4. The information processing apparatus according to claim 3, wherein the predetermined hierarchical group is composed of a first group of nodes divided into upper and lower hierarchies in the first tree structure, or is composed of a group of nodes classified according to the nodes located at a predetermined hierarchical level in the first tree structure.
5. The information processing apparatus according to claim 2, wherein the response unit identifies the weight coefficient of the second node that has the maximum second weight coefficient, determines one or more first nodes that have a relatively large influence on the variation of the weight coefficient of the identified second node, and generates follow-up information to provide the user with additional information regarding the status of the determined first nodes.
6. The information processing apparatus according to claim 2, wherein, when there are multiple second nodes associated with the first node to which the first weight coefficient has been assigned, the response unit generates a follow-up question to the user to supplement the status of other first nodes associated with any of the multiple second nodes, and responds to the user.
7. The information processing apparatus according to claim 1, wherein the proposal unit selects a second node in which the second weight coefficient is relatively large, and generates the proposal information for the first input information based on the data regarding the handling method of the selected second node.
8. The information processing apparatus according to claim 2, wherein the proposal unit generates the proposal information for the first input information and the second input information and proposes it to the user when the sum of the first weight coefficients assigned to the first node included in the first tree structure is equal to or greater than a predetermined proposal criterion value.
9. The information processing apparatus according to claim 2, wherein the response unit calculates a bias index indicating the bias in the distribution of the first weight coefficients for each predetermined hierarchical group to which the first node to which the first weight coefficient is assigned belongs, and calculates an influence index indicating the degree to which the fluctuation of the weight coefficient of the second node to which the second weight coefficient is maximum affects each first node associated with the second node to which the second weight coefficient is maximum, and determines the first node based on the index with the largest absolute value among the calculated bias index and influence index, and generates follow-up information to ask the user to supplement the status of the determined first node and responds to the user.
10. The information processing apparatus according to claim 9, wherein the multiple bias indicators and influence indicators are adjusted using adjustment values that change dynamically based on environmental data relating to the consulting.
11. An information processing system comprising: a storage unit that stores a first tree structure comprising a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship; a second tree structure comprising a plurality of second nodes that store data relating to a method of dealing with the situation, arranged in a hierarchical relationship; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information regarding a method of dealing with the first input information based on the second weight coefficient assigned to the second node and proposes it to the user.
12. An information processing method performed by a processor, comprising: a storage step of storing a first tree structure comprising a plurality of first nodes configured in a hierarchical relationship for storing data relating to a situation, and a second tree structure comprising a plurality of second nodes configured in a hierarchical relationship for storing data relating to the situation; a receiving step of receiving first input information from a user; a first assignment step of comparing the received first input information with each of the first nodes included in the first tree structure, calculating a first weight coefficient for each first node based on the similarity with the first input information, and assigning the calculated first weight coefficient to the first node; a second assignment step of assigning a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal step of generating proposal information relating to the first input information based on the second weight coefficients assigned to the second nodes and proposing it to the user.
13. A program that causes a computer to function as: a storage unit that stores a first tree structure composed of a plurality of first nodes that store data relating to a situation, arranged in a hierarchical relationship; a second tree structure composed of a plurality of second nodes that store data relating to a method of dealing with the situation, arranged in a hierarchical relationship; a receiving unit that receives first input information from a user; a first assignment unit that compares the received first input information with each of the first nodes included in the first tree structure, calculates a first weight coefficient for each first node based on the similarity with the first input information, and assigns the calculated first weight coefficient to the first node; a second assignment unit that assigns a second weight coefficient based on the first weight coefficient to each of the second nodes included in the second tree structure based on the degree of association defined between the nodes of the first tree structure and the second tree structure; and a proposal unit that generates proposal information relating to a method of dealing with the first input information based on the second weight coefficients assigned to the second nodes and proposes it to the user.