Problem solving support device and problem solving support method

Abstract

Provided is a problem solving support device that makes it possible to propose a pertinent solution to a problem for a client. A problem solving support device 1 includes: a storage unit for storing at least a first tree structure and a second tree structure, the first tree structure being composed of a plurality of first nodes that store data related to a problem, and the second tree structure being composed of a plurality of second nodes that store data related to a solution; a first assignment unit for individually comparing each of the first nodes included in the first tree structure with subject data to calculate, for each of the first nodes, a first weighting coefficient based on the degree of similarity with the subject data and for assigning the calculated first weighting coefficient to the first node; and a second assignment unit for assigning a second weighting coefficient based on the first weighting coefficient to each of the second nodes included in the second tree structure on the basis of the degree of relevance representing the strength of relevance defined between same-layer nodes arranged in the same layer of the first tree structure and the second tree structure.

Description

Problem-solving support device and problem-solving support method

[0001] The present invention relates to a problem solving support device and a problem solving support method.

[0002] Patent Document 1 below discloses a knowledge information creation support device for reducing the burden of creating knowledge information such as FAQs created from past customer interaction histories. This support device extracts only utterances containing question information from a large volume of past interaction logs, and groups similar utterances (questions) together to create FAQs.

[0003] Patent No. 2021-157534

[0004] However, problems, including customer needs, are diverse, and so are solutions to those problems. Therefore, when proposing solutions to problems to customers, it is desirable to propose solutions that are appropriate for the problems.

[0005] The support system of Patent Document 1 is limited to creating FAQs based on questions in the response log, and is insufficient as a mechanism for proposing solutions to problems, and there is room for improvement.

[0006] An object of the present invention is to provide a problem solving support device and a problem solving support method that are capable of proposing purposeful solutions to customer problems.

[0007] A problem-solving support device according to one aspect of the present invention includes a memory unit that stores at least a first tree structure composed of a plurality of first nodes that store data related to the problem, and a second tree structure composed of a plurality of second nodes that store data related to the solution; a first allocation unit that compares each of the first nodes included in the first tree structure with target data, calculates a first weighting factor for each first node based on the similarity between the first node and the target data, and assigns the calculated first weighting factor to the first node; and a second allocation unit that assigns a second weighting factor based on the first weighting factor to each second node included in the second tree structure based on a degree of association that represents the strength of the association between same-level nodes located at the same level in the first tree structure and the second tree structure.

[0008] According to this aspect, each first node in a first tree structure related to the problem is compared with data for which a solution is proposed, and a first weighting factor based on the similarity with the target data is assigned to each first node, and a second weighting factor based on the first weighting factor is assigned to each second node in a second tree structure based on the relevance between nodes in the same hierarchy in the first tree structure and the second tree structure related to the solution. This makes it possible to assign a weighting factor based on the similarity with data for which a solution is proposed to each node in the problem-solving structure formed by the first tree structure and the second tree structure, and to identify nodes from the problem-solving structure that are highly relevant to the target data and propose them as solutions.

[0009] In the above-described aspect, the second allocation unit may assign a second weighting factor based on the first weighting factor to each second node included in the second tree structure based on a degree of association representing the strength of the association between each first lowest-level node located at the lowest level of the first tree structure and each second lowest-level node located at the lowest level of the second tree structure.

[0010] According to this aspect, it is possible to identify a second node that is highly related to a first node based on the degree of association between the lowest-level nodes of the first tree structure and the second tree structure that constitute the problem-solving structure.

[0011] In the above-described aspect, the second allocation unit may use the relevance to calculate a second weighting factor for each of the same-level nodes in the second tree structure, and use the calculated second weighting factor to calculate the second weighting factor for each of the second nodes located above the same-level node in the second tree structure, sequentially from the lower level.

[0012] According to this aspect, the relevance between the data for which a solution is proposed and a first node at a specific level in the first tree structure is reflected in a second node at a specific level in the second tree structure, and the relevance can be reflected sequentially from the lower levels to second nodes located at higher levels of the reflected second node.

[0013] In the above-described aspect, the system may further include a first vector conversion unit that converts the target data into a first vector, which is a sequence of vectors, and a second vector conversion unit that converts the data stored in each of the first nodes into a second vector, which is a sequence of vectors, and the first allocation unit may calculate the similarity using the first vector and the second vector.

[0014] According to this aspect, it is possible to calculate the similarity between the data for which a solution is proposed and the data stored in each of the first nodes using vectors representing the respective features.

[0015] In the above-described aspect, the similarity may be represented by a third vector that is a sequence of vectors, and the first allocating unit may calculate the first weighting factor using a vector included in the third vector.

[0016] According to this aspect, it is possible to calculate the first weighting coefficient between each first node and the data for which a solution is proposed, using the similarity represented by a vector.

[0017] In the above-described aspect, the system may further include a presentation unit that presents at least a portion of the first tree structure formed by the first nodes to which the first weighting coefficient has been assigned by the first allocation unit, and the second tree structure formed by the second nodes to which the second weighting coefficient has been assigned by the second allocation unit, as an inference result for the target data.

[0018] According to this aspect, it is possible to present solutions to problems contained in the target data by using nodes of the first tree structure and the second tree structure to which weighting factors are assigned based on the degree of relevance to the target data for which a solution is proposed.

[0019] In the above-described aspect, each of the first node and the second node may further include attribute data representing attributes of the data to be stored, and may further include a receiving unit that receives input of the attribute data, and the presentation unit may present, as an inference result, the first node and the second node that correspond to the input attribute data.

[0020] According to this aspect, it is possible to present problems and solutions that match the specified attribute data.

[0021] In the above-described aspect, when presenting a first node and a second node as an inference result, the presentation unit may present the first node in association with a second node that is in a hierarchical relationship with the same-level node in a second tree structure that has a high degree of association with the same-level node in a first tree structure that is in a hierarchical relationship with the first node.

[0022] According to this aspect, it is possible to present solutions that are highly relevant to a problem included in data for which a solution is to be proposed, organized in a hierarchically associated format.

[0023] In the above-described aspect, the second allocation unit may use a fully connected layer when associating each of the same hierarchical nodes in the first tree structure with each of the same hierarchical nodes in the second tree structure to calculate the second weighting coefficient.

[0024] According to this aspect, the first tree structure and the second tree structure can be associated with each other using a fully connected layer.

[0025] In addition, a problem-solving support method according to another aspect of the present invention is a method executed by a processor, and includes: a storage step for storing at least a first tree structure composed of a plurality of first nodes that store data related to the problem, and a second tree structure composed of a plurality of second nodes that store data related to the solution; a first allocation step for comparing each of the first nodes included in the first tree structure with the target data, calculating a first weighting factor for each first node based on the similarity between the first node and the target data, and assigning the calculated first weighting factor to the first node; and a second allocation step for assigning a second weighting factor based on the first weighting factor to each second node included in the second tree structure based on the relevance degree that represents the strength of the relationship defined between same-level nodes located at the same level in the first tree structure and the second tree structure.

[0026] According to this aspect, each first node in a first tree structure related to the problem is compared with data for which a solution is proposed, and a first weighting factor based on the similarity with the target data is assigned to each first node, and a second weighting factor based on the first weighting factor is assigned to each second node in a second tree structure based on the relevance between nodes in the same hierarchy in the first tree structure and the second tree structure related to the solution. This makes it possible to assign a weighting factor based on the similarity with data for which a solution is proposed to each node in the problem-solving structure formed by the first tree structure and the second tree structure, and to identify nodes from the problem-solving structure that are highly relevant to the target data and propose them as solutions.

[0027] According to the present invention, it is possible to provide a problem solving support device and a problem solving support method that can propose purposeful solutions to customer problems.

[0028] FIG. 1 is a block diagram illustrating the configuration of a problem solving support device according to an embodiment; FIG. 2 is a conceptual diagram illustrating the function of a first vector conversion unit; FIG. 3 is a conceptual diagram illustrating the function of a second vector conversion unit; FIG. 4 is a conceptual diagram illustrating the function of a first allocation unit; FIG. 5 is a conceptual diagram illustrating the function of a first allocation unit; FIG. 6 is a conceptual diagram illustrating the function of a second allocation unit; FIG. 7 is a conceptual diagram illustrating the function of a presentation unit; FIG. 8 is a conceptual diagram illustrating the function of a presentation unit; FIG. 9 is a flowchart illustrating the operation of a problem solving support device according to an embodiment;

[0029] A preferred embodiment of the present invention will be described with reference to the accompanying drawings. In each drawing, the same reference numerals denote the same or similar configurations. However, the drawings are schematic, and specific dimensions and the like should be determined in light of the following description. Furthermore, the drawings may include portions with different dimensional relationships and ratios.

[0030] [Configuration of the Problem Solving Support Device] An example of the configuration of a problem solving support device 1 according to an embodiment will be described with reference to Fig. 1. The problem solving support device 1 includes, as its physical configuration, a processor 11, a storage device 12, and a communication interface 13, for example.

[0031] The processor 11 is, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), etc. The processor 11 executes a program 121 stored in the storage device 12 to realize various functions described below.

[0032] The storage device 12 is a computer-readable recording medium such as a disk drive or a semiconductor memory, and functions as a storage unit that stores a program 121 for implementing various functions of the problem solving support device 1, various data 122 used by the program 121, and the like.

[0033] The various data 122 includes, for example, data on issues and data on solutions. The data on issues and the data on solutions are each related to a common specific event. Examples of the specific event include personnel affairs and factory automation (FA).

[0034] For example, if the data on the issues is related to personnel, the data on the solutions will also be related to personnel. In this example, it is preferable to separate the data by tagging or categorizing them so that data related to issues and solutions does not include data related to matters other than personnel (e.g., FA).

[0035] The data related to the problem and the data related to the solution can be managed using a tree structure consisting of a plurality of nodes. For example, the tree structure can include a first tree structure consisting of a plurality of first nodes that store data related to the problem, and a second tree structure consisting of a plurality of second nodes that store data related to the solution.

[0036] The number of tree structures to be managed is not limited to two, and may be three or more. For example, in the case of three types, a first tree structure relating to why, a second tree structure relating to what, and a third tree structure relating to how may be provided. In the case of five types, a first tree structure relating to requirements, a second tree structure relating to target values, a third tree structure relating to objects, a fourth tree structure relating to actions, and a fifth tree structure relating to means may be provided.

[0037] Each of the first node and the second node may further include attribute data representing attributes of the data stored therein. For example, if the field of support is human resources consulting, the attribute data preferably includes the job title and affiliation of the person handling the data.

[0038] In this embodiment, for the sake of convenience, a case will be described in which the data stored in each node is customized for human resources consulting, but the customized data is not limited to that for human resources consulting. For example, the data may be for FA consulting, and the data can be applied to any event for which consulting is possible.

[0039] The communication interface 13 functions as a communication unit that connects to a network and communicates with other devices on the network.

[0040] The problem solving support device 1 has, as functional components for realizing various functions, for example, a reception unit 111, a vector conversion unit 112, an allocation unit 113, and a presentation unit 114. Each of these will be described below.

[0041] [Reception Unit 111] The reception unit 111 receives input of attribute data such as job title and affiliation. The attribute data can be input, for example, by a user of the problem solving support device 1 operating an input device. The input device may be provided in the problem solving support device 1, or may be connected to the problem solving support device 1 via a network.

[0042] [Vector Conversion Unit 112] The vector conversion unit 112 has (1-1) a function as a first vector conversion unit and (1-2) a function as a second vector conversion unit. Each function will be described below.

[0043] (1-1) Function as a first vector conversion unit: The function as a first vector conversion unit is to convert data for which a solution is proposed into a first vector, which is a sequence of vectors. For example, text information such as minutes containing content that may become a problem or voice of the customer (VOC) can be used as the target data.

[0044] The target data is not limited to text information. For example, images or videos may be used as the target data. In this case, it is preferable to extract characteristic elements from the images or videos and convert them into text information.

[0045] The function of the first vector conversion unit will be specifically described with reference to FIG.

[0046] First, as shown in FIG. 2, it is assumed that text information such as "Employees are not motivated. They are not willing to try new things..." is input as data Dx for which a solution is to be proposed.

[0047] Next, the vector conversion unit 112 divides the sentences included in the target data Dx using, for example, morphological analysis or a rule base. In Fig. 2, the sentences Sx after division are shown as, for example, ["Employees have no motivation," "They don't try new things," ...].

[0048] Next, the vector conversion unit 112 inputs the divided sentence Sx into a natural language processing (NLP) model such as Bert (Bidirectional Encoder Representations from Transformers), and defines the sequence of vectors output from the model as a first vector Vx.

[0049] In FIG. 2, for example, the sequence of vectors [0.98 0.01 0.1 . . .] output when "Employees are not motivated" is input to the model, and the sequence of vectors [0.08 0.51 0.00 . . .] output when "They are not willing to try new things" is input to the model are displayed as first vectors Vx.

[0050] (1-2) Function as a second vector conversion unit: The function as a second vector conversion unit is to convert the data stored in each of the first nodes constituting the first tree structure into a second vector, which is a sequence of vectors.

[0051] The function of the second vector conversion unit will be specifically described with reference to Fig. 3. For ease of explanation, the description will be made using four first nodes Na, Nb, Nc, and Nd among the first node group constituting the first tree structure Ta shown in Fig. 3.

[0052] The vector conversion unit 112 inputs the data stored in each of the first nodes Na, Nb, Nc, and Nd into a natural language processing model such as Bert, and converts the sequence of vectors output from the model into second vectors Va, Vb, Vc, and Vd.

[0053] In FIG. 3, for example, the sequence of vectors [0.92 0.31 0.31 ...] output when the model is input with "Taking on new challenges on my own" stored in the first node Na is displayed as the second vector Va, and the sequence of vectors [0.7 0.02 0.2 ...] output when the model is input with "Understanding my current mission" stored in the first node Nb is displayed as the second vector Vb.

[0054] Similarly, the sequence of vectors [0.15 0.32 0.33...] output when the model is input with "I am improving my skills" stored in the first node Nc is displayed as the second vector Vc, and the sequence of vectors [0.54 0.64 0.92...] output when the model is input with "I am able to take time for activities other than my direct mission" stored in the first node Nd is displayed as the second vector Vd.

[0055] In the same manner as described above, the vector conversion unit 112 calculates second vectors corresponding to the other first nodes constituting the first tree structure Ta.

[0056] [Allocating Unit 113] Returning to the explanation of Fig. 1, the allocating unit 113 has (2-1) a function as a first allocating unit and (2-2) a function as a second allocating unit. Each function will be explained below.

[0057] (2-1) Function as a first allocation unit: The function as a first allocation unit is to compare each of the first nodes included in the first tree structure with the target data Dx, calculate a first weighting coefficient for each first node based on the similarity with the target data Dx, and allocate the calculated first weighting coefficient to the first node used in the calculation.

[0058] The function of the first allocation unit will be specifically described with reference to Figures 4 and 5. As in the above, for convenience of explanation, the description will be made using four first nodes Na, Nb, Nc, and Nd from the first node group that constitutes the first tree structure Ta.

[0059] First, the allocation unit 113 calculates a similarity between the first vector Vx and the second vectors Va, Vb, Vc, and Vd. This similarity can be expressed as a third vector, which is a sequence of vectors, and is preferably calculated using, for example, cosine similarity.

[0060] Cosine similarity is a method for evaluating the similarity between two texts. Specifically, cosine similarity evaluates the similarity of two vectors based on the value calculated by dividing the dot product of the two vectors by the magnitude of the two vectors. The calculated value is normalized to a range of -1 to 1, with 1 indicating similarity, -1 indicating dissimilarity, and 0 indicating neither.

[0061] 4, the third vector Vax representing the similarity between the first vector Vx and the second vector Va is displayed as [0.1, 0.9, ...]. Similarly, the third vector Vbx representing the similarity between the first vector Vx and the second vector Vb is displayed as [0.9, 0.3, ...], the third vector Vcx representing the similarity between the first vector Vx and the second vector Vc is displayed as [0.5, 0.5, ...], and the third vector Vdx representing the similarity between the first vector Vx and the second vector Vd is displayed as [0.1, 0.1, ...].

[0062] Next, the allocation unit 113 calculates first weighting factors for the first nodes Na, Nb, Nc, and Nd using the vectors included in the third vectors Vax, Vbx, Vcx, and Vdx. The first weighting factors may be calculated, for example, by calculating the sum of the n largest vectors included in each of the third vectors Vax, Vbx, Vcx, and Vdx. n can be determined arbitrarily.

[0063] 5 , the sum of the top three vectors [0.9, 0.5, 0.3] of the vectors [0.1, 0.9, 0.2, 0.5, 0.3, ...] included in the third vector Vax is [1.7], which is displayed as the first weighting factor Wi (where i is a subscript) of the first node Na. Similarly, the sum of the top three vectors of the vectors included in the third vector Vbx is [1.2], which is displayed as the first weighting factor Wi of the first node Nb, the sum of the top three vectors of the vectors included in the third vector Vcx is [0.2], which is displayed as the first weighting factor Wi of the first node Nc, and the sum of the top three vectors of the vectors included in the third vector Vdx is [0.1], which is displayed as the first weighting factor Wi of the first node Nd.

[0064] Similarly to the above, the allocating unit 113 calculates the first weighting factors Wi corresponding to each of the other first nodes constituting the first tree structure Ta, and then allocates each of the calculated first weighting factors Wi to the first nodes used in the calculation.

[0065] Here, in the hierarchical relationship shown in Figure 5, the lower node that has the highest degree of relevance (strongest relationship) to the higher-level first node Na is the first node Nb, which has the largest first weighting coefficient Wi among the first nodes Nb, Nc, and Nd.

[0066] (2-2) Function as second allocation unit: The function as the second allocation unit is to allocate a second weighting factor based on the first weighting factor of the first lowest-level node to each second node included in the second tree structure based on the relevance factor representing the strength of the relation between each node located at the lowest level of the first tree structure (hereinafter referred to as the "first lowest-level node") and each node located at the lowest level of the second tree structure (hereinafter referred to as the "second lowest-level node"). It is preferable to use, for example, a propagation matrix described below as the relevance factor.

[0067] The function of the second allocation section will be specifically described with reference to FIG.

[0068] First, the allocation unit 113 calculates the second weighting coefficients wj (j is a subscript) of each of the second lowest layer nodes Tbb using the first weighting coefficients Wi of each of the first lowest layer nodes Tab and the propagation matrix M.

[0069] The propagation matrix M is a matrix in which the number of rows is the number of nodes in the first lowest layer node Tab and the number of columns is the number of nodes in the second lowest layer node Tbb. Fig. 6 shows, as an example, a 5-row, 4-column propagation matrix M determined based on the number of nodes in the first lowest layer node Tab (5) and the number of nodes in the second lowest layer node Tbb (4). It is preferable to set, for example, the strength of association (degree of association) between the corresponding first lowest layer node Tab and second lowest layer node Tbb in each component of the propagation matrix M.

[0070] The strength of this association may be set, for example, as a numerical value between 0 and 1, with 1 being the strongest and 0 being the weakest. By using such a propagation matrix M, the second weighting coefficient can be determined by a fully connected layer between the first lowest layer node Tab and the second lowest layer node Tbb. Note that the second weighting coefficient may also be determined using a structure other than a fully connected layer.

[0071] Specifically, as shown in FIG. 7, the allocation unit 113 calculates the second weight coefficient wj of each of the second lowest hierarchical nodes Tbb based on the row vector [0.52 1.55 0.75 0.49] obtained by multiplying the row vector [1.2 0.2 0.1 0.1 1.0], whose components are the first weight coefficients Wi of each of the first lowest hierarchical nodes Tab, by the propagation matrix M, which has 5 rows and 4 columns.

[0072] Next, the allocation unit 113 uses the calculated second weighting coefficient wj of each second lowest layer node Tbb to calculate second weighting coefficients for each second node located above the second lowest layer node Tbb, in order from the lowest layer. This will be specifically described with reference to FIG. 8 .

[0073] As shown in Figure 8, the allocation unit 113 assigns [2.07], which is the sum of [0.52], which is the second weighting coefficient Wj of the second node Ne, and [1.55], which is the second weighting coefficient Wj of the second node Nf, to the second weighting coefficient Wj of the second node Ni, which is located above the second nodes Ne and Nf, in the second lowest hierarchical node Tbb.

[0074] In addition, the allocation unit 113 assigns the second weighting coefficient Wj of the second node Ng of the second lowest hierarchical node Tbb, which is [0.75], to the second weighting coefficient Wj of the second node Nj that is located above the second node Ng, and assigns the second weighting coefficient Wj of the second node Nh, which is [0.49], to the second weighting coefficient Wj of the second node Nk that is located above the second node Nh.

[0075] Furthermore, the allocation unit 113 assigns the value [3.31] obtained by adding the second weighting coefficient Wj of the second node Ni, which is [2.07], the second weighting coefficient Wj of the second node Nj, which is [0.75], and the second weighting coefficient Wj of the second node Nk, which is [0.49], to the second weighting coefficient Wj of the second node Nl, which is positioned higher than the second nodes Ni, Nj, and Nk.

[0076] Here, when a third tree structure is provided in addition to the first and second tree structures, it is preferable to calculate the third weight coefficient for each of the third lowest layer nodes in the same manner as above, using the second weight coefficient Wj for each of the second lowest layer nodes Tbb and a propagation matrix determined based on the second and third tree structures. The same applies when four or more tree structures are provided.

[0077] [Presentation Unit 114] Returning to the explanation of Fig. 1, the presentation unit 114 can present at least a part of the first tree structure Ta, which is configured by first nodes to which the first weighting factor Wi is assigned by the first allocation unit, and the second tree structure Tb, which is configured by second nodes to which the second weighting factor Wj is assigned by the second allocation unit, as an inference result for the data Dx for which a solution is proposed.

[0078] When presenting the inference results, the inference results may be displayed on a display device or may be output from a printer. The display device and the printer may be provided in the problem solving support device 1, or may be connected to the problem solving support device 1 via a network.

[0079] The presentation unit 114 may display the inference results by, for example, highlighting the node with the largest weight coefficient or a predetermined value or more for each level of the tree structure compared to other nodes. To highlight a node, for example, the background color, pattern, or text within the node frame may be changed to something more conspicuous. This will be described in detail with reference to FIG. 9 .

[0080] In Figure 9, as an inference result, the first nodes Nb, Nd and second nodes Nj, Nm, which have the largest weight coefficients in each hierarchy among the first nodes constituting the first tree structure Ta and the second nodes constituting the second tree structure Tb, are displayed more prominently than the other nodes.

[0081] Furthermore, when presenting the first node and the second node as the inference result, the presenting unit 114 may present the first node presented as the problem in association with the second node that is in a hierarchical relationship with the second lowest layer node Tbb that has a high degree of association with the first lowest layer node Tab that is in a hierarchical relationship with the first node. This will be specifically described with reference to FIG.

[0082] 10 illustrates a case where, for example, the node corresponding to the problem is the first node Nc, and second nodes Nj and Nm are presented as nodes corresponding to the solution. In this case, the first lowest hierarchical node Tab that is hierarchically related to the first node Nc corresponding to the problem is the two first nodes Ng and Nh. Of the two first nodes Ng and Nh, the first node Nh with the largest weighting coefficient is the first node Nh, which has a weighting coefficient of 1.0.

[0083] Therefore, among the first nodes constituting the first tree structure Ta corresponding to the task, the first nodes Nc and Nh are displayed more prominently than the other nodes. Note that in Fig. 10, the layer immediately below the first node Nc is the first lowest layer node Tab, but if there is one or more further layers between the first node Nc and the first lowest layer node Tab, three or more first nodes on the path connecting the first node Nc and the first node Nh may be displayed more prominently than the other nodes.

[0084] On the other hand, the second lowest hierarchical node Tbb of the second tree structure Tb has four second nodes Ni, Nj, Nk, and Nl. Of the four second nodes Ni, Nj, Nk, and Nl, the second node Nj has the largest weighting coefficient of 1.55. The second node Nm is located in the hierarchy above this second node Nj.

[0085] Therefore, among the second nodes constituting the second tree structure Tb corresponding to the solutions, the second nodes Nj and Nm are displayed more prominently than the other nodes. Note that the weighting coefficients of the four second nodes Ni, Nj, Nk, and Nl can be adjusted by changing each component of the propagation matrix M.

[0086] In this way, by displaying the first nodes Nc, Nh and the second nodes Nj, Nm more prominently than the other nodes, solutions to problems can be organized and presented in a hierarchically associated format.

[0087] In FIG. 10 , the degree of association between the first lowest-level node Tab and the second lowest-level node Tbb is used to present the first and second nodes as inference results, but this is not limiting. For example, the degree of association between nodes at the same level in the first tree structure Ta and the second tree structure Tb may be used. Specifically, the degree of association between a first node at the nth level in the first tree structure Ta and a second node at the nth level in the second tree structure Tb may be used to present the first and second nodes as inference results. Here, n in the nth level is a natural number. The relationship between the levels and n is as follows: the highest level is the first level, followed by the second, third, fourth, ..., nth levels toward the lowest level.

[0088] Furthermore, when presenting the first node and the second node as the inference result, the presenting unit 114 may present, of the first node and the second node, the first node and the second node that correspond to the attribute data accepted by the accepting unit 111. This will be specifically described with reference to FIG.

[0089] 11 is a schematic diagram showing an example of presenting first nodes based on job titles, which are attribute data. A job title is set as attribute data for each first node. For example, the job title for the first node Na in the highest hierarchy is set to "department manager," the job title for the first nodes Nb and Nc in the middle hierarchy is set to "section manager," and the job title for the first nodes Nd, Ne, Nf, Ng, and Nh in the lowest hierarchy is set to "general employee." The higher the first node in the hierarchy, the more content that matches the higher-ranking job title is stored.

[0090] For example, when "department manager" is specified as the position in Fig. 11, first nodes from the top to the bottom hierarchy are presented. At this time, it may be possible to arbitrarily set which hierarchy from the top hierarchy to be presented. In Fig. 11, when "department manager" is specified as the position, nodes belonging to two hierarchies, the top hierarchy and the middle hierarchy, are presented.

[0091] It is preferable that the nodes in each layer of the tree structure be set so that the content gradually changes from abstract to concrete as the layer descends: specifically, the most abstract content is stored in the top layer nodes, intermediate content in the middle layer nodes, and the most concrete content in the bottom layer nodes.

[0092] Nodes in a hierarchical relationship function similarly to clustering labels, and nodes with similar content are subordinate to nodes in a higher hierarchy.

[0093] [Operation of the Problem Solving Support Device] An example of the operation of the problem solving support device 1 according to the embodiment will be described with reference to FIG.

[0094] First, the vector conversion unit 112 of the problem solving support device 1 converts data Dx for which a solution is to be proposed into a first vector, which is a sequence of vectors (step S101).

[0095] Next, the vector conversion unit 112 of the problem solving support device 1 converts the data stored in each of the first nodes constituting the first tree structure Ta into a second vector, which is a sequence of vectors (step S102).

[0096] Next, the allocation unit 113 of the problem solving support device 1 compares each of the first nodes included in the first tree structure Ta with the target data Dx, calculates a first weighting factor Wi for each first node based on the similarity with the target data Dx, and allocates the calculated first weighting factor Wi to the first node used in the calculation (step S103).

[0097] Next, the allocation unit 113 of the problem solving support device 1 allocates a second weighting factor Wj based on the first weighting factor Wi of the first lowest hierarchy node to each second node included in the second tree structure Tb based on the relevance determined between each of the first lowest hierarchy nodes and each of the second lowest hierarchy nodes (step S104).

[0098] Next, the presentation unit 114 of the problem-solving support device 1 presents at least a part of the first tree structure Ta, which is composed of the first nodes to which the first weighting factor Wi has been assigned in step S103, and the second tree structure Tb, which is composed of the second nodes to which the second weighting factor Wj has been assigned in step S104, as an inference result indicating a solution to the data Dx for which a solution is proposed (step S105).

[0099] As described above, the problem solving support device 1 according to the embodiment can compare each first node in a first tree structure Ta related to the problem with the data Dx for which a solution is proposed, and assign a first weighting factor Wi to each first node based on the similarity with the target data Dx, and can assign a second weighting factor Wj based on the first weighting factor Wi to each second node in a second tree structure Tb based on the degree of association between the first tree structure Ta and the lowest-level node in a second tree structure Tb related to the solution.

[0100] This allows each node of the problem-solving structure consisting of the first tree structure Ta and the second tree structure Tb to be assigned a weighting coefficient based on the similarity with the data Dx for which a solution is proposed, and makes it possible to identify nodes from the problem-solving structure that are highly relevant to the target data Dx and propose them as solutions.

[0101] Therefore, the problem solving support device 1 according to the embodiment makes it possible to propose a purposeful solution to a customer's problem.

[0102] The present invention is not limited to the above-described embodiment, and can be embodied in various other forms without departing from the spirit of the present invention. Therefore, the above-described embodiment is merely an example in all respects and should not be interpreted as being limiting. For example, the order of the above-described processing steps can be arbitrarily changed or executed in parallel as long as no contradiction occurs in the processing content.

[0103] [Additional Note] Aspects of the present embodiment include the following disclosure.

[0104] (Supplementary Note 1) A problem solving support device (1) comprising: a storage unit (12) that stores at least a first tree structure composed of a plurality of first nodes that store data related to a problem, and a second tree structure composed of a plurality of second nodes that store data related to a solution; a first allocation unit (113) that compares each of the first nodes included in the first tree structure with target data, calculates a first weighting factor for each of the first nodes based on the similarity with the target data, and allocates the calculated first weighting factor to the first node; and a second allocation unit (113) that allocates a second weighting factor based on the first weighting factor to each of the second nodes included in the second tree structure based on a degree of association that represents the strength of an association defined between same-level nodes arranged at the same level in the first tree structure and the second tree structure.

[0105] (Supplementary Note 2) The problem solving support device (1) described in Supplementary Note 1, wherein the second allocation unit (113) allocates a second weighting factor based on the first weighting factor to each of the second nodes included in the second tree structure based on a degree of association representing the strength of the association between each of the first lowest hierarchy nodes placed in the lowest hierarchy of the first tree structure and each of the second lowest hierarchy nodes placed in the lowest hierarchy of the second tree structure.

[0106] (Supplementary Note 3) The problem solving support device (1) described in Supplementary Note 1 or 2, wherein the second allocation unit (113) uses the relevance to calculate the second weighting factor for each of the same-level nodes in the second tree structure, and uses the calculated second weighting factor to calculate the second weighting factor for each of the second nodes located above the same-level nodes in the second tree structure, sequentially from the lower levels.

[0107] (Supplementary Note 4) The problem solving support device (1) according to any one of Supplementary Notes 1 to 3, further comprising: a first vector conversion unit (112) that converts the target data into a first vector, which is a sequence of vectors; and a second vector conversion unit (112) that converts data stored in each of the first nodes into a second vector, which is a sequence of vectors, respectively; wherein the first allocation unit (113) calculates the similarity using the first vector and the second vector.

[0108] (Supplementary Note 5) The problem solving support device (1) according to Supplementary Note 4, wherein the similarity is represented by a third vector that is a sequence of vectors, and the first allocation unit (113) calculates the first weighting factor using a vector included in the third vector.

[0109] (Supplementary Note 6) The problem solving support device (1) according to any one of Supplementary Notes 1 to 5, further comprising a presentation unit (114) that presents at least a part of the first tree structure constituted by the first nodes to which the first weighting coefficients have been assigned by the first allocation unit (113) and the second tree structure constituted by the second nodes to which the second weighting coefficients have been assigned by the second allocation unit (113) as an inference result for the target data.

[0110] (Supplementary Note 7) The problem solving support device (1) described in Supplementary Note 6, wherein each of the first node and the second node further includes attribute data representing attributes of the data to be stored, and further includes a receiving unit (111) that receives input of the attribute data, and the presentation unit (114) presents, as the inference result, the first node and the second node that correspond to the input attribute data.

[0111] (Supplementary Note 8) The problem solving support device (1) according to Supplementary Note 6, wherein when presenting the first node and the second node as the inference result, the presentation unit (114) presents the first node to be presented in association with the second node in the second tree structure that has a high degree of association with the same-level node in the first tree structure that has a hierarchical relationship with the first node.

[0112] (Supplementary Note 9) The problem solving support device (1) according to any one of Supplementary Notes 1 to 8, wherein the second allocation unit (113) uses a fully connected layer when associating each of the same-level nodes in the first tree structure with each of the same-level nodes in the second tree structure to calculate the second weight coefficients.

[0113] (Supplementary Note 10) A problem-solving support method executed by a processor (11), comprising: a storage step of storing at least a first tree structure constituted by a plurality of first nodes that store data related to a problem, and a second tree structure constituted by a plurality of second nodes that store data related to a solution; a first allocation step of comparing each of the first nodes included in the first tree structure with target data, calculating a first weighting factor for each of the first nodes based on the similarity with the target data, and allocating the calculated first weighting factor to the first node; and a second allocation step of assigning a second weighting factor based on the first weighting factor to each of the second nodes included in the second tree structure, based on a degree of association that represents the strength of an association defined between same-level nodes that are arranged at the same level in the first tree structure and the second tree structure.

[0114] 1... problem solving support device, 11... processor, 12... storage device, 13... communication interface, 111... reception unit, 112... vector conversion unit, 113... allocation unit, 114... presentation unit, 121... program, 122... data

Claims

1. A problem solving support device comprising: a memory unit that stores at least a first tree structure composed of a plurality of first nodes that store data related to a problem, and a second tree structure composed of a plurality of second nodes that store data related to a solution; a first allocation unit that compares each of the first nodes included in the first tree structure with target data, calculates a first weighting coefficient for each of the first nodes based on the similarity between the first node and the target data, and assigns the calculated first weighting coefficient to the first node; and a second allocation unit that assigns a second weighting coefficient based on the first weighting coefficient to each of the second nodes included in the second tree structure based on a degree of association that represents the strength of the association between same-level nodes that are located at the same level in the first tree structure and the second tree structure.

2. The problem solving support device described in claim 1, wherein the second allocation unit assigns a second weighting coefficient based on the first weighting coefficient to each of the second nodes included in the second tree structure based on a degree of association representing the strength of the association defined between each of the first lowest hierarchy nodes located in the lowest hierarchy of the first tree structure and each of the second lowest hierarchy nodes located in the lowest hierarchy of the second tree structure.

3. The problem solving support device described in claim 1, wherein the second allocation unit uses the relevance to calculate the second weighting coefficient for each of the same hierarchical nodes in the second tree structure, and uses the calculated second weighting coefficient to calculate the second weighting coefficient for each of the second nodes located above the same hierarchical nodes in the second tree structure, successively from the lower hierarchical levels.

4. The problem solving support device of claim 1, further comprising: a first vector conversion unit that converts the target data into a first vector, which is a sequence of vectors; and a second vector conversion unit that converts each of the data stored in the first nodes into a second vector, which is a sequence of vectors; wherein the first allocation unit calculates the similarity using the first vector and the second vector.

5. The problem solving support device according to claim 4, wherein the similarity is represented by a third vector which is a sequence of vectors, and the first allocation unit calculates the first weighting coefficient using a vector included in the third vector.

6. The problem solving support device of claim 1, further comprising a presentation unit that presents at least a portion of the first tree structure constituted by the first nodes to which the first weighting coefficient has been assigned by the first allocation unit, and the second tree structure constituted by the second nodes to which the second weighting coefficient has been assigned by the second allocation unit, as an inference result for the target data.

7. A problem solving support device as described in claim 6, wherein each of the first node and the second node further includes attribute data representing attributes of the data to be stored, and further comprising a receiving unit that receives input of the attribute data, and the presentation unit presents, of the first node and the second node, the first node and the second node that correspond to the input attribute data as the inference result.

8. The problem solving support device as described in claim 6, wherein when presenting the first node and the second node as the inference result, the presentation unit presents the first node to be presented in association with the second node that is in a hierarchical relationship with the same hierarchical node in the second tree structure that has a high degree of association with the same hierarchical node in the first tree structure that is in a hierarchical relationship with the first node.

9. The problem solving support device according to claim 1, wherein the second allocation unit uses a fully connected layer when associating each of the same hierarchical nodes in the first tree structure with each of the same hierarchical nodes in the second tree structure to calculate the second weighting coefficient.

10. A problem solving support method executed by a processor, comprising: a storage step for storing at least a first tree structure composed of a plurality of first nodes storing data related to a problem, and a second tree structure composed of a plurality of second nodes storing data related to a solution; a first allocation step for comparing each of the first nodes included in the first tree structure with target data, calculating a first weighting coefficient for each of the first nodes based on the similarity with the target data, and assigning the calculated first weighting coefficient to the first node; and a second allocation step for assigning a second weighting coefficient based on the first weighting coefficient to each of the second nodes included in the second tree structure based on a degree of association representing the strength of association defined between same-level nodes located at the same level in the first tree structure and the second tree structure.

Images