WO2023238511A1

WO2023238511A1 - Ideation assistance device, ideation assistance program, and ideation assistance method

Info

Publication number: WO2023238511A1
Application number: PCT/JP2023/015317
Authority: WO
Inventors: 伸夫苗村
Original assignee: 株式会社日立製作所
Priority date: 2022-06-07
Filing date: 2023-04-17
Publication date: 2023-12-14
Also published as: JP2023179025A

Abstract

The present invention comprises: an idea database 102 that stores ideas, the number of votes for each idea, and a citation relation of each idea; an idea selection unit 105 that selects an idea stored in the idea database 102; an input unit 101 that receives input of votes with respect to each idea selected by the idea selection unit 105, and a proposed idea; a citation definition unit 103 that calculates the citation relation and number of citations of each idea; and an idea classification unit 104 that uses the number of votes and number of citations of each idea to classify the ideas. Thus, provided are an ideation assistance device, an ideation assistance program, and an ideation assistance method that identify an idea suitable for reference and contribute to the derivation of a better idea.

Description

Idea support device, idea support program, and idea support method

The present invention relates to an ideation support device, an ideation support program, and an ideation support method.

Patent Document 1 describes how to systematically generate ideas by selecting good ideas using the collective intelligence of a large number of unspecified participants online, and actively piggybacking on good ideas using a brainwriting framework. It describes an idea support device that improves the quality and quantity of ideas. Patent Document 1 describes, "an idea database that stores ideas, evaluations of ideas, and parent-child relationships of ideas, an arithmetic unit that selects an idea to be displayed as a draft based on the evaluation of ideas, and a system that displays the draft and allows participants to view the draft. and an interface for inputting proposals, which are new ideas associated with the original, using the original as a parent.

Patent Document 2 describes a technology for automatically grouping each idea presented by participants according to its content without manual operation when conducting a discussion for idea generation between remote locations. ing. Patent Document 2 states that ``Until the client terminal performs a change operation on the group, idea memos are grouped by clustering, and after the client terminal performs a change operation on the group, the idea memo and the group are grouped. Grouping is performed based on the results of learning the correspondence between

Japanese Patent Application Publication No. 2021-157509 JP2020-8987A

In recent years, in product and service design, it has become increasingly important to not only improve performance, which is easy to quantify, but also to acquire innovative ideas that bring about changes in consumers' lives and values.

Various brainstorming techniques, such as Osborn's checklist and brainwriting, have been proposed as ways to support idea generation. Brainwriting is where each participant in a brainstorming session writes down three of his or her ideas on a sheet of paper, hands the sheet to the next participant, and the next participant develops each idea using the previous three ideas as a reference. This is a forced association method that allows you to generate a large number of ideas by repeating the process of coming up with three of your own ideas and writing them down.

These techniques assume offline brainstorming between one to ten people, but in recent years, online brainstorming using computer devices and the Internet has been proposed. In online brainstorming, the number of participants and the number of ideas to be handled are enormous, making it difficult for participants to read through all the ideas and identify the good ones.

On the other hand, for example, in Patent Document 1, brainwriting is carried out online, and participants propose ideas and simultaneously evaluate several ideas by voting to identify good ideas with a large number of votes. A method to encourage reference has been proposed. Further, Patent Document 2 proposes a method of automatically grouping ideas proposed as electronic data.

One of the benefits of brainstorming with multiple people is that you can refine your ideas by proposing new ideas based on ideas that others have come up with. On the other hand, in online brainstorming where many participants participate at a time that is convenient for each person, it becomes difficult for the idea creator to look through all existing ideas, select good ones, and refer to them. .

Although Patent Document 1 identifies good ideas based on the number of votes and encourages reference to these good ideas, ideas with a large number of votes are not necessarily suitable for reference. In other words, ideas with a large number of votes may already be difficult to refine through reference, and there is a high possibility that no matter how much reference is used, no further improvement in quality can be expected, so ideas that are more suitable for reference are identified. There is a need for improvement, and there is room for improvement.

Furthermore, the method of Patent Document 2 groups ideas based on their contents, and does not group ideas by determining whether each idea is good or bad, so there is room for improvement.

The present invention was made in view of this situation, and provides an ideation support device, an ideation support program, and an ideation support method that identify ideas suitable for reference and contribute to deriving better ideas. The purpose is to

The present invention includes a plurality of means for solving the above-mentioned problems, and one example thereof is a database that stores ideas, the number of votes for each idea, and citation relationships, and a database that stores ideas, the number of votes for each idea, and citation relationships, and an idea selection unit for selecting an idea; an input unit for inputting votes for the proposed idea and each of the ideas selected in the idea selection unit; and calculating the citation relationship and number of citations for each of the ideas. and an idea classification section that classifies the ideas using the number of votes and the number of citations for each idea.

According to the present invention, it is possible to identify ideas suitable for reference and contribute to deriving better ideas. Problems, configurations, and effects other than those described above will be made clear by the description of the following examples.

1 is an overall schematic configuration diagram of an ideation support device according to a first embodiment; FIG. 1 is a diagram showing a configuration of an interface of an ideation support device according to a first embodiment; FIG. FIG. 3 is a diagram showing the configuration of an interface dedicated to an initial stage in the ideation support device according to the first embodiment. FIG. 3 is a diagram for explaining classification of ideas using voting probability and citation probability in the ideation support device according to the first embodiment. FIG. 3 is a diagram for explaining an idea selection unit in the ideation support device according to the first embodiment. FIG. 3 is a diagram for explaining visualization of a single family idea using a tree structure in the output unit of the ideation support device according to the first embodiment. FIG. 3 is a diagram for explaining visualization of ideas for a plurality of family lines using an effective graph in the output unit of the ideation support device according to the first embodiment.

Examples of the idea support device, idea support program, and idea support method of the present invention will be described below with reference to the drawings. In the drawings used in this specification, the same or corresponding components are given the same or similar symbols, and repeated description of these components may be omitted.

<Example 1>
Embodiment 1 of the idea support device, idea support program, and idea support method of the present invention will be described using FIGS. 1 to 7.

First, the overall configuration of the ideation support device will be explained using FIG. 1. FIG. 1 shows the configuration of an ideation support device 100 according to a first embodiment.

The ideation support device 100 shown in FIG. 1 includes an input section 101, an idea database 102, a quotation definition section 103, an idea classification section 104, an idea selection section 105, and an output section 106.

Each of the configurations, functions, processing units, processing means, etc. described above can be realized in part or in whole by hardware, for example, by designing an integrated circuit. The hardware constituting the unit 101 and the output unit 106 can be a keyboard, a display, a mobile terminal, or a display (which can also serve as a touch panel) on the participant side.

On the other hand, in the ideation support device 100, there is a function to display a screen forming part of the input section 101 as shown in FIGS. 2 and 3, an idea database 102, a quotation definition section 103, an idea classification section 104, an idea selection section. The computer 105 can be configured with a computer having the above-mentioned liquid crystal display, etc., an input device, a storage device, a CPU, a memory, etc., but it may also be in the form of a cloud service, and is not particularly limited.

The operation of the ideation support device 100 may be controlled by various programs. This program is stored in an internal recording medium and an external recording medium, and is read and executed by the CPU.

Note that the operation control processing executed by the ideation support device 100 may be combined into one program, each may be divided into multiple programs, or a combination thereof may be used. Further, part or all of the program may be realized by dedicated hardware or may be modularized. Furthermore, various programs may be installed on each computer by a program distribution server or a storage medium.

The input unit 101 is a unit for inputting the proposed ideas and votes for each idea selected by the idea selection unit 105, and is preferably the main body that executes the input step or input process.

This input section 101 is equipped with an evaluation/proposal screen 200 shown in FIG. There is.

In FIG. 2, a total of four draft display sections 201 and four draft evaluation sections 202 are arranged for each of Ideas 1 to 4, but any number of draft display sections 201 and draft evaluation sections 202 may be displayed as long as there is one or more. Note that since there is no draft at the initial stage of brainstorming, the draft may be left blank, or an interface like the one shown in FIG. 3 dedicated to the initial stage may be used. Further, in order to improve convenience for participants, a brainstorming theme 205 may be displayed on the evaluation/proposal screen 200.

The draft display section 201 displays the already mentioned idea selected according to the idea selection section 105 as a draft to be used as a reference for the next idea.

The draft evaluation section 202 is a button for participants to input their evaluation of the draft, and by clicking the "Vote" button and marking it as "voted," the draft evaluation section 202 is a button that allows participants to input their evaluation of the draft by a specified amount (for example, Increase by 1).

For the multiple draft evaluation units 202 displayed, "voting" may be limited to just one like a radio button, or multiple selections may be allowed and "voting" can be done for any number of drafts. It may be possible to do so. In Figure 2, a button-type evaluation is used, but it is also possible to make the evaluation a point system and provide a slider, radio buttons, and a free entry field for entering numbers. It is also possible to evaluate each aspect separately, such as performance and profitability.

The proposal reception section 203 is a free entry field for participants to enter their own ideas, and inputs ideas that are referenced to the ideas of the draft evaluation section 202. However, the participants do not necessarily need to refer to the ideas of the draft evaluation section 202, and may write in unrelated ideas.

When the submit button 204 is clicked after input to the draft evaluation section 202 and proposal reception section 203 is completed, the input data is transmitted to the quotation definition section 103 and recorded in the idea database 102.

The idea database 102 is a variety of recording media that saves the ideas themselves, the number of votes for each idea, and citation relationships, and is preferably the main body that executes the storage step and storage process.

In this idea database 102, all ideas entered in the proposal receiving section 203 are stored together with ID numbers, parent numbers, and the number of votes. Preferably, the proposal date and time, proposer information, number of citations, number of evaluations, evaluator information, and voter information are also saved.

A citation relationship is defined for the ideas (proposals) entered in the proposal receiving section 203 as children whose parents are some (0 to 4) of the drafts displayed in the 1 to 4 draft display sections 201. That is, if a proposal was obtained by referring to a certain draft, a citation relationship is defined with the proposal as a child and the draft as a parent, and the number of citations of the parent is increased by one.

Each item of the idea database 102 will be explained below.

The ID number is a natural number assigned to each idea. ID numbers may be assigned in the order of proposal. In the idea, the content input by the participant to the proposal reception unit 203 may be stored as text. The date and time when the submit button 204 was clicked may be stored as the proposal date and time. The name and identification number of the participant who input the idea may be recorded in the proposer information. The ID of the draft defined as the parent of the proposal may be stored in the parent number. If there is no draft at the initial stage of brainstorming, or if there is no citation relationship between the draft and the proposal, the parent number is set to 0. The number of citations is the number of child ideas that have that idea as a parent. The number of times the idea has been displayed on the evaluation/proposal screen 200 is stored in the number of evaluations. As for the number of votes, the number of times the idea is displayed on the evaluation/proposal screen 200 and voted for by the draft evaluation section 202 is stored. As shown in FIG. 2, if the button-type evaluation is 1 and the score from one vote is 1, the number of votes is stored as the number of votes. On the other hand, when giving points using sliders, radio buttons, or free input, the total sum of evaluations by each participant is saved. The evaluator information stores the name and identification number of the participant whose idea was displayed on the evaluation/proposal screen 200. The name and identification number of the participant who voted for the idea among the participants whose idea was displayed on the evaluation/proposal screen 200 may be stored in the voter information. If evaluations are given as points using a slider or the like instead of voting, the names and identification numbers of participants who gave evaluations higher than a standard score (such as 4 or higher on a 5-point scale) may be saved.

The citation definition unit 103 is a part that calculates the citation relationship and number of citations for each idea, and is preferably the main body that executes the citation definition step and the citation definition process.

The citation definition section 103 defines the citation relationship between the proposal of the proposal reception section 203 and the draft of the draft display section 201. For example, the quotation definition unit 103 calculates the degree of similarity between the idea proposed by the input unit 101 and the idea selected by the idea selection unit 105 using natural language processing, and the citation definition unit 103 calculates the degree of similarity between the idea proposed by the input unit 101 and the idea selected by the idea selection unit 105. Ideas exceeding a threshold are defined as being cited by ideas in the input unit 101.

More specifically, in defining the citation relationship, sentences constituting each idea can be converted into a multidimensional vector using a natural language processing model, and the cos similarity between the proposal and each original draft can be used.

For example, if four proposals A, B, C, and D are displayed on the evaluation/proposal screen 200, and there is proposal It is possible to calculate four types of cos similarity and define a citation relationship using any one of drafts A, B, C, and D having the largest cos similarity as a parent.

However, in this method, the citation relationship is defined regardless of the magnitude of the cos similarity, so even if the cos similarity of AX, B-X, C-X, and D-X is sufficiently small, the citation relationship is will be defined. In order to avoid this, there is a method of providing a threshold value (the above-mentioned first threshold value) for the cos similarity. In this method, only when the cos similarity is equal to or higher than a predetermined threshold, a citation relationship is established between the proposal and the original proposal A, B, C, or D, which has the highest cos similarity. Define. Alternatively, a citation relationship is defined with all drafts A, B, C, and D whose cos similarity exceeds a threshold as parents.

In the former method, there is one parent, so a directed graph with ideas as nodes and citation relationships as directed edges forms a tree structure. The latter method allows multiple parents, so the resulting graph structure is not necessarily a tree structure, but because ideas are proposed as a time series, it is a directed acyclic graph.

There is no limit to the natural language processing model used to vectorize idea sentences, and any model may be used, such as a neural network or a simpler bag of words. Furthermore, when the languages of the original draft and the proposal are different, the original draft may be translated into a language suitable for the participants using automatic translation and then displayed, and the cos similarity may be calculated in the translated language.

Through the above method, similar ideas with citation relationships are connected in a graph structure, and eventually one or more multiple graph structures are generated, and ideas are automatically clustered for each graph structure. .

The idea classification unit 104 classifies ideas using the number of votes and number of citations for each idea. The idea classification unit 104 is the main body that executes the idea classification step and the idea classification process. Preferably, the idea classification unit 104 refers to the number of citations and the number of votes for each idea, and if the number of citations or the citation probability calculated from the number of citations is equal to or higher than a second threshold, and the number of votes or the voting probability calculated from the number of votes is Ideas below the third threshold are classified as reference ideas whose citation is recommended, and ideas whose number of votes or voting probability calculated from the number of votes is above the fourth threshold are classified as promising ideas.

For example, the idea classification unit 104 classifies ideas based on the number of citations and the number of votes for each idea. Since the classification regarding the content of ideas is performed by the quotation definition unit 103, the idea classification unit 104 determines whether the ideas are good or bad. Here, we will particularly consider identifying ideas that have excellent content (hereinafter referred to as "promising ideas") and ideas that are suitable for reference (reference ideas).

For potential ideas, the voting probability is calculated by dividing the number of votes for each idea by the number of evaluations, and, for example, ideas whose voting probability is equal to or higher than a threshold (the above-mentioned fourth threshold) can be identified and classified as potential ideas. The threshold value may be set as a constant value such as 0.75, or may be set dynamically using a quantile value such as the 75% quantile of all ideas acquired so far. Instead of using the number of evaluations, a candidate whose number of votes exceeds a threshold value may simply be determined as a promising proposal.

The number of citations and the number of votes will be used together to identify reference proposals. The reference proposal calculates the citation probability and voting probability by dividing the number of citations and votes for each idea by the number of evaluations. 3) The following ideas can be identified and classified as reference ideas. This is because while a reference proposal that is suitable for reference has a high probability of being cited, it leaves room for refinement, so the probability of voting is low.

Ideas that have both a high voting probability and a high citation probability are promising ideas and are not suitable for reference because no further improvement can be expected and they are not considered as reference ideas. The threshold value may be a fixed numerical value as in the case of the probable proposal, or a quantile value may be used. Further, the threshold value may be determined separately for the probable proposal and the reference proposal. For example, the threshold for reference proposals may be set to 0.5 or more for citation probability and 0.5 or less for voting probability, and the threshold for potential proposals may be set for 0.75 or more for voting probability. In this case, ideas are classified as shown in Figure 4.

For reference proposals as well, without using the number of evaluations, those with the number of citations exceeding the threshold and the number of votes below the threshold may be used as reference proposals. In addition, a reference idea may be defined as one in which the total number of votes of the children of the idea or the average voting probability of the children, which is calculated by dividing the total number of votes of the children by the total number of evaluations of the children, exceeds a threshold value. . In addition, potential proposals and reference proposals may be limited to those that satisfy the above criteria and have been evaluated a certain number of times or more. For example, only ideas that have been evaluated twice or more are candidates for potential ideas or reference ideas. This is because when the number of evaluations is extremely small, the reliability of the voting probability and citation probability is low.

The idea selection unit 105 is a part that selects ideas to be displayed to participants, and is preferably the main body that executes the idea selection step and idea selection process.

The idea selection unit 105 selects an idea to be displayed on the original proposal display unit 201 from among potential ideas and reference ideas based on the classification results of the idea classification unit 104 and information in the idea database 102.

Figure 5 shows the idea selection procedure.

After starting the process, it is determined whether the number of data in the idea database 102 is greater than or equal to a predetermined initial number of families (step S1). If the number is less than the initial number of families, the idea is not displayed on the draft display section 201 (step S5). A family tree is a cluster of ideas that are connected when a directed graph is followed in the direction of a directed edge, and the initial family number is a constant that determines the number of clusters at the start of idea generation, and may be any natural number.

If the number of ideas is less than the initial number of families, no ideas are selected and the original draft is not displayed. In this case, the parent number of the proposed idea is 0, making it an initial idea without an original draft. By tracing ideas in the direction of directed edges from an initial idea, it is possible to identify ideas that belong to a family tree.

If the number of ideas is greater than or equal to the initial number of families, the ideas to be displayed are selected in three ways. First, two ideas are selected from among the ideas that have been evaluated less than 2 times (step S2). The two ideas may be selected at random, but the natural language processing used in the citation definition unit 103 may also be used. In this case, one idea is randomly selected from the ideas that have been evaluated less than 2 times, and the idea that has the smallest cos similarity with the selected idea and that has been evaluated less than 2 times is selected as another idea. This makes it possible to increase the variety of ideas displayed.

Next, one idea is selected from the reference plans (step S3). If there are multiple reference ideas, one may be selected at random, but bias in the displayed ideas can be prevented by preferentially selecting the reference idea that has been evaluated less frequently. That is, if the minimum value of the number of evaluations among the reference plans is 2, one may be selected at random from among the reference plans with the number of evaluations of 2. Alternatively, similar to parent selection in evolutionary calculations used in mathematical optimization, the fitness, which represents the ease of selection, is calculated from the number of evaluations, the citation probability, or both, and the fitness is used for roulette selection or tournament selection. You may select a reference plan by When considering the fitness maximization problem, the citation probability itself, the reciprocal of the number of evaluations, the weighted sum of these, etc. can be used as the fitness.

Finally, one idea is selected from the promising ideas (step S4). This can also be selected in the same way as the reference proposal. When considering the fitness maximization problem, the voting probability itself, the reciprocal of the number of evaluations, the weighted sum of these, etc. can be used as the fitness when selecting a likely option. Here, four ideas to be displayed are selected using three methods, but the selection method may be only one of the three, or the number of ideas to be selected may be different from the above.

The output unit 106 outputs the citation relationship of each idea calculated by the citation definition unit 103 and the classification of each idea by the idea classification unit 104. Preferably, the output unit 106 can output the citation relationships between ideas in a directed graph or tree structure.

For example, the output unit 106 can change the color, shape, size, or font of nodes in a directed graph or tree structure for each classification of ideas. Furthermore, the output unit 106 can surround the idea with a line for each family lineage connected by tracing directed edges from the initial idea, and use different node shapes, colors, sizes, and fonts for each family lineage.

Figure 6 shows an example of visualizing a certain family idea using a tree structure.

In Figure 6, the initial idea is placed at the bottom as the root node of the tree structure, and the ideas are connected by edges based on citation relationships. This allows ideas with many citations to have many upward edges, making them easy to distinguish. However, the number of edges represents the number of citations, does not take into account the number of evaluations, and does not correspond to the citation probability. Therefore, in order to visualize the reference plan, the node shape of the reference plan is made different from other nodes.

Furthermore, in FIG. 6, the color of each node corresponds to the voting probability, making it possible to distinguish between likely proposals. The shape of the node for the initial idea has also been changed. Since it is only necessary to distinguish between a reference idea, a promising idea, and an initial idea, in addition to the shape and color of the nodes, it is also possible to change other elements such as the size of the nodes, the color of the border, and the font.

In Figure 6, the root node of the initial idea is placed at the bottom, but it may be placed at the top or in the center like in a mind map.

Figure 7 shows an example of visualizing multiple family tree ideas using a directed graph.

In FIG. 7, when the citation definition unit 103 allows multiple drafts to become parents, the idea obtained as shown in FIG. 7 becomes a directed acyclic graph. In addition, when selecting four ideas to display from different families, child ideas whose parents are original ideas from multiple families may occur, and when ideas belonging to each family are clustered as shown in Figure 7, multiple ideas may be created. Ideas arise that belong to the family lineage of. When four ideas to be displayed are selected from the same family tree, or when there is only one parent draft, each family tree has an independent graph structure or tree structure.

When the graph structure of multiple family lines is connected as shown in Figure 7, in order to understand each family line, it is recommended to surround the family lines with lines as shown in Figure 7, or to use different node shapes, colors, and fonts for each family line. You may. In addition, the natural language processing used in the citation definition unit 103 is used to obtain a sentence vector for each idea, and the sentence vector is converted into a two-dimensional or three-dimensional image using a dimension reduction method such as principal component analysis or an autoencoder. The position of each node on a plane or space may be determined by mapping to a vector. By visualizing these nodes by adding citation-related directed edges, the position of the node indicates the degree of similarity between the ideas, making it possible to grasp at a glance the magnitude of change between the child idea and the parent idea.

In this way, ideas can be classified based on the number of votes and number of citations for ideas, and can be used for selecting and visualizing ideas.

Next, the effects of this embodiment will be explained.

The ideation support device 100 according to the first embodiment of the present invention described above includes an idea database 102 that stores ideas, the number of votes for each idea, and citation relationships, and an idea selection unit 105 that selects ideas stored in the idea database 102. , an input unit 101 for inputting votes for proposed ideas and each idea selected by the idea selection unit 105; a citation definition unit 103 for calculating the citation relationship and number of citations for each idea; and an idea classification unit 104 that classifies ideas using the number of votes and number of citations.

According to the present invention, by defining citation relationships between ideas, it is possible to classify ideas based on the number of citations and identify ideas suitable for reference, leading to suggestions of better ideas. You can increase your chances.

Furthermore, since it further includes an output unit 106 that visualizes the citation relationship of each idea calculated by the citation definition unit 103 and the classification of each idea by the idea classification unit 104, participants can You will be able to check the status of your ideas and get more great ideas.

Further, the citation definition unit 103 uses natural language processing to calculate the degree of similarity between the idea proposed by the input unit 101 and the idea selected by the idea selection unit 105, and selects the idea with the highest calculated degree of similarity or the first idea. By defining that an idea exceeding a threshold value is cited by an idea in the input unit 101, it becomes possible to more appropriately deal with cases in which it is preferable not to define quotations.

The idea classification unit 104 also refers to the number of citations and votes for each idea, and determines whether the number of citations or the citation probability calculated from the number of citations is equal to or higher than the second threshold and the number of votes or the voting probability calculated from the number of votes is the third threshold. Ideas below the threshold are classified as reference ideas recommended for citation, and the idea selection unit 105 selects ideas to be displayed in the input unit 101 from among the reference ideas, thereby selecting ideas that are more suitable for reference and have more room for refinement. Become more specific.

Furthermore, the idea classification unit 104 classifies ideas for which the number of votes or the voting probability calculated from the number of votes is equal to or higher than a fourth threshold as potential ideas, and the idea selection unit 105 selects ideas to be displayed on the input unit 101 from among the potential ideas. By making selections, you can better identify ideas that are more complete.

The output unit 106 also outputs the citation relationships between ideas in a directed graph or tree structure, and in particular, changes the color, shape, size, or font of nodes in the directed graph or tree structure for each classification of ideas. Visually easy to understand by using lines around each family lineage that connects the idea or idea by tracing directed edges from the initial idea, and using different node shapes, colors, sizes, and fonts for each family lineage. This will allow participants to deepen their understanding.

<Example 2>
An idea support device, an idea support program, and an idea support method according to a second embodiment of the present invention will be described.

In the second embodiment, the idea database 102 and the citation definition unit 103 of the first embodiment are used to collect information from the proposers of reference ideas and promising ideas, and the participants who have a large average or total improvement in the number of votes for their proposed ideas. , a construction process/method for building a team. This will be explained below.

A large number of participants are expected to come to idea generation using the ideation support device 100, but after acquiring a large number of ideas from many participants, some ideas will be selected in an offline workshop, etc., and then further detailed information will be provided. When considering this, it is desirable to limit the number of participants to a level at which a conversation can be established. Example 2 provides this process/method.

Three criteria are used when selecting participants from the perspective of idea generation. Here, we will consider an example in which six participants in a workshop are to be selected.

First, select two people who will propose a reference plan. If there are three or more proposers of reference plans, the two proposers are selected in descending order of the number of reference plans they have proposed. The proposer of a reference proposal becomes an initiator to provide a noteworthy perspective.

Next, improve the reference plan and select participants who have a higher voting probability. This calculates the improvement amount for each idea, which is the voting probability minus the average voting probability of the parent idea, and calculates the average or total improvement amount for the ideas proposed for each participant. You can choose by calculating. The average value or total sum of improvement amounts is calculated for each participant, and the two participants are selected in descending order of either of these values. These participants will be primarily responsible for improving ideas.

Finally, select two potential proposers in the same way as the reference proposal proposers. The person who proposes a promising idea suddenly proposes a great idea and becomes a provider of diversity that changes perspectives. In addition to these participants, facilitators and initiators join to build the team. In this way, teams can be generated based on the classification results of ideas.

The other configurations and operations are substantially the same as those of the ideation support device, ideation support program, and ideation support method of the first embodiment described above, and the details will be omitted.

The ideation support device, ideation support program, and ideation support method according to the second embodiment of the present invention also provide substantially the same effects as the ideation support device, ideation support program, and ideation support method according to the first embodiment described above.

In addition, by building a team from those who proposed reference ideas and promising ideas, and participants who have a large average or total improvement in the number of votes for their proposed ideas, you can have efficient discussions when you want to deepen the discussion. will be able to expand.

<Example 3>
An idea support device, an idea support program, and an idea support method according to a third embodiment of the present invention will be described.

Embodiment 3 is a prediction model that uses natural language processing to learn, in the idea selection unit 105 of Embodiment 1, the relationship between an input of an original proposal cited by a likely proposal and an output of a possible proposal, and predicts a possible proposal from the original proposal. This is a form in which ideas output from the prediction model are displayed on the input unit 101 using a reference plan as input. This will be explained below.

When generating ideas, consider generating children of reference ideas based on the relationship between a promising idea and its parent. That is, a model is created that pairs the potential ideas obtained so far with their parents, and uses the ideas of the parents as input to predict the potential ideas that are children. There are no restrictions on the method of model generation, and possible methods include generating sentence vectors using natural language processing and learning the relationship between parent and child sentence vectors using a neural network or the like.

After learning the parent-child relationship of the likely ideas, ideas are generated by inputting the reference idea and predicting the ideas that will be children of the reference idea. The ideas obtained in this way are displayed on the evaluation/proposal screen 200 in the input section 101.

The number of votes or voting probability for an idea automatically generated in this way is the reward and fitness level when optimizing the hyperparameters of a model that predicts child ideas from parent ideas using reinforcement learning or hyperheuristics. can do.

The ideation support device, ideation support program, and ideation support method according to the third embodiment of the present invention also provide substantially the same effects as the ideation support device, ideation support program, and ideation support method of the first embodiment described above.

In addition, the idea selection unit 105 uses natural language processing to learn the relationship between the input of the original draft cited by the leading proposal and the output of the leading draft, and constructs a predictive model that predicts the leading proposal from the original draft, and selects a reference proposal. By displaying ideas output from the prediction model as input on the input unit 101 and repeating idea generation and voting, the prediction model of the child idea is refined and better ideas can be generated. . In this way, ideas can be automatically generated based on the idea classification results.

<Others>
Note that the present invention is not limited to the above-described embodiments, and includes various modifications. The above-mentioned embodiments have been described in detail to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to having all the configurations described.

Furthermore, it is also possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is also possible to add, delete, or replace some of the configurations of each embodiment with other configurations.

100: Idea support device 101: Input unit 102: Idea database (database)
103: Citation definition section 104: Idea classification section 105: Idea selection section 106: Output section 200: Evaluation/proposal screen 201: Draft display section 202: Draft evaluation section 203: Proposal reception section 204: Submit button 205: Theme

Claims

a database for storing ideas, the number of votes for each idea, and citation relationships;
an idea selection unit that selects the ideas stored in the database;
an input unit for inputting votes for the proposed ideas and each of the ideas selected in the idea selection unit;
a citation definition unit that calculates the citation relationship and the number of citations for each of the ideas;
An idea support device comprising: an idea classification unit that classifies the ideas using the number of votes and the number of citations for each of the ideas.
The idea support device according to claim 1,
An ideation support device further comprising: an output unit that outputs the citation relationship of each of the ideas calculated by the citation definition unit and the classification of each of the ideas by the idea classification unit.
The idea support device according to claim 2,
The quotation definition unit calculates the degree of similarity between the idea proposed by the input unit and the idea selected by the idea selection unit using natural language processing, and selects the idea or the idea for which the calculated degree of similarity is maximum. An idea support device, characterized in that the idea exceeding a first threshold is defined as being cited by the idea of the input section.
The idea support device according to claim 3,
The idea classification unit refers to the number of citations and the number of votes of each idea, and the number of citations or the citation probability calculated from the number of citations is equal to or higher than a second threshold and the number of citations or the number of votes is calculated from the number of votes. Classify the ideas whose voting probability is below a third threshold as reference ideas recommended for citation,
An idea generation support device, wherein the idea selection unit selects the idea to be displayed on the input unit from among the reference ideas.
The idea support device according to claim 4,
The idea classification unit classifies the ideas for which the number of votes or the voting probability calculated from the number of votes is equal to or higher than a fourth threshold as potential ideas;
An idea generation support device, wherein the idea selection unit selects the idea to be displayed on the input unit from among the potential ideas.
The idea support device according to claim 5,
The idea support device is characterized in that the output unit outputs the citation relationships between the ideas in a directed graph or a tree structure.
The idea support device according to claim 6,
The idea support device is characterized in that the output unit changes the color, shape, size, or font of the nodes of the directed graph or the tree structure for each classification of the idea.
The idea support device according to claim 7,
The output unit is characterized in that the idea is surrounded by a line for each family lineage connected by tracing directed edges from the initial idea, and uses different node shapes, colors, sizes, and fonts for each family lineage. Idea support device.
The idea support device according to claim 5,
An ideation support device characterized in that a team is constructed from proposers of the reference idea and the likely idea, and participants who have a large average value or total sum of improvements in the number of votes for the proposed idea.
The idea support device according to claim 5,
The idea selection unit constructs a prediction model that predicts the likely idea from the original idea by learning a relationship between inputting an original proposal cited by the likely idea and outputting the likely idea through natural language processing; An idea support device, characterized in that the idea outputted from the prediction model using a reference idea as input is displayed on the input section.
a saving step of saving ideas, the number of votes and citation relationships of each said idea;
an idea selection step of selecting the idea to be saved in the saving step;
an input step of inputting votes for the proposed ideas and each of the ideas selected in the idea selection step;
a citation definition step of calculating the citation relationship and the number of citations for each of the ideas;
A program that causes a processing device to execute an idea classification step of classifying the ideas using the number of votes and the number of citations for each of the ideas.
a storage step of storing ideas, the number of votes for each of the ideas, and citation relationships;
an idea selection step of selecting the ideas to be saved in the storage step;
an input step of inputting votes for the proposed ideas and each of the ideas selected in the idea selection step;
a citation definition step of calculating the citation relationship and number of citations for each of the ideas;
An ideation support method, comprising: an idea classification step of classifying the ideas using the number of votes and the number of citations for each of the ideas.