JP7129163B2 - Provision device, provision method and provision program - Google Patents

Description

The present invention relates to a provision device, a provision method, and a provision program.

Conventionally, there is known a technique of searching or generating information related to input information based on analysis results of input information and outputting the retrieved or generated information as a response. As an example of such technology, the words, sentences, and context contained in the input text are converted into multidimensional vectors and analyzed, and based on the analysis results, text similar to the input text, A technique of natural language processing is known for inferring a text following a text and outputting an analogy result.

JP 2010-175614 A

“Dynamics on expanding spaces: modeling the emergence of novelties”, Vittorio Loreto, Vito D. P. Servedio, Steven H. Strogatz, Francesca Tria <Internet> https://arxiv.org/abs/1701.00994 (searched December 06, 2017) ) “Yule Process and Simon Process”, Yasuhiro Hashimoto <Internet> https://dencity.jp/misc/yule-simon_process/ (searched December 06, 2017)

Here, it is conceivable to use the technology described above to assist the dialogue between users. However, with the above-described technology, information that can be predicted by the participants, such as text similar to the input text and text following the input text, is only output at a predetermined timing. For this reason, the above-described technology may not be able to efficiently promote dialogue between users.

The present application has been made in view of the above, and aims to improve the efficiency of dialogue between users.

The providing apparatus according to the present application includes an estimating unit for estimating the growth speed of graph information showing technical elements belonging to fields to be discussed as nodes, and relationships between the elements as links between nodes, and the estimating unit and a providing unit for providing auxiliary information for assisting the dialogue according to the estimated growth rate.

According to one aspect of the embodiment, it is possible to improve the efficiency of interaction between users.

FIG. 1 is a diagram illustrating an example of processing executed by a providing device according to an embodiment. FIG. 2 is a diagram illustrating a configuration example of a provision device according to an embodiment; FIG. 3 is a diagram illustrating an example of information registered in a technical graph database according to the embodiment; FIG. 4 is a flowchart illustrating an example of the flow of provision processing according to the embodiment. FIG. 5 is a diagram illustrating an example of a hardware configuration;

Embodiments for implementing the provision device, the provision method, and the provision program (hereinafter referred to as "embodiments") according to the present application will be described in detail below with reference to the drawings. Note that the provision device, the provision method, and the provision program according to the present application are not limited by this embodiment. Also, in each of the following embodiments, the same parts are denoted by the same reference numerals, and overlapping descriptions are omitted.

[Embodiment]
[1-1. Example of provision device]
First, an example of processing executed by the providing apparatus will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of processing executed by a providing device according to an embodiment. In FIG. 1, the provision device 10 is an information processing device that executes provision processing described below, and is realized by, for example, a server device, a cloud system, or the like.

More specifically, the providing device 10 can communicate with any device such as an input/output device 100 (see FIG. 2, for example) via a predetermined network N such as the Internet. Here, the input/output device 100 acquires the user's utterance using a voice acquisition device such as a microphone that acquires voice. Then, the input/output device 100 converts the utterance into text data using any speech recognition technology, and transmits the converted text data to the providing device 10 . In addition, the input/output device 100 reads aloud the text data received from the providing device 10 using a device that outputs sound, such as a speaker. Note that the input/output device 100 may display the text data received from the providing device 10 on a predetermined display device.

Note that the input/output device 100 is realized by an information processing device such as a server device such as a smart device such as a smartphone or a tablet, a desktop PC (Personal Computer) or a notebook PC. Note that the input/output device 100 may be realized by, for example, the same information processing device, or may be realized by a device such as a robot.

[1-2. Regarding the processing of the providing device]
Here, the user's utterance is converted into a distributed representation, and based on the results of comparison with other utterances and distributed representations of information, a response to the user's utterance and a utterance that assists the dialogue between users are generated. Such processing can be considered. However, the conventional technology only outputs information that the participants can predict, such as text similar to the input text or text following the input text. For this reason, the above-described technology may not be able to efficiently promote dialogue between users. For example, with the above-described technology, it is not possible to perform output according to the situation in dialogue.

Here, the providing apparatus 10 executes the following proposal process. First, the providing device 10 estimates the growth speed of graph information indicating the relationship between each element as a link between nodes, with the technical elements belonging to the target field of the dialogue as nodes. Then, the providing device 10 provides supplementary information for assisting the dialogue according to the estimated growth speed.

For example, the providing apparatus 10 generates graph information (hereinafter referred to as "technology graph") showing the relationship between each element as a link between the nodes, using the technical elements that are the target of the dialogue as nodes. Such graph information is information indicating from what technical element the technical element that is the target of the dialogue is generated. For example, if the first technical element is a combination of the second technical element and the third technical element, or if the technology is developed from the second technical element and the third technical element, it corresponds to the first technical element The node is connected to the node corresponding to the second technical element and the node corresponding to the third technical element via links.

Here, the providing apparatus 10 may use any information as a technical element as long as it indicates a technique belonging to a certain field. For example, the providing apparatus 10 may use contents presumed to be ideas generated in a dialogue as technical elements, and may plot relationships between the technical elements into a technical graph. For example, if a plurality of existing ideas are mentioned within a predetermined period of time and then a new idea is mentioned, the providing device 10 adds a node indicating the new idea to the technology graph. At the same time, each node representing an existing idea is connected to a node corresponding to a new idea by a link.

In such a technology graph, the increasing trend of the number of nodes indicates the increasing trend of technical elements belonging to the field, that is, the increasing trend of new technical elements. If such a technical graph is developed according to dialogues between users (for example, meetings, brainstorming, etc.), it is possible to show how the dialogues are progressing. In other words, when an idea is formed in a dialogue, if that idea is converted into a node as a technical element and a link is formed with other ideas that were the source of that idea, the idea will be generated in the dialogue. If so, it can be said that the number of nodes and links included in the technical graph is on the increase.

Therefore, the providing device 10 generates a technical graph corresponding to the dialogue, and estimates the development mode of the generated technical graph. More specifically, the providing device 10 estimates an increasing tendency of the number of nodes. In addition, the providing device 10 estimates the state of the dialogue corresponding to the technical graph based on the estimated increasing trend of the number of nodes. Then, when the estimated state satisfies a predetermined condition, the providing device 10 provides the user with information that assists the dialogue (hereinafter referred to as "auxiliary information").

For example, when the growth speed of the technical graph is below a predetermined threshold, it is estimated that the dialogue between users is stagnating. However, depending on the content of the dialogue, such as when the topic is in a field where it is difficult to generate new ideas, the dialogue between users may get stagnant. cannot output. Therefore, the providing device 10 estimates the growth speed of the technical graph in the future. For example, the providing device 10 estimates the growth rate of the technology graph in the future based on the growth mode of the technology graph generated based on the interaction up to that point.

For example, the providing device 10 counts the number of new ideas generated in the dialogue, that is, the increase in the number of nodes included in the technology graph, per unit time. Then, the providing apparatus 10 calculates the average value of the increase in the number of nodes counted per unit time as the future growth rate of the technology graph. Note that the providing apparatus 10 may use an arbitrary speed estimation technique to estimate the future increase in the number of nodes, and use the estimated increase as the growth speed of the technology graph.

On the other hand, the providing device 10 estimates the growth speed of the technical graph according to the content of the actual dialogue at the processing timing. For example, the providing device 10 estimates the growth speed of the technical graph within 20 minutes from the processing timing. Then, when the growth speed of the technical graph corresponding to the actual dialogue is slower than the estimated growth speed of the technical graph, the providing device 10 determines that the dialogue between the users is stagnant, and Provide supporting information. Note that the providing device 10 may provide auxiliary information when the estimated growth speed of the technical graph is simply slower than a predetermined growth speed.

That is, the provision device 10 estimates the growth speed of the technical graph, and when the estimated growth speed of the technical graph slows down, determines that the dialogue such as brainstorming is stagnant, and moves on to the next topic or introduces a new topic. Facilitate as if forced to suggest. As a result, the providing device 10 can make interaction between users more efficient.

[1-3. About growth speed]
Here, the providing apparatus 10 may estimate the growth speed of the technology graph by any method as long as it estimates the growth speed of the technology graph based on the increasing tendency of the number of nodes included in the technology graph. .

For example, if a new idea is generated each time a user speaks, it is considered that the dialogue is progressing efficiently. On the other hand, if the number of user remarks is large compared to the number of ideas, it may not be possible to say that the dialogue is progressing efficiently. Therefore, the providing apparatus 10 may estimate the growth speed of the technical graph based on the number of nodes increased per unit time and the number of utterances in dialogue per unit time. For example, the providing device 10 may estimate the growth speed of the technical graph based on the ratio between the increase in the number of nodes and the number of statements. To give a more specific example, the providing device 10 may use a value obtained by dividing the number of increases in the number of nodes by the number of statements as the growth rate of the technology graph.

[1-4. Supplementary information]
Here, the providing apparatus 10 may use information of arbitrary content as auxiliary information. For example, when it is determined that the dialogue is stagnant, the providing device 10 may provide information to the effect that the dialogue is stagnant to the user. Information may be provided that indicates the relationship between the idea and the existing ideas.

Further, the providing device 10 may provide auxiliary information based on the growth mode of the technical graph. For example, the providing device 10 may estimate new nodes that will appear in the future from existing nodes using a model that has learned the characteristics of the growth mode of the technical graph, and provide information indicating the estimated nodes as auxiliary information. .

For example, the providing device 10 can display technical graphs such as the time it takes an existing node to generate a new node, the characteristics of an idea corresponding to an existing node that has a relationship with the new node, the number of existing nodes that generate a new node, and so on. Let various models learn the characteristics of the growth mode of Then, when the growth speed of the technical graph satisfies a predetermined threshold, that is, when it is estimated that the dialogue is stagnant, the providing apparatus 10 uses the learned model to determine whether the newly generated node and the existing node Estimate the relationship with

For example, when information about nodes and links included in a technical graph at a certain timing is input, the providing device 10 outputs information indicating newly added nodes and links at the next timing, so as to learn a model. I do. Then, when it is estimated that the dialogue is stagnant, the providing device 10 inputs the information of the technical graph at the processing timing to the model, thereby linking nodes estimated to be newly added in the future. Get information indicating

After that, the providing apparatus 10 may propose, as supplementary information, information indicating nodes and links that are estimated to be newly added in the future. For example, the providing device 10 identifies an existing node that is connected to a newly added node by a link, and inquires of the user, together with the idea corresponding to the identified existing node, whether a new idea will arise from those ideas. Information may be provided as ancillary information.

[1-5. Future technology graph]
Here, the providing apparatus 10 estimates parameters indicating the development of the technology graph based on the increase of nodes and links in the technology graph in the past, and the number of nodes and links included in the technology graph based on the estimated parameters. By changing , it is possible to predict the future development of the technical graph and thus the likelihood of generating new ideas in the dialogue.

For example, the development of technology in a certain field can be regarded as a change in the number of nodes and links in the technology graph, and such changes in the number of nodes and links can be virtually reproduced by a power distribution. (For example, see Non-Patent Document 1). Therefore, the providing apparatus 10 estimates the trend of change in the number of nodes and the number of links assuming that a new node connected to that node by a link is generated from an existing node according to a power distribution. good. Further, the providing apparatus 10 may estimate the future growth rate of the technical graph assuming that the technical graph develops according to an arbitrary power distribution such as the Yule process, Simon process, modified Yule process, or the like.

Here, various parameters indicating the development of the technical graph may be set based on actual measurements in the past. For example, the providing device 10 learns a model that generates various parameters indicating the development of the technology graph from the increase mode of nodes and links included in the technology graph, and uses the learned model based on past interactions. Various parameters may be generated from the mode of increase in nodes and links.

[1-6. About Polya's Pot Algorithm]
Here, as shown in Non-Patent Document 1, a method of mathematically representing the development of the technical field by the Polya's pot algorithm is conceivable. Therefore, the providing device 10 may develop the technology graph using Polya's pot algorithm.

An example of the Polya's Pot Algorithm is described below. For example, in the Polya's Jar Algorithm, one ball is drawn at random from a pot of different colored balls, and it is determined whether or not the drawn ball is the color of the first drawn ball. be. If the ball taken out has the same color as the ball taken out in the past, the taken out ball is returned to the jar, and a predetermined number of balls of the same color as that ball are added to the jar. . On the other hand, if the removed ball is of the same color as the first removed ball, the removed ball is returned to the urn, and a predetermined number of balls of the same color as the removed ball are added to the urn, each with a different color. A given number of balls, each with a new color, is placed in the urn.

Such a Polya's bottle algorithm can be regarded as a stochastic model of the development of technical elements in a certain field. For example, if a ball placed in a jar is a technical element, the probability that the color of the ball taken out of the jar is the color of the first ball taken out from the jar is associated with the probability of a new technical element occurring. be able to. Also, if the ball taken out is the color that was taken out for the first time, the number of balls of a new color that is newly put into the urn will be the number of new balls that will be generated in the future from that technology when a new technical element occurs. can be associated with a parameter indicating the number of technical elements. In addition, the number of balls that are put into the urn together with the removed balls and that have the same color as the balls that are removed from the urn can be associated with the relationship between the technical elements, that is, the parameter that causes the link. can.

Therefore, the providing device 10 uses Polya's Pot Algorithm to develop the technology graph and estimate the future development trend in the target field. For example, when any element is selected from a set of multiple types of elements, the providing apparatus 10 adds an element of a type different from the selected element together with the selected element to the set, and then selects each type of element. Based on the probability that an element is selected, the trend of change in the number of nodes and the number of links may be estimated, assuming that existing nodes generate new nodes that are connected to the nodes by links. .

[1-7. About the initial model of the technical graph]
Here, the providing device 10 may employ a graph having an arbitrary configuration as the initial model of the technical graph. For example, the providing device 10 uses the Barabashi-Albert model to generate an initial state of the technology graph, and from the nodes included in the generated technology graph, it is assumed that a new node connected to the node by a link is generated. , the trend of change in the number of nodes and the number of links may be estimated.

For example, the providing device 10 generates a complete graph with m nodes as an initial model of the technical graph. Subsequently, the providing device 10 adds one new node and sets a link connecting the added node and another node with a predetermined probability. By repeating such processing until the number of nodes reaches a predetermined number, the providing device 10 may generate an initial model. In addition to the techniques described above, the providing apparatus 10 may employ various techniques for generating an initial model of a complex graph, such as the Watts-Strogatz model, to generate an initial model of a technical graph. .

[1-8. Regarding graph convolution]
Note that the providing apparatus 10 estimates whether or not the dialogue is stagnant by considering the speed at which core ideas are generated among the ideas generated in the dialogue, and it is estimated that the dialogue is stagnant. In this case, auxiliary information may be output.

For example, when convolutional analysis of a technical graph is performed using a graph convolution technique such as graph convolution, it is possible to obtain the graph structure of the main nodes in the dialogue, that is, the nodes corresponding to the core ideas in the dialogue. can be done. Therefore, the providing device 10 performs convolutional analysis of the technology graph to newly generate a technology graph indicating the relationship between main nodes in the dialogue, and estimates the growth speed of the newly generated technology graph.

When the estimated growth rate is below a predetermined threshold, the providing apparatus 10 determines that even if many statements and ideas are generated in the dialogue, core ideas are not generated, that is, the dialogue is stagnant. It is also possible to estimate that the information is correct and output the auxiliary information. In this way, the providing device 10 can output auxiliary information according to whether or not a core idea is generated by performing convolutional analysis of the technical graph.

[1-9. About an example of proposal processing]
Next, an example of proposal processing executed by the providing apparatus 10 will be described with reference to FIG. First, the providing device 10 acquires statements in dialogue between users, such as brainstorming (step S1). In such a case, the providing device 10 generates a technical graph in which the ideas generated in the dialogue are nodes and the relationships between the ideas are links (step S2). For example, the providing device 10 uses various character analysis techniques to set the techniques and ideas included in the user's utterances as nodes, and the co-occurrences of the techniques and ideas included in the utterances as links. Generate technical graphs. For example, the providing apparatus 10 links nodes corresponding to technologies and ideas included in the same statement.

Subsequently, the providing device 10 causes the model to learn the characteristics of the mode of occurrence of nodes and links (step S3). For example, when the providing apparatus ₁₀ receives link information T1 indicating the state of the link at timing _T1 and node information _T1 indicating the state of the node at timing T1, the providing apparatus ₁₀ receives the link information at timing _T2 . _The learning model is _trained so as to _output link information T2 indicating the state of the node and node information T2 indicating the state of the node at timing T2.

Subsequently, the providing apparatus 10 determines the degree of stagnation of the dialogue based on the difference between the growth speed of the technical graph and the growth speed estimated by the learning model (step S4). For example, let the processing timing be _Tn . In such a case, the providing device 10 generates a technical graph based on the statements from timing T _n−1 to timing T _n and estimates the actual growth speed of the generated technical graph. For example, the providing apparatus 10 estimates the number of new ideas actually extracted from utterances from timing T _n−1 to timing T _n as the actual growth rate. On the other hand, the providing device 10 estimates the technical graph at timing T _n by inputting the technical graph at timing T _n−1 into the learning model.

Then, the providing apparatus 10 takes the difference between the number of nodes in the technology graph at timing T _n−1 and the number of nodes in the technology graph at estimated timing T _n as the estimated growth rate. Then, when the actual growth rate is lower than the estimated growth rate, the provision device 10 determines that the dialogue is stagnant.

Here, when it is estimated that the dialogue is stagnant, the providing device 10 uses a learning model to estimate newly generated nodes from existing nodes, and creates links connecting the estimated nodes and existing nodes. Estimate (step S5). For example, the providing device 10 generates a technology graph at timing _Tn (i.e., a technology graph including as nodes ideas and technologies generated from timing _T0 to timing _Tn ), and inputs the generated technology graph to a learning model. By doing so, the technical graph at timing T _n+1 is estimated.

In addition, the providing apparatus 10 compares the technical graph at timing _Tn with the technical graph at estimated timing Tn ₊ 1, and determines the newly added node from timing _Tn to timing _Tn+1 (hereinafter referred to as "new node"). ) are identified. Also, the providing apparatus 10 estimates a link indicating the relationship between the specified new node and the existing node (that is, the node included in the technology graph at timing _Tn ).

Then, the providing device 10 provides information indicating the estimated link to the user (step S6). For example, the providing device 10 identifies existing nodes to which the estimated links are connected, outputs information indicating ideas and technologies corresponding to the identified existing nodes, and asks, "Do you have any new ideas related to these? ”, etc., suggest to users to consider ideas and technologies corresponding to new nodes from existing nodes.

[2. Configuration of Providing Device]
An example of the functional configuration of the providing apparatus 10 that implements the above-described providing process will be described below. FIG. 2 is a diagram illustrating a configuration example of a provision device according to an embodiment; As shown in FIG. 2 , the providing device 10 has a communication section 20 , a storage section 30 and a control section 40 .

The communication unit 20 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 20 is connected to the network N by wire or wirelessly, and transmits and receives information to and from the input/output device 100 .

The storage unit 30 is implemented by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 30 also stores a technical graph database 31 and a model database 32 .

Technical graphs are registered in the technical graph database 31 . For example, FIG. 3 is a diagram showing an example of information registered in the technical graph database according to the embodiment. In the example shown in FIG. 3, the technical graph database 31 has items such as "timing", "breast ID (Identifier)", "link ID", "first node ID", and "second node ID". Information is registered for each technical graph.

Here, "timing" is information indicating at what timing the technical graph is the technical graph. The "Brest ID" is information indicating which brain the technical graph corresponds to, that is, which conversation of which user the technical graph corresponds to. A "link ID" is information for identifying a link included in a technical graph. The "first node ID" and "second node ID" identify nodes connected by the link indicated by the associated "link ID", that is, nodes indicating two related technical elements. It is information to do.

For example, in the example shown in FIG. 3, the technical graph information indicated by "technical graph #1" includes timing "timing #1", breast ID "breast #1", link ID "link #1", first node ID "Node #1" and the second node ID "Node #2" are associated and registered. Such information is the technical graph indicated by "technical graph #1", and the technical graph at the timing indicated by timing "timing #1" is the technical graph corresponding to the dialogue indicated by brain ID "beast #1". It indicates that the node indicated by the first node ID "node #1" and the node indicated by the second node ID "node #2" are connected by the link indicated by the link ID "link #1".

Note that in the example shown in FIG. 3, conceptual values such as "timing #1", "breast #1", "link #1", "node #1", etc. are described, but in reality, characters indicating timing Columns and numerical values, character strings and numerical values indicating breasts, character strings and numerical values indicating links and nodes, etc. are registered. Also, in the technical graph database 31, it is assumed that the information of the technical graph indicated by "technical graph #1" is registered for each timing. Also, the information shown in FIG. 3 is merely an example, and data in any format may be registered in the technical graph database 31 as long as it has a graph structure.

Returning to FIG. 2, the description is continued. A trained model is registered in the model database 32 . For example, the model database 32 includes an input layer for inputting input information that is input to a model, a plurality of intermediate layers for sequentially performing predetermined processing on the input information input to the input layer, and a plurality of intermediate layers. data of a model having an output layer that generates output information corresponding to input information based on the output of the last intermediate layer among the intermediate layers of . More specifically, the model database 32 registers data indicating the connection relationship of each node and the connection coefficient between nodes.

Here, in the model, the value is calculated based on the first element belonging to any layer from the input layer to the output layer other than the output layer, and the weight of the first element and the first element. By performing an operation based on the first element and the weight of the first element, with each element belonging to each layer other than the output layer as the first element for the information input to the input layer including the second element, The computer is made to function so that the information input to the input layer and the corresponding information are output from the output layer. The model may be assumed to be used as a program module that is part of artificial intelligence software.

In such a model, when information indicating nodes included in a technology graph at a certain timing and information indicating links are input to the input layer, information indicating nodes included in the technology graph in the future is output from the output layer. A computer is operated to output information indicating a link. For example, the model is used in a computer with a CPU and memory. Specifically, the CPU of the computer, according to instructions from the learned model stored in the memory, applies the learned weighting coefficients and responses in the neural network to the information of the technical graph input to the input layer of the model. It performs operations based on functions, etc., and operates to output information indicating future technical graphs from the output layer.

Here, when the model is realized by a neural network having one or more intermediate layers such as DNN, the first element included in each model is regarded as any node of the input layer or intermediate layer. The second element corresponds to the node to which the value is transmitted from the node corresponding to the first element, that is, the next node, and the weight of the first element corresponds to the node corresponding to the first element is the weight, ie the connection coefficient, taken into account for the value transmitted to the node corresponding to the second element.

Here, the providing device 10 uses the information registered in the technical graph database 31 to learn the learning model. That is, the information registered in the technical graph database 31 includes the input layer to which the technical graph at a predetermined timing is input, the output layer, and any layer from the input layer to the output layer, other than the output layer. and a second element whose value is calculated based on the weight of the first element and the first element, and based on the weight that reflects the characteristics of the technical graph at the input predetermined timing It is the data for causing the computer to function so as to output, from the output layer, the information indicating the technical graph at the timing later than the predetermined timing, which is the technical graph that was input by performing the calculation.

For example, the providing device 10 uses, as learning data, a first technical graph generated from a certain dialogue, based on statements from the beginning of the dialogue to a predetermined first timing, and a first technical graph from the beginning of the dialogue to the first timing. A second technical graph is generated based on the utterances up to a second timing later. Then, the providing apparatus 10 learns the model so that the information output by the model approaches the second technical graph when the first technical graph is input to the model.

Any learning method such as back propagation can be applied to model learning. By repeatedly executing such learning at various timings, the providing device 10 can create a model that has learned the characteristics of the increase mode of nodes and links in the technology graph corresponding to a certain dialogue, and obtains a model from the technology graph at a certain timing. A model can be generated that estimates the future technology graph. In other words, the weights between nodes included in the model reflect the characteristics of the increase in nodes and links in the technology graph corresponding to a certain interaction, and are used by a computer to estimate future technology graphs from input technology graphs. It is a parameter for executing

The control unit 40 is a controller, and various programs stored in a storage device inside the providing apparatus 10 are transferred to the RAM or the like by a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). It is realized by being executed as a work area. Also, the control unit 40 is a controller, and may be implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As shown in FIG. 2 , the controller 40 has a generator 41 , an estimator 42 and a provider 43 . The generator 41 generates a technical graph for each dialogue. For example, the generation unit 41 acquires a user's utterance in a dialogue from the input/output device 100 . Then, the generating unit 41 generates, for each unit time, a technology graph in which the technologies and ideas included in the user's utterance are set as nodes and links indicating the relationships between the technologies and ideas are set. Then, the generation unit 41 registers the generated technical graph in the technical graph database 31 .

The generating unit 41 generates technical graphs at predetermined time intervals, and registers the generated technical graphs in the technical graph database 31 . That is, the generating unit 41 generates, at predetermined time intervals, a technical graph showing the relationship between the ideas and technologies that the user has said so far in a dialogue such as brainstorming, and stores the generated technical graph in the technical graph database 31. sign up.

The estimating unit 42 estimates the growth speed of a technical graph that indicates the relationship between each element as a link between the nodes, with the technical elements belonging to the target field of the dialogue as nodes. For example, the estimation unit 42 estimates the growth speed of the technical graph in the future. To give a more specific example, the estimation unit 42 estimates the growth speed of the technical graph based on the number of nodes increased per unit time and the number of utterances in dialogue per unit time. For example, the estimating unit 42 determines a value obtained by dividing the increase in the number of nodes by the number of statements as the growth speed of the technical graph.

For example, the estimating unit 42 reads a model corresponding to the dialogue from the model database 32 when estimating the growth speed of the technical graph for a certain dialogue. The estimating unit 42 also reads the latest technical graph that is a technical graph corresponding to the dialogue from the technical graph database 31 and inputs the read technical graph into the model. As a result, the estimation unit 42 obtains from the model information indicating the technical graph estimated as the future technical graph.

Subsequently, the estimating unit 42 estimates the growth speed of the technical graph using the estimated future technical graph and the technical graph up to that point. That is, the estimation unit 42 estimates the future growth speed of the technical graph. For example, the estimation unit 42 calculates the difference between the number of nodes included in the estimated future technology graph and the number of nodes included in the latest technology graph generated by the generation unit 41 as the estimated growth speed. Note that the estimation unit 42 may use a value obtained by dividing the difference in the number of nodes by the average number of utterances as the estimated growth speed.

On the other hand, the estimation unit 42 waits until the generation unit 41 generates a new technical graph corresponding to the dialogue. Then, when the generation unit 41 generates a new technology graph, the estimation unit 42 determines the number of nodes included in the technology graph newly generated by the generation unit 41 and the number of nodes included in the technology graph generated last time by the generation unit 41. The actual growth rate is calculated as the difference between the number of nodes Note that the estimating unit 42 may use a value obtained by dividing the difference in the number of nodes by the average number of utterances as the actual growth speed.

The estimating unit 42 then provides the providing unit 43 with the comparison result between the actual growth rate and the estimated growth rate. For example, the estimating unit 42 determines whether the actual growth speed is lower than the estimated growth speed, whether the actual growth speed or the estimated growth speed exceeds a predetermined threshold, etc., and provides the determination result to the providing unit 43. Output.

The providing unit 43 provides supplementary information for assisting the dialogue according to the estimated growth speed. For example, the providing unit 43 provides auxiliary information when the growth speed of the technical graph corresponding to the actual dialogue is slower than the estimated growth speed of the technical graph. Moreover, the provision part 43 provides auxiliary information, when the growth speed of the estimated technical graph is slower than a predetermined growth speed. The providing unit 43 also uses a model that has learned the characteristics of the growth mode of the technical graph to estimate new nodes that will appear in the future from existing nodes, and provides information indicating the estimated nodes as auxiliary information.

For example, the provision unit 43 determines that the dialogue between users is stagnant when the actual growth rate is lower than the estimated growth rate, or when the actual growth rate or the estimated growth rate is lower than a predetermined threshold. Supplementary information will be provided in such cases. For example, the providing unit 43 generates information for outputting from the input/output device 100 a message indicating that the dialogue is stagnant, and transmits the generated information to the input/output device 100 as auxiliary information.

In addition, the provision part 43 performs the following processes, when making the information which shows the estimated node into auxiliary information. First, the providing unit 43 reads the latest technical graph, which is the technical graph corresponding to the dialogue, from the technical graph database 31 . Also, the providing unit 43 reads a model corresponding to the dialogue from the model database 32 . Then, the providing unit 43 inputs the latest technical graph for the read model and obtains the future technical graph estimated by the model.

Here, the providing unit 43 compares the future technology graph and the latest technology graph, and identifies new nodes that are estimated to occur in the future. The providing unit 43 also identifies an existing node connected to the identified node by a link. Then, the providing unit 43 generates auxiliary information indicating that a new node may be generated from the identified existing node, and transmits the generated auxiliary information to the input/output device 100 .

[3. Example of Flow of Processing Executed by Providing Apparatus]
Next, an example of the flow of provision processing executed by the provision device 10 will be described with reference to FIG. FIG. 4 is a flowchart illustrating an example of the flow of provision processing according to the embodiment. First, the providing device 10 acquires a user's speech in a dialogue via the input/output device 100 (step S101). In such a case, the providing device 10 generates a technical graph from the acquired user's statement (step S102). For example, the providing device 10 uses technical elements such as ideas and techniques included in user's remarks as nodes, and generates links connecting technical elements having co-occurrence.

Subsequently, the providing device 10 causes the model to learn the growth mode of the technical graph (step S103). Then, the providing apparatus 10 determines whether or not the growth speed of the technical graph is below a predetermined threshold (step S104). For example, the provision device 10 obtains the estimated growth rate using a model and estimates whether the actual growth rate is lower than the estimated growth rate. Then, when the growth rate of the technical graph is below the predetermined threshold (step S104: Yes), the providing apparatus 10 uses the learning model to estimate new nodes generated from the existing nodes (step S105).

For example, the providing device 10 estimates a future technology graph by inputting the latest technology graph into the model, and identifies estimated nodes that are newly added nodes in the future technology graph. Then, the providing device 10 estimates a link indicating the relationship between the existing node and the estimated node (step S106), provides information indicating the estimated estimated node and the link (step S107), and ends the process. . On the other hand, when the growth speed of the technical graph is equal to or higher than the predetermined threshold (step S104: No), the providing device 10 executes step S101.

[4. Modification]
An example of the provision processing by the provision device 10 has been described above. However, embodiments are not so limited. Variations of the provision processing executed by the provision device 10 will be described below.

[4-1. Device configuration〕
In the example described above, the provision device 10 executed the provision processing within the provision device 10 . However, embodiments are not so limited. For example, the providing device 10 includes a generating device that sets the initial model and parameters of the technical graph, an estimating device that estimates whether or not the dialogue is stagnant based on the growth speed of the technical graph, and information based on the estimation result may be realized by the cooperative operation of providing devices that provide the In such a case, the generating device has the generating unit 41 shown in FIG. 2, the estimating device has the estimating unit 42 shown in FIG. 2, and the providing device has the providing unit 43 shown in FIG. Become. Also, the providing device 10 may store the technical graph database 31 in an external storage server.

Further, the providing device 10 may have a function as a learning device for learning a model. In such a case, the control unit 40 of the provision device 10 has, for example, a learning unit for performing learning processing. The learning unit corrects the connection coefficients (i.e., weighting coefficients) between the nodes of the model so that when a technical graph at a certain timing is input to the model, a technical graph at a later timing is output. This allows the model to learn the characteristics of how the technical graph grows based on interactions.

[4-2. others〕
Further, among the processes described in the above embodiments, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being performed manually can be performed manually. All or part of this can also be done automatically by known methods. In addition, information including processing procedures, specific names, and various data and parameters shown in the above text and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution and integration of each device is not limited to the one shown in the figure, and all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured.

Moreover, each of the embodiments described above can be appropriately combined within a range that does not contradict the processing contents.

[5. program〕
Also, the providing apparatus 10 according to the above-described embodiments is implemented by a computer 1000 configured as shown in FIG. 5, for example. FIG. 5 is a diagram illustrating an example of a hardware configuration; A computer 1000 is connected to an output device 1010 and an input device 1020, and an arithmetic device 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected via a bus 1090. have

The arithmetic device 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050, programs read from the input device 1020, and the like, and executes various processes. The primary storage device 1040 is a memory device such as a RAM that temporarily stores data used by the arithmetic device 1030 for various calculations. The secondary storage device 1050 is a storage device in which data used for various calculations by the arithmetic device 1030 and various databases are registered, and is implemented by a ROM (Read Only Memory), HDD, flash memory, or the like.

The output IF 1060 is an interface for transmitting information to be output to the output device 1010 that outputs various types of information such as a monitor and a printer. It is realized by a connector conforming to a standard such as HDMI (registered trademark) (High Definition Multimedia Interface). Also, the input IF 1070 is an interface for receiving information from various input devices 1020 such as a mouse, keyboard, scanner, etc., and is realized by, for example, USB.

Note that the input device 1020 includes, for example, optical recording media such as CDs (Compact Discs), DVDs (Digital Versatile Discs), PDs (Phase change rewritable discs), magneto-optical recording media such as MOs (Magneto-Optical discs), and tapes. It may be a device that reads information from a medium, a magnetic recording medium, a semiconductor memory, or the like. Also, the input device 1020 may be an external storage medium such as a USB memory.

Network IF 1080 receives data from other devices via network N and sends the data to arithmetic device 1030, and also transmits data generated by arithmetic device 1030 via network N to other devices.

The arithmetic device 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070 . For example, arithmetic device 1030 loads a program from input device 1020 or secondary storage device 1050 onto primary storage device 1040 and executes the loaded program.

For example, when the computer 1000 functions as the provision device 10 , the arithmetic device 1030 of the computer 1000 implements the functions of the control unit 40 by executing programs loaded on the primary storage device 1040 .

[6. effect〕
As described above, the providing device 10 estimates the growth speed of a technology graph that indicates the relationship between each element as a link between nodes, with the technology elements belonging to the field to be the target of dialogue as nodes. Then, the providing device 10 provides supplementary information for assisting the dialogue according to the estimated growth speed. In this way, the providing device 10 estimates the progress of the dialogue (for example, the generation of ideas and technical elements) based on the technical graph, and provides the auxiliary information at the timing according to the estimation result. As a result of being able to provide the auxiliary information in a timely fashion, it is possible to improve the efficiency of dialogue between users.

In addition, the providing device 10 estimates the growth speed of the technical graph in the future, and provides auxiliary information when the growth speed of the technical graph corresponding to the actual dialogue is slower than the estimated growth speed of the technical graph. do. Further, the providing device 10 provides auxiliary information when the estimated growth speed of the technical graph is slower than the predetermined growth speed. Therefore, the provision device 10 provides the auxiliary information when the dialogue is stagnant, so that the efficiency of the dialogue between the users can be improved.

Also, the providing apparatus 10 estimates the growth speed of the technical graph based on the number of nodes increased per unit time and the number of utterances in dialogue per unit time. For example, the providing apparatus 10 uses a value obtained by dividing the number of increases in the number of nodes by the number of statements as the growth rate of the technology graph. Therefore, the providing apparatus 10 can appropriately estimate whether or not the dialogue is stagnant.

The providing device 10 also uses a model that has learned the characteristics of the growth mode of the technical graph to estimate new nodes that will appear in the future from existing nodes, and provides information indicating the estimated nodes as auxiliary information. Therefore, the providing device 10 can estimate an idea or technical element newly generated in the dialogue, and provide information indicating the estimated idea or technical element as auxiliary information.

As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to carry out the invention in other forms with modifications.

Also, the "section, module, unit" described above can be read as "means" or "circuit". For example, the generating unit can be read as generating means or a generating circuit.

REFERENCE SIGNS LIST 10 provision device 20 communication unit 30 storage unit 31 technical graph database 32 model database 40 control unit 41 generation unit 42 estimation unit 43 provision unit 100 input/output device

Claims (6)

Graph information in which links are set between a plurality of nodes corresponding to a plurality of technical elements having co-occurrence, with technical elements belonging to a predetermined field to be the subject of dialogue as nodes, which is known at a first time point. the difference between the first graph information whose nodes are the technical elements currently in use and the second graph information whose nodes are the technical elements known at a second time point prior to the first time point , and the second graph information and , the future growth rate of the graph corresponding to the predetermined field is estimated based on the average value of the difference from the third graph information whose nodes are the technical elements known at the third point in time prior to the second point in time. an estimator that
and a providing unit that provides supplementary information set in advance as information that prompts dialogue regarding the predetermined field when the growth rate estimated by the estimating unit is below a predetermined threshold. The estimating unit calculates a growth rate of the graph information based on the number of nodes increased in the first graph information with respect to the second graph information and the number of utterances in the dialogue from the second point in time to the first point in time. 2. The providing device according to claim 1, characterized by estimating 3. The providing apparatus according to claim 2, wherein the estimating unit sets a value obtained by dividing the number of increases in the number of nodes by the number of times of utterances as the growth speed of the graph information. The provision unit estimates new nodes that will be generated in the future from existing nodes using a model that has learned the characteristics of the growth mode of the graph information, and estimates information indicating the estimated nodes and information that encourages dialogue regarding the predetermined field. 4. The providing device according to any one of claims 1 to 3, wherein information set in advance as the auxiliary information is provided as the auxiliary information. A provision method executed by a provision device,
Graph information in which links are set between a plurality of nodes corresponding to a plurality of technical elements having co-occurrence, with technical elements belonging to a predetermined field to be the subject of dialogue as nodes, which is known at a first time point. the difference between the first graph information whose nodes are the technical elements currently in use and the second graph information whose nodes are the technical elements known at a second time point prior to the first time point , and the second graph information and , the future growth rate of the graph corresponding to the predetermined field is estimated based on the average value of the difference from the third graph information whose nodes are the technical elements known at the third point in time prior to the second point in time. an estimation step to
and a providing step of providing supplementary information set in advance as information that encourages dialogue regarding the predetermined field when the growth rate estimated by the estimating step is below a predetermined threshold. Graph information in which links are set between a plurality of nodes corresponding to a plurality of technical elements having co-occurrence, with technical elements belonging to a predetermined field to be the subject of dialogue as nodes, which is known at a first time point. the difference between the first graph information whose nodes are the technical elements currently in use and the second graph information whose nodes are the technical elements known at a second time point prior to the first time point , and the second graph information and , the future growth rate of the graph corresponding to the predetermined field is estimated based on the average value of the difference from the third graph information whose nodes are the technical elements known at the third point in time prior to the second point in time. an estimation procedure for
and a provision program for causing a computer to execute: a provision procedure for providing supplementary information preset as information that prompts dialogue on the predetermined field when the growth rate estimated by the estimation procedure is below a predetermined threshold.

Images