JP2013012012A - Dialogue rule alteration device, dialogue rule alteration method, and dialogue rule alteration program - Google Patents

Abstract

Disclosed is a dialog rule changing device and a dialog rule changing method capable of changing a dialog rule of a dialog agent system in accordance with the user's preference without imposing a burden on the user.
An interaction rule changing device includes a detection node having a processing procedure corresponding to an input message, an operation node having a processing procedure corresponding to a response message, and a connection relationship between the detection node and the operation node. A dialogue rule DB 134 for storing data in which at least one of a plurality of categories is attached to a message corresponding to each node as a dialogue rule, and an action history on the user's web An interaction rule changing unit 144 that changes the interaction rule based on the user interest category determined based on the message category corresponding to each node.
[Selection] Figure 2

Description

  The present invention relates to an interaction rule changing device, an interaction rule changing method, and an interaction rule changing program for changing an interaction rule of an interaction agent system that outputs a response message to an input message in accordance with an interaction rule set in advance.

  In recent years, an interactive agent system that outputs a message corresponding to a message to the user terminal in response to receiving the message from the user terminal has become widespread. In such an interactive agent system, a method for outputting various messages to a message received from a user terminal or outputting a message suitable for the user of the user terminal has been proposed.

  For example, Patent Literature 1 stores a relationship between a message from a user and a message output corresponding to the message as a knowledge structure (dialog rule), and in response to receiving a message from the user, A method for changing the dialogue rule based on the word included in the message has been proposed.

JP 2009-223720 A

  However, in the method described in Patent Document 1, when a user wants to change a predetermined message in the dialog rule, the dialog agent system and the message are not updated until a message directly related to the predetermined message is output from the dialog agent system. It is necessary to repeat the exchange. Therefore, in the method described in Patent Document 1, it is complicated for the user to change the conversation rule corresponding to a desired message.

  Furthermore, in the method described in Patent Document 1, the interaction rule is changed by the user exchanging messages with the interaction agent system. For this reason, in the exchange of messages, the user inputs a message that does not conform to his / her preference, and as a result, there is a possibility that the conversation rule is changed to a user's preference. Therefore, in the method described in Patent Document 1, there is a possibility that the user cannot change the conversation rule according to his / her preference.

  The present invention provides an interaction rule changing device, an interaction rule changing method, and an interaction rule changing program capable of changing an interaction rule of an interaction agent system according to the user's preference without imposing a burden on the user. With the goal.

  (1) An interaction rule changing device for changing the interaction rule of an interaction agent system that outputs a response message to an input message in accordance with a predetermined interaction rule, and a processing procedure corresponding to the input message is defined Data composed of a detection node, an operation node for which a processing procedure corresponding to the response message is determined, and an edge indicating a connection relationship between the detection node and the operation node, Dialog rule storage means for storing data to which a message category is attached as the dialog rule, a user interest category determined based on a user's behavior history on the web, and a message category corresponding to each of the nodes A dialogue rule changing means for changing the dialogue rule based on the dialogue rule. Le changing device.

  The dialog rule changing device of (1) determines a dialog rule based on the user interest category determined based on the user's behavior history on the web and the message category corresponding to each node by the dialog rule changing means. change.

  As described above, the dialog rule changing device (1) changes the dialog rule according to the user interest category based on the user's behavior history, and therefore does not need to accept the dialog rule changing operation from the user. Furthermore, since the user interest category based on the user's behavior history is in accordance with the user's preference, the dialog rule changing device in (1) is in accordance with the user's preference without imposing a burden on the user. The interaction rule of the interaction agent system can be changed.

  (2) A dialog rule generating unit that generates a new dialog rule including at least one of the detection node and the action node to which the user interest category is attached, and the dialog rule changing unit is stored in the dialog rule storage unit. The dialog rule changing device according to (1), wherein a new dialog rule generated by the dialog rule generating means is added to the stored dialog rule.

  In the dialog rule changing device of (2), the dialog rule generating means generates a new dialog rule including at least one of the detection node and the action node to which the user interest category is attached, and the dialog rule changing means stores the dialog rule storage. A new dialogue rule generated for the dialogue rule stored in the means is added. By doing in this way, the dialog rule change apparatus of (2) can add the dialog rule corresponding to a user interest category newly, and can increase the dialog rule corresponding to a user preference.

  (3) The category has a hierarchy, and the interaction rule changing unit is a detection node or action node to which a category higher than the user interest category in the interaction rule stored in the interaction rule storage unit is attached. The dialog rule changing device according to (2), wherein a new dialog rule generated by the dialog rule generating means is added so as to be subordinate.

  In the dialog rule changing device of (3), the dialog rule is applied to the detection node or the action node to which the higher category of the user interest category in the dialog rule stored in the dialog rule storage means is added by the dialog rule changing means. A new dialogue rule generated by the generation unit is added so as to be in a lower level. By doing in this way, the dialog rule changing device of (3) can make a transition from the detection node or the action node with the higher category of the user interest category to the dialogue corresponding to the user interest category without a sense of incongruity.

  (4) The dialogue rule generation unit extracts a node connected to a lower level of a node to which a higher category of the user interest category in the dialogue rule stored in the dialogue rule storage unit is extracted. Change the category attached to the node to the user interest category and change the message corresponding to the extracted node to a message corresponding to both the user interest category and the higher category. The dialog rule changing device according to (3), which generates a dialog rule.

  In the dialog rule changing device of (4), the dialog rule generating means extracts a node connected to a lower level of the node to which the higher category of the user interest category in the dialog rule stored in the dialog rule storing means is attached, Change the category attached to the extracted node to the user interest category, and change the message corresponding to the extracted node to a message corresponding to both the user interest category and the higher-level category. Generate rules. By doing in this way, the dialog rule change apparatus of (4) can generate | occur | produce the node to which the user interest category was attached | subjected efficiently using the structure of the existing node.

  (5) The dialogue rule changing unit is indirectly connected to an action node to which the user interest category in the dialogue rule stored in the dialogue rule storage unit is assigned and a higher level of the action node via a plurality of edges. The dialog rule changing device according to any one of (1) to (4), wherein an edge that connects a detection node that is connected to the communication rule is added to the dialog rule stored in the dialog rule storage means.

  In the dialog rule changing device of (5), the dialog rule changing means assigns an action node to which the user interest category in the dialog rule stored in the dialog rule storage means is attached, and the action node via a plurality of edges. An edge that connects a detection node that is indirectly connected to a higher level is added to the dialogue rule stored in the dialogue rule storage means. By doing in this way, the dialog rule changing device of (5) can make a transition to a dialog corresponding to the user's taste with a small number of dialogs.

  (6) The dialogue rule changing means is indirectly connected to a higher level through a plurality of edges to the operation node to which the user interest category in the dialogue rule stored in the dialogue rule storage means is attached. The interaction rule changing device according to any one of (1) to (5), wherein a response message corresponding to an operation node being changed is changed to a response message including a name of the user interest category.

  In the dialog rule changing device of (6), the dialog rule changing means indirectly indirectly through a plurality of edges with respect to the operation node to which the user interest category in the dialog rule stored in the dialog rule storage means is attached. The response message corresponding to the operation node connected to the user is changed to a response message including the name of the user interest category. Here, when a response message including the name of the user interest category is output, the user is more likely to input a message corresponding to the user interest category than when the response message is not output. Therefore, the dialog rule changing device of (6) can increase the probability of transition to an operation node corresponding to the user interest category.

  (7) The dialogue rule storage means stores a transition probability to the transition destination operation node associated with the edge in association with the edge, and the dialogue rule change means stores the dialogue rule storage. In the case where a plurality of edges corresponding to detection nodes to which a higher category of the user interest category is added are stored in the interaction rule stored in the means, the action node to which the user interest category is attached When the corresponding edge is included in the plurality of edges, the plurality of edges so that the transition probability of the edge corresponding to the operation node with the user interest category is higher than the transition probability of the other edges. The dialog rule changing device according to any one of (1) to (6), which adjusts each transition probability.

  In the dialog rule changing device of (7), when a plurality of edges corresponding to a detection node to which a higher category of the user interest category is assigned are stored, an edge corresponding to an operation node to which the user interest category is assigned Are included in the plurality of edges, the transition probability of each of the plurality of edges is set so that the transition probability of the edge corresponding to the operation node assigned with the user interest category is higher than the transition probability of the other edges. adjust. By doing in this way, the dialog rule changing device of (7) increases the probability of transition to the operation node to which the user interest category is attached, and copes with the user's preference between the user and the dialog agent system. It is possible to make it easy to have a dialogue.

  (8) An action history storage unit that stores an action history on the web of the user; an interest category determination unit that determines the user interest category based on the action history stored in the action history storage unit; The interaction rule changing unit changes the interaction rule based on the user interest category determined by the interest category determination unit and the message category corresponding to each of the nodes. 7) The dialogue rule changing device according to any one of 7).

  In this way, in the dialog rule changing device of (8), the interest category determining means determines the user interest category based on the action history stored in the action history storage means, and the dialog rule changing means determines the interest category. The interaction rule can be changed based on the user interest category determined by the determination unit and the message category corresponding to each node.

  (9) A dialogue rule changing method in which a computer changes the dialogue rule of the dialogue agent system that outputs a response message to an input message according to a predetermined dialogue rule, and a processing procedure corresponding to the input message is determined. The detection node, the operation node for which the processing procedure corresponding to the response message is determined, and the edge indicating the connection relationship between the detection node and the operation node, And changing the interaction rule, which is data to which the message category is attached, based on the user interest category determined based on the user's behavior history on the web, and the message category corresponding to each of the nodes, How to change dialogue rules.

  (10) A dialog rule changing program for causing a computer to execute the change of the dialog rule of the dialog agent system that outputs a response message to an input message according to a predetermined dialog rule, and a processing procedure corresponding to the input message Each of the nodes is a data comprising a detection node for which a response procedure is determined, an operation node for which a processing procedure corresponding to the response message is determined, and an edge indicating a connection relationship between the detection node and the operation node The interaction rule, which is the data with the message category attached to the user, is changed based on the user interest category determined based on the user's behavior history on the web and the message category corresponding to each node. Step for causing the computer to execute an interactive rule change program Lamb.

  According to these dialogue rule changing method and dialogue rule program, the same effects as the invention according to (1) are obtained.

  According to the present invention, there is provided an interaction rule changing device, an interaction rule changing method, and an interaction rule changing program capable of changing an interaction rule of an interaction agent system according to the user's preference without imposing a burden on the user. can do.

It is a figure which shows the relationship between the dialogue rule change apparatus which concerns on 1st Embodiment, and its peripheral device. It is a figure which shows the function structure of the dialogue rule change apparatus which concerns on 1st Embodiment. It is a figure which shows action history DB which concerns on 1st Embodiment. It is a figure which shows category dictionary DB which concerns on 1st Embodiment. It is a figure which shows category DB which concerns on 1st Embodiment. It is a figure which shows dialog rule DB which concerns on 1st Embodiment. It is a schematic diagram of the dialogue rule which concerns on 1st Embodiment. It is a flowchart which shows the flow of the process which outputs a response message according to the dialogue rule predetermined with respect to the input message by the dialogue rule change apparatus which concerns on 1st Embodiment. It is a figure which shows the function structure of the dialogue rule change apparatus which concerns on 2nd Embodiment. It is a figure which shows dialog rule DB which concerns on 2nd Embodiment. It is a schematic diagram of the dialogue rule which concerns on 2nd Embodiment. It is the figure (the 1) which shows the example by which the dialogue rule which concerns on 2nd Embodiment was changed. It is a figure (the 2) which shows the example by which the dialogue rule which concerns on 2nd Embodiment was changed. It is a figure (the 3) which shows the example by which the dialogue rule which concerns on 2nd Embodiment was changed.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[First Embodiment]
[Function configuration]
FIG. 1 is a diagram illustrating a relationship between the dialog rule changing device 1 according to the first embodiment and its peripheral devices. The dialog rule changing device 1 is communicably connected to the dialog agent system 2 and the plurality of web servers 3 via a communication network such as a LAN.

  The dialogue rule changing device 1 changes the dialogue rule of the dialogue agent system 2 that outputs a response message in accordance with a dialogue rule set in advance with respect to the input message. Details of the dialog rule changing device 1 will be described later.

  The dialogue agent system 2 is communicably connected to the dialogue rule changing device 1 via a communication network such as a LAN, and is connected to a plurality of user terminals 4 via a communication network such as a LAN or the Internet. The dialogue agent system 2 receives an input message from the user terminal 4. Then, the dialogue agent system 2 outputs a response message corresponding to the accepted input message to the user terminal 4 in accordance with a preset dialogue rule.

  The web server 3 is a search server, a mini blog server, a blog server, or the like, and provides various services to a plurality of user terminals 4 on the web.

  In addition, the arrow shown in FIG. 1 has shown the flow of data, and the character string shown with the arrow has shown the content of data. For example, an arrow from the dialogue rule changing device 1 to the dialogue agent system 2 indicates that the dialogue rule is transmitted (or referred to) from the dialogue rule changing device 1 to the dialogue agent system 2. An arrow from the dialogue agent system 2 to the dialogue rule changing device 1 indicates that a dialogue log is transmitted from the dialogue agent system 2 to the dialogue rule changing device 1. An arrow from the web server 3 to the interaction rule changing device 1 indicates that information related to the action history is transmitted from the web server 3 to the interaction rule changing device 1.

FIG. 2 is a diagram illustrating a functional configuration of the conversation rule changing device 1 according to the first embodiment.
The interaction rule changing device 1 includes an operation unit 11, a display unit 12, a storage unit 13, and a control unit 14.

The operation unit 11 includes, for example, a keyboard and a pointing device (mouse, tracking ball, etc.), and receives a direct input from the operator.
The display unit 12 is configured by various displays such as a liquid crystal display and a plasma display, for example, and displays information related to the function of the dialog rule changing device 1.

  The storage unit 13 includes an action history DB 131 as an action history storage unit, a category dictionary DB 132 as a category dictionary information storage unit, a category DB 133, and an interaction rule DB 134 as an interaction rule storage unit.

  The action history DB 131 stores a use history of the web server 3 for the user terminal 4 as an action history. The action history includes at least one of the search history in the web server 3, the content posted on the blog or miniblog, and the content of the web page accessed by the user terminal 4.

FIG. 3 is a diagram illustrating the action history DB 131 according to the first embodiment.
As illustrated in FIG. 3, the behavior history DB 131 stores a user ID for identifying a user, a behavior type of the user's behavior history, and the content of the behavior history in association with each other. The content of the action history indicates the search keyword when the action type is “search”, indicates the posted content when the action type is “mini-blog” or “blog”, and indicates the viewed web when it is “web browsing” Shows a snippet of the page. Note that the behavior history stored in the behavior history DB 131 includes other behavior histories such as a dialogue log with the user terminal 4 in the dialogue agent system 2, a message body of an email, and a purchase history of a product in web shopping. .

FIG. 4 is a diagram showing the category dictionary DB 132 according to the first embodiment.
As shown in FIG. 4, the category dictionary DB 132 stores the category and the word corresponding to the category in association with each other as category dictionary information.

FIG. 5 is a diagram showing the category DB 133 according to the first embodiment.
As shown in FIG. 5, the category has a plurality of hierarchies. That is, the category DB 133 stores an upper category and a lower category, which is a category one level lower than the upper category, in association with each other.

  As shown in FIG. 5, for example, a “Roppongi” category exists as an upper category of the “Curry store: Roppongi” category, and a “Jimbocho” category exists as a category higher than the “Curry store: Jimbocho” category. Furthermore, as an upper category of the “Roppongi” category and the “Jimbocho” category, there is a “region” category.

The dialogue rule DB 134 stores dialogue rules.
The dialogue rule is composed of a detection node in which a processing procedure corresponding to an input message is defined, an operation node in which a processing procedure corresponding to a response message is defined, and an edge indicating a connection relationship between the detection node and the operation node. The data of the graph structure.
In the first embodiment, only one dialogue rule is stored in the dialogue rule DB 134, and this dialogue rule is used by users of a plurality of user terminals 4 via the dialogue agent system 2.

FIG. 6 is a diagram showing the dialogue rule DB 134 according to the first embodiment.
As shown in FIG. 6, the interaction rule DB 134 includes a node table 135 that stores information related to nodes, and an edge table 136 that stores information related to edges.

  The node table 135 associates and stores a node ID for identifying a node, a node type indicating a node type, a processing content indicating a processing procedure corresponding to a message, and a category assigned to the node.

  That is, in the dialogue rule stored in the dialogue rule DB 134, a category corresponding to the message (message corresponding to the processing procedure of each node) is assigned to each node (detection node and operation node). Here, the category is attached to the message means that the category information (for example, category name or category code for identifying the category) indicating the category is attached to the message.

  The edge table 136 stores a connection source node ID indicating the node ID of the connection source node and a connection destination node ID indicating the node ID of the connection destination node in association with each other.

  One of the connection source node and the connection destination node is a detection node, and the other is an operation node. That is, the connection source node and the connection destination node indicate a connection relationship between the detection node and the operation node, that is, an edge. When there are a plurality of operation nodes corresponding to the detection node corresponding to the interaction rule of the first embodiment, the interaction agent system 2 randomly identifies one operation node and makes a transition to the operation node.

FIG. 7 is a schematic diagram of a dialogue rule according to the first embodiment.
A broken line block shown in FIG. 7 indicates a detection node, and a solid line block indicates an operation node. Also, node IDs are assigned to these blocks. Moreover, the arrow which connects each block has shown the edge, and specifically, the start point (side without an arrow) shows a connection origin node, and the end point (side with an arrow) has shown the connection destination node. For example, the arrow connecting the node with the node ID “N1” and the node with the node ID “N2” is an operation in which the connection source node is the detection node with the node ID “N1” and the connection destination node is the node ID “N2”. Indicates a node.

  Returning to FIG. 2, the control unit 14 comprehensively controls each function related to the dialogue rule changing device 1. The control unit 14 includes a totaling unit 141, an interest category determination unit 142 as an interest category determination unit, an interaction rule generation unit 143 as an interaction rule generation unit, and an interaction rule change unit 144 as an interaction rule change unit. .

The totaling unit 141 totals the action history of each of a plurality of users for each category.
Specifically, the totalization unit 141 performs the following process for each user ID stored in the action history DB 131. That is, the totalization unit 141 performs morphological analysis of the action history stored in the action history DB 131 and extracts words. The totaling unit 141 classifies the extracted words into categories stored in the category dictionary DB 132, and totals the classified history for each category, thereby totaling the action history for each category. In addition, when extracting the word from the action history, the counting unit 141 may extract only one word when a plurality of the same words are included in one action history.

  Further, when the action history stored in the action history DB 131 is a snippet of a browsed web page, the counting unit 141 may classify the snippet into a category using a document classifier. That is, after learning the document classifier using a combination of a snippet and a category corresponding to the snippet as teacher data in advance, the counting unit 141 uses the document classifier to classify the snippet of the browsed web page as a category. You may make it classify | categorize into.

  The interest category determination unit 142 determines the user interest category, which is the category of the user's interest topic, for each user ID based on the behavior history stored in the behavior history DB 131. Specifically, the interest category determination unit 142 determines, for each user ID, one or a plurality of categories having a relatively high aggregate value aggregated by the aggregation unit 141 as the user interest category.

  The interaction rule generation unit 143 generates, for each of a plurality of users, a new interaction rule including at least one of a detection node and an action node to which the user interest category determined by the interest category determination unit 142 is attached.

  Specifically, the dialogue rule generation unit 143 extracts a node connected to a lower level of a node to which a higher category of the user interest category in the dialogue rule stored in the dialogue rule DB 134 is added. The attached category is changed to the user interest category, and a message corresponding to the extracted node is changed to a message corresponding to both the user interest category and the upper category, thereby generating a new interaction rule.

The processing of the dialogue rule generation unit 143 will be described in detail below.
First, the dialogue rule generation unit 143 refers to the category DB 133 and identifies one higher-level category among categories one level higher than the user interest category. Then, the dialogue rule generation unit 143 identifies the node (detection node and action node) to which the identified higher category is added in the dialogue rule stored in the dialogue rule DB 134, and directly or directly below this node. Extract one or more indirectly connected nodes. Here, extraction refers to duplicating a new node based on the node that is the extraction target in the conversation rule stored in the conversation rule DB 134, and the content of the node that is the extraction target is changed. is not.

  Here, when the identified higher category is a category one level higher than the user interest category, the dialogue rule generation unit 143 performs the process from the node to which the higher category is attached to the node one level lower. One or more nodes are extracted for.

  On the other hand, when the identified higher category is a category higher in the hierarchy than the user interest category, the interaction rule generation unit 143 starts from the node with the higher category to the node lower in the hierarchy. A plurality of nodes are extracted for. For example, when the upper category is a category that is three layers higher than the user interest category, the interaction rule generation unit 143 extracts the lower layer of three layers from the node with the higher category.

  Subsequently, the interaction rule generation unit 143 changes the category attached to the node to the user interest category for at least a part of the extracted nodes, and sends a message corresponding to the extracted node to the user interest category, A new interaction rule is generated by changing the message to correspond to both the higher category of the user interest category.

  Here, when the identified upper category is a category one level higher than the user interest category, the dialogue rule generation unit 143 changes the node having the higher category from the node to which the upper category is added to the node one layer lower. The attached category is changed to the user interest category.

  On the other hand, when the identified higher category is a category that is higher in the hierarchy than the user interest category, the interaction rule generation unit 143 is attached to the lowermost node among the extracted nodes in the multiple hierarchy. Change the category to the user interest category. Then, the dialogue rule generation unit 143 changes the category assigned to the node one level higher than the lowest layer node to the category one level higher than the user interest category. As described above, the dialogue rule generation unit 143 changes the category assigned to each node in the plurality of hierarchies to a category corresponding to each hierarchy.

  Further, the dialogue rule generation unit 143 changes the message of the node whose category is changed to the user interest category as follows. First, the dialogue rule generation unit 143 extracts an action history associated with the user ID corresponding to this user interest category from the action history DB 131. Then, the dialogue rule generation unit 143 performs morphological analysis on the extracted action history to extract words, and from the extracted words, co-occurrence for both the user interest category and a category higher than the user interest category Identify words with relatively high rates. And the dialogue rule production | generation part 143 changes the message before a change to the message containing the word with a relatively high co-occurrence rate. In addition, when specifying a word with a high co-occurrence rate, the dialogue rule generation unit 143 is not limited to the action history stored in the action history DB 131, but based on the search result of the web search and the content of the web page. You may make it identify the word with a high incidence rate.

  A specific example will be described with reference to FIGS. For example, when the user interest category is “curry store: Jimbocho”, the dialogue rule generation unit 143 specifies “region” as a higher category. As shown in FIG. 5, the “region” category is a category two levels higher than the “curry store: Jimbocho” category. Subsequently, the dialogue rule generation unit 143 includes a plurality of nodes (node IDs “N4”, “N6”, “N7”) from the node assigned the upper category “region” to the node lower by two layers. ”Node) is extracted (duplicated). Subsequently, the dialogue rule generation unit 143 selects the category “curry: Roppongi” attached to the lowermost node (nodes with node IDs “N6” and “N7”) among the extracted nodes in the plurality of layers. Change the user interest category to “Curry: Jimbocho”. Here, the name of the store corresponding to the user interest category “Curry: Jimbocho” is searched by searching the external gourmet site etc. with “Curry” and “Jimbocho” for the messages of the lowest layer node among the extracted nodes of multiple layers. Change to a message containing Subsequently, the dialogue rule generation unit 143 sets the category “Roppongi” attached to the node in the hierarchy one level higher than the lowest layer node (the node having the node ID “N4”) as a category higher than the user interest category. Change to “Jimbocho”.

  In the above embodiment (first embodiment), when the interaction rule generation unit 143 extracts a node from the interaction rule stored in the interaction rule DB 134, the user interest category is assigned to the extracted node. However, the present invention is not limited to this. For example, a lower node of a node to which the user interest category is attached may be extracted, and the message and category of the lower node may be changed.

  This case will be described with reference to FIGS. For example, when the user interest category is “Jimbocho”, the dialogue rule generation unit 143 specifies “region” as a higher category of this “Jimbocho” and specifies “Roppongi” as a lower category of this “region”. To do. Then, the dialogue rule generation unit 143 extracts (duplicates) a plurality of nodes (nodes with node IDs “N4”, “N6”, and “N7”) for nodes below the node to which “Roppongi” is attached. To do. Subsequently, the dialogue rule generation unit 143 assigns the category assigned to the node having the user interest category “Roppongi” (the node having the node ID “N4”) among the extracted nodes in the plurality of layers to the user interest category “ Change to "Jimbocho". Here, messages of nodes lower than this node (nodes with node ID “N4”) (nodes with node IDs “N6” and “N7”) are sent to external gourmet sites, etc. “Curry” and “Jimbocho” To change to a message including the store name corresponding to the user interest category “Curry: Jimbocho”.

  The dialogue rule changing unit 144 is based on the user interest category determined based on the user's behavior history on the web and the category attached to the message corresponding to each node (detection node and action node). The dialogue rule stored in the DB 134 is changed.

  Specifically, the dialog rule changing unit 144 adds a new dialog rule generated by the dialog rule generating unit 143 to the dialog rule stored in the dialog rule DB 134. Here, the dialogue rule changing unit 144 refers to the category DB 133 and specifies a higher category of the user interest category. Then, the dialog rule changing unit 144 sets the new dialog rule so that the new dialog rule generated by the dialog rule generating unit 143 is lower than the detected node or action node to which the specified higher category is attached. Add

  Here, the dialog rule changing unit 144 generates a node ID of the generated new dialog rule so that it does not overlap with the node ID stored in the node table 135, and also generates an edge corresponding to the generated node ID. Generate. In addition, when the highest node of the generated new dialog rule is the detection node, the dialog rule changing unit 144 sets the operation node with the identified higher category as the connection source node, and sets the highest level of the new dialog rule. An edge having a higher node as a connection destination node is generated. In addition, when the highest node of the generated new dialog rule is an action node, the dialog rule changing unit 144 sets the detected node with the identified higher category as the connection source node, and sets the highest level of the new dialog rule. An upper node generates an edge as a connection destination node. Then, the dialog rule changing unit 144 stores the generated new dialog rule and the generated edge in the dialog rule DB 134 (the node table 135 and the edge table 136).

[flowchart]
Next, the flow of processing in the dialogue rule changing device 1 will be described.
FIG. 8 is a flowchart showing a flow of processing for outputting a response message in accordance with a predetermined dialogue rule for the input message by the dialogue rule changing apparatus 1 according to the first embodiment.

  In step S1, the control unit 14 (aggregation unit 141) aggregates the user's behavior history for each category. Specifically, the control unit 14 (aggregation unit 141) performs a morphological analysis of the action history stored in the action history DB 131 to extract words, and the extracted words are stored in the category dictionary DB 132. Classify into: Then, the control unit 14 (aggregation unit 141) aggregates the action history for each category by totaling the classified words for each category. The tabulation process in step S1 is performed for each of a plurality of user action histories.

  In step S2, the control unit 14 (interest category determination unit 142) determines, for each of a plurality of users, one or a plurality of categories having a relatively high aggregate value of the categories aggregated in step S1 as user interest categories. .

  In step S3, the control unit 14 (interaction rule generation unit 143) creates, for each of a plurality of users, a new interaction rule including at least one of a detection node and an action node to which the user interest category determined in step S2 is attached. Generate. Specifically, the control unit 14 (interaction rule generation unit 143) extracts a node connected to a lower level of a node to which a higher category of the user interest category in the interaction rule stored in the interaction rule DB 134 is attached, For at least one of the extracted nodes, the category attached to the node is changed to the user interest category, and the message corresponding to the node is changed to a message corresponding to both the user interest category and the higher-level category. To generate a new interaction rule. The process of step S3 is performed for all user interest categories determined in step S2.

  In step S4, the control unit 14 (interaction rule changing unit 144) changes the interaction rule stored in the interaction rule DB 134. Specifically, the control unit 14 (dialog rule changing unit 144) refers to the category DB 133 to identify a higher category of the user interest category, and to the detection node or action node to which the identified higher category is attached. On the other hand, a new interaction rule is added so that the new interaction rule generated in step S3 is at a lower level. The process in step S4 is performed for all the dialogue rules generated in step S3.

  As described above, according to the first embodiment, the interaction rule changing device 1 includes the user interest category determined by the interaction rule changing unit 144 based on the user's behavior history on the web, and the message category corresponding to each node. Based on the above, change the interaction rule.

  Thus, since the interaction rule changing device 1 changes the interaction rule according to the user interest category based on the user's behavior history, it is not necessary to accept an operation for changing the interaction rule from the user. Furthermore, since the user interest category based on the user's behavior history is in accordance with the user's preference, the dialogue rule changing apparatus 1 does not place a burden on the user, and the dialogue agent system in accordance with the user's preference. Two interaction rules can be changed.

  In addition, the dialog rule changing device 1 generates a new dialog rule including at least one of the detection node and the action node to which the user interest category is attached by the dialog rule generating unit 143, and the dialog rule changing unit 144 generates the dialog rule A new dialogue rule generated for the dialogue rule stored in the DB 134 is added. By doing in this way, the dialog rule change apparatus 1 can add the dialog rule corresponding to a user interest category newly, and can increase the dialog rule corresponding to a user preference.

  Further, the dialog rule changing device 1 extracts and extracts, by the dialog rule generation unit 143, nodes connected to the lower level of the node to which the higher category of the user interest category in the dialog rules stored in the dialog rule DB 134 is attached. New interaction rule by changing the category attached to the selected node to the user interest category and changing the message corresponding to the extracted node to the message corresponding to both the user interest category and the upper category. Is generated. By doing in this way, the interaction rule change apparatus 1 can generate | occur | produce the node to which the user interest category was attached | subjected efficiently using the structure of the existing node.

  In addition, the dialog rule changing device 1 generates a dialog rule for a detection node or action node to which a category higher than the user interest category in the dialog rule stored in the dialog rule DB 134 is attached by the dialog rule changing unit 144. The new dialogue rule generated by the unit 143 is added so as to be subordinate. By doing in this way, the interaction rule changing device 1 can make a transition from the detection node or the action node with the higher category of the user interest category to the interaction corresponding to the user interest category without a sense of incongruity.

  Further, the dialogue rule changing device 1 changes one dialogue rule stored in the dialogue rule DB 134 based on each of a plurality of user action histories. By doing in this way, the dialogue rule change device 1 can efficiently increase the scale of the dialogue rule.

  Further, in the conversation rule changing device 1, the interest category determining unit 142 determines the user interest category based on the behavior history stored in the behavior history DB 131, and the conversation rule changing unit 144 uses the interest category determining unit 142 to determine the user interest category. The interaction rules can be changed based on the determined user interest category and the message category corresponding to each node.

[Second Embodiment]
[Function configuration]
FIG. 9 is a diagram showing a functional configuration of the dialog rule changing device 1A according to the second embodiment of the present invention.
The dialog rule changing device 1A shown in the present embodiment has a plurality of dialog rules corresponding to each of a plurality of users and a transition destination (connection destination) associated with the edge of the dialog rule. Unlike the first embodiment, the point that the transition probability to the operation node of () is stored in association with each other and the point that the conversation rule can be individually changed for each of the plurality of interaction rules is different from the first embodiment. This is the same as the dialogue rule changing device 1 shown in the embodiment.

  The storage unit 13A of the interaction rule changing device 1A includes an interaction rule DB 134A, and the control unit 14A of the interaction rule changing device 1A includes an interaction rule changing unit 144A.

  The dialogue rule DB 134A stores a plurality of dialogue rules, and each of the plurality of dialogue rules corresponds to a plurality of users.

FIG. 10 is a diagram showing the dialogue rule DB 134A according to the second embodiment.
As illustrated in FIG. 10, the interaction rule DB 134A stores a node table 135A that stores information related to nodes and an edge table 136A that stores information related to edges.

The node table 135A stores a user ID, a node ID, a node type, a processing content, and a category in association with each other.
The edge table 136A stores a user ID, a connection source node ID, a connection destination node ID, and a transition probability from the connection source node ID to the connection destination node ID in association with each other. That is, the edge table 136A stores the transition probability to the operation node of the transition destination (connection destination) associated with the edge in association with the edge.

  In the second embodiment, one node table 135A and one edge table 136A are stored in one dialog rule DB 134A. However, a plurality of dialog rule DBs 134A are provided or one dialog rule DB 134A is stored. Thus, a plurality of node tables 135A and edge tables 136A may be stored, or a plurality of node tables 135A and edge tables 136A may be stored individually in the user terminal 4 so as to correspond to each of a plurality of users. Then, the dialog rule generated by the dialog rule generation unit 143 may be added only to the dialog rule corresponding to the user so that only one user can use it.

FIG. 11 is a schematic diagram of a dialogue rule according to the second embodiment.
As shown in FIG. 11, the dialog rule according to the second embodiment is different in that a transition probability is added to the dialog rule according to the first embodiment. That is, the numerical value given to each edge is a transition probability.

  The dialog rule changing unit 144A has a function of changing the dialog rule corresponding to the user ID associated with the user interest category in addition to the functions described in the first embodiment.

  In other words, the dialog rule changing unit 144A is indirectly connected to the operation node to which the user interest category in the dialog rule stored in the dialog rule DB 134A is attached and the higher level of the operation node via a plurality of edges. The edge that connects the detected node is added to the interaction rule stored in the interaction rule DB 134A.

  Specifically, the dialogue rule changing unit 144A refers to the category DB 133 and identifies a category that is one or more layers above the user interest category. Then, the dialogue rule changing unit 144A specifies a detection node that is indirectly connected to the node to which the user interest category is attached among the detection nodes to which any one of the identified higher categories is attached. . Then, the dialogue rule changing unit 144A generates an edge connecting the node to which the user interest category is attached and the identified detection node, and corresponds to the generated edge and the user interest category. By storing the user ID in the dialogue rule DB 134A, the newly generated edge is added to the dialogue rule. In addition, after the newly generated edge is added, the dialogue rule changing unit 144A changes the transition probability so that the transition probabilities of the lower edges associated with the connection source node of this edge become equal. Here, when there are a plurality of operation nodes to which the same user interest category is attached, the dialogue rule changing unit 144A selects one operation node at random and performs processing for generating an edge.

  A specific example will be described with reference to FIGS. 5, 10, and 11. For example, in the dialogue rule shown in FIG. 11, when the user ID corresponding to the user interest category is “u001” and the user interest category is “curry store: Roppongi”, the dialogue rule changing unit 144A sets “ Identify curry. Then, the dialog rule changing unit 144A is a detection node to which the “curry” category is attached, and the nodes to which the user interest category “curry store: Roppongi” is attached (node IDs “N6” and “N7”). The node having the node ID “N1” is specified as the detection node indirectly connected to the node. Here, since there are a plurality of operation nodes to which the same user interest category is attached, the dialogue rule changing unit 144A selects the node having the node ID “N6” as one operation node. Then, the dialogue rule changing unit 144A includes a node (node ID “N6”) to which the user interest category “curry store: Roppongi” is attached, and the identified detection node (node ID “N1”). Are generated, and the generated edge and the user ID “u001” are associated with each other and stored in the interaction rule DB 134A, whereby the newly generated edge is added to the interaction rule. In this case, the dialogue rule in which the node is newly added is as shown in FIG.

  The interaction rule changing unit 144A is an operation node that is indirectly connected to a higher level through a plurality of edges with respect to an operation node to which a user interest category in the interaction rule stored in the interaction rule DB 134A is attached. Is changed to a response message including the name of the user interest category.

  Specifically, the dialogue rule changing unit 144A refers to the category DB 133 and identifies a category that is one or more layers above the user interest category. Then, the dialogue rule changing unit 144A specifies an operation node to which any one of the specified higher categories is attached. The identified node is an operation node that is indirectly connected to the user interest category via a plurality of edges. Then, the dialogue rule changing unit 144A changes the response message corresponding to the identified action node to a response message including the name of the user interest category. In addition, including the name of a user interest category means including all or part of the name of a user interest category.

  A specific example will be described with reference to FIGS. 5, 10, and 11. For example, when the user ID corresponding to the user interest category is “u001” and the user interest category is “curry: keema”, the dialogue rule changing unit 144A specifies “curry: type” as the upper category. Subsequently, the dialogue rule changing unit 144A identifies the operation node to which the “curry: type” category is attached (the node having the node ID “N8”) as the operation node to which any one of the upper categories is assigned. . Subsequently, as shown in FIG. 13, the dialog rule changing unit 144A uses the response message of this action node as a response message including the name of the user interest category (for example, “What kind of curry do you like? I like Keema curry”) Change to

  Further, the dialogue rule changing unit 144A, when the plurality of edges corresponding to the detection node to which the higher category of the user interest category is added to the dialogue rule stored in the dialogue rule DB 134A, When the edge corresponding to the operation node with the category is included in the plurality of edges, the transition probability of the edge corresponding to the operation node with the category indicating the user interest category becomes the transition probability of the other edge. The transition probabilities of each of the plurality of edges are adjusted so as to be higher than each other.

  A specific example will be described with reference to FIGS. 5, 10, and 11. For example, when the user ID corresponding to the user interest category is “u001” and the user interest category is “Roppongi”, the dialogue rule changing unit 144A specifies “curry” as the upper category. Subsequently, the dialogue rule changing unit 144A determines the user interest category “Roppongi” for the transition probabilities of a plurality of edges having the detection node (the node with the node ID “N1”) to which the “curry” category is attached as the connection source node. The transition probability of an edge having an operation node (node ID “N3”) to which the category indicating “” is attached as a connection destination node is higher than the transition probability of an edge having another operation node as a connection destination node. Adjust to. For example, as illustrated in FIG. 14, the dialogue rule changing unit 144A sets the transition probability of an edge whose connection source node is a node with a node ID “N1” and whose connection destination node is a node with a node ID “N3” to “0”. .8 ”, the transition probability of the edge where the connection source node is the node with the node ID“ N1 ”and the connection destination node is the node with the node ID“ N2 ”is adjusted to“ 0.2 ”.

  In addition, the dialogue rule changing unit 144A is an operation node having the same detection node as a connection source node when a plurality of user interest categories are determined for one user, and each user interest category is different. When there is an operation node marked with, the transition probabilities of these operation nodes with different user interest categories are adjusted. In other words, the dialog rule changing unit 144A adjusts the transition probability so that the transition probability of the user interest category having a large total value aggregated by the totalization unit 141 is relatively high.

  According to the second embodiment, in addition to the effects described in the first embodiment, there are the following effects.

  The interaction rule changing device 1A is indirectly connected to the operation node to which the user interest category in the interaction rule stored in the interaction rule DB 134A is assigned by the interaction rule changing unit 144A, and the operation node via the plurality of edges. An edge that connects a detection node that is automatically connected is added to the interaction rule stored in the interaction rule DB 134A. By doing in this way, the dialog rule changing device 1A can make a transition to a dialog corresponding to the user's taste with a small number of dialogs.

  Further, the dialog rule changing device 1A is indirectly connected to a higher level through a plurality of edges with respect to the operation node to which the user interest category in the dialog rule stored in the dialog rule DB 134A is attached by the dialog rule changing unit 144A. The response message corresponding to the operation node connected to is changed to a response message including the name of the user interest category. Here, when a response message including the name of the user interest category is output, the user is more likely to input a message corresponding to the user interest category than when the response message is not output. Therefore, the interaction rule changing device 1A can increase the probability of transition to an operation node corresponding to the user interest category.

  Further, in the case where a plurality of edges corresponding to the detection node to which the higher category of the user interest category is stored, the interaction rule changing device 1A has an edge corresponding to the operation node to which the user interest category is attached. When included in the multiple edges, adjust the transition probabilities of each of the multiple edges so that the transition probabilities of the edges corresponding to the action nodes with user interest categories are higher than the transition probabilities of other edges To do. By doing in this way, the dialog rule changing device 1A increases the probability of transition to the operation node to which the user interest category is assigned, and the dialog corresponding to the user's preference between the user and the dialog agent system 2. Can be made easier.

  In the present embodiment (first embodiment and second embodiment), the dialog rule changing devices 1 and 1A, the dialog agent system 2, the web server 3, and the user terminal 4 are applied to a computer and its peripheral devices. . Each unit in the present embodiment is configured by hardware included in a computer and its peripheral devices, and software that controls the hardware.

  The hardware includes a storage unit, a communication device, a display device, an input device, and an output device in addition to a CPU (Central Processing Unit) as a control unit. Examples of the storage unit include a memory (RAM: Random Access Memory, ROM: Read Only Memory, etc.), a hard disk drive (HDD: Hard Disk Drive), and an optical disk (CD: Compact Disc, DVD: Digital Versatile Drive, etc.). It is done. Examples of the communication device include various wired and wireless interface devices. Examples of the display device include various displays such as a liquid crystal display and a plasma display. Examples of the input device include a keyboard, a pointing device (mouse, tracking ball, touch pad, etc.), a position acquisition device using GPS (Global Positioning System), a timer, a microphone, and the like. Examples of the output device include a speech synthesizer for outputting speech.

  The software includes a computer program and data for controlling the hardware. The computer program and data are stored in the storage unit, and are appropriately executed and referenced by the control unit. The computer program and data can be distributed via a communication line, or can be recorded on a computer-readable medium such as a CD-ROM and distributed.

  The above embodiment has been described above. However, the above embodiment is merely an example, and components, data structures, flowcharts, and the like illustrated in the drawings can be changed as appropriate. In addition, the effects of the present invention described in the above embodiment are merely enumerated the most preferable effects resulting from the above embodiment, and are not limited to those described in the above embodiment.

  In the above embodiment, the totalization unit 141 classifies the words included in the action history based on the category dictionary DB 132, but the classification may be performed using a document classifier.

  In the above embodiment, the dialog rule is stored in the dialog rule changing device 1 or 1A. However, the present invention is not limited to this, and the dialog rule may be stored in another system or server such as the dialog agent system 2. . In this case, the interaction rule changing device 1 or 1A may change the interaction rule stored in another system or server.

  In the above embodiment (second embodiment), the transition probability for the operation node is set. However, the transition probability may be set for each time zone and region. That is, the transition probabilities stored in the edge table 136A are stored for each time zone (for example, daytime, evening, night, etc.), and the dialogue agent system 2 applies the dialogue rule according to the time zone. The probability may be changed. For example, when it is a time zone of midnight or early morning, the dialogue agent system 2 may output a message related to a recipe without outputting a message related to restaurant recommendation.

  Further, the transition probability stored in the edge table 136A is stored for each region, and the dialog agent system 2 acquires the position from the user terminal 4, and changes the transition probability applied to the dialog rule based on the position. You may do it. Alternatively, the edge table 136A may store transition probabilities for each time zone and each region, and the dialog agent system 2 may appropriately change these transition probabilities based on the time zone and the position of the user terminal. For example, when the user terminal 4 is located in Roppongi, the dialogue agent system 2 may output a message related to Roppongi without outputting a message related to Jimbocho.

DESCRIPTION OF SYMBOLS 1 Dialog rule change apparatus 11 Operation part 12 Display part 13, 13A Storage part 14, 14A Control part 131 Action history DB
132 Category Dictionary DB
133 Category DB
134, 134A Dialog rule DB
141 Aggregation unit 142 Interest category determination unit 143 Dialog rule generation unit 144, 144A Dialog rule change unit

Claims (10)

An interaction rule changing device for changing the interaction rule of an interaction agent system that outputs a response message according to an interaction message set in advance with respect to an input message,
It is composed of a detection node in which a processing procedure corresponding to the input message is defined, an operation node in which a processing procedure corresponding to the response message is defined, and an edge indicating a connection relationship between the detection node and the operation node. Dialogue rule storage means for storing, as the dialogue rule, data with a message category attached to each node,
An interaction rule changing means for changing the interaction rule based on a user interest category determined based on a user's behavior history on the web and a message category corresponding to each node;
A dialogue rule changing device comprising: Interaction rule generation means for generating a new interaction rule including at least one of the detection node and the action node to which the user interest category is attached,
The interaction rule changing unit adds a new interaction rule generated by the interaction rule generation unit to the interaction rule stored in the interaction rule storage unit.
The dialog rule changing device according to claim 1. The category has a hierarchy,
The dialog rule changing means is generated by the dialog rule generating means for a detection node or action node to which a category higher than the user interest category in the dialog rule stored in the dialog rule storage means is attached. Add new conversation rules to the lower level,
The dialog rule changing device according to claim 2. The dialogue rule generation means extracts a node connected to a lower level of a node to which a higher category of the user interest category in the dialogue rule stored in the dialogue rule storage means is added, and attaches the extracted node to the extracted node. A new interaction rule by changing a message corresponding to the extracted node to a message corresponding to both the user interest category and the higher-level category. Generate,
The dialog rule changing device according to claim 3. The interaction rule changing unit is indirectly connected to an operation node to which the user interest category in the interaction rule stored in the interaction rule storage unit is attached, and a higher level of the operation node via a plurality of edges. Adding an edge connecting to the detected detection node to the interaction rule stored in the interaction rule storage means;
The dialog rule changing device according to claim 1. The dialogue rule changing means is indirectly connected to a higher level through a plurality of edges with respect to an operation node to which the user interest category in the dialogue rule stored in the dialogue rule storage means is attached. Changing the response message corresponding to the operation node to a response message including the name of the user interest category;
The dialog rule changing device according to claim 1. The dialogue rule storage means stores the transition probability to the transition destination operation node associated with the edge in association with the edge,
In the case where a plurality of edges corresponding to a detection node to which a higher category of the user interest category is added to the interaction rule stored in the interaction rule storage unit, the interaction rule changing unit is stored. When an edge corresponding to the operation node to which the user interest category is attached is included in the plurality of edges, the transition probability of the edge corresponding to the operation node to which the user interest category is attached is the transition probability of another edge. Adjusting the transition probability of each of the plurality of edges to be higher than
The dialogue rule changing device according to any one of claims 1 to 6. An action history storage means for storing an action history on the user's web;
Interest category determination means for determining the user interest category based on the action history stored in the action history storage means,
The interaction rule changing unit changes the interaction rule based on a user interest category determined by the interest category determining unit and a message category corresponding to each node.
The dialogue rule changing device according to any one of claims 1 to 7. A dialog rule changing method in which a computer changes the dialog rule of a dialog agent system that outputs a response message in accordance with a predetermined dialog rule for an input message,
It is composed of a detection node in which a processing procedure corresponding to the input message is defined, an operation node in which a processing procedure corresponding to the response message is defined, and an edge indicating a connection relationship between the detection node and the operation node. A dialogue rule that is a data to which the category of the message is attached to each node.
Change based on the user interest category determined based on the user's behavior history on the web and the message category corresponding to each node;
How to change dialogue rules. A dialog rule changing program for causing a computer to change the dialog rule of the dialog agent system that outputs a response message in accordance with a predetermined dialog rule for an input message,
It is composed of a detection node in which a processing procedure corresponding to the input message is defined, an operation node in which a processing procedure corresponding to the response message is defined, and an edge indicating a connection relationship between the detection node and the operation node. And an interaction rule that is data with a message category attached to each node.
Changing based on a user interest category determined based on a user's behavior history on the web and a message category corresponding to each of the nodes;
An interaction rule changing program for causing the computer to execute

Images