JP4489552B2 - System and terminal arrangement method - Google Patents

Description

  The present invention relates to an interaction control system, a terminal device, and an interaction control method for performing man-machine interaction for acquiring user information through dialogue.

  In order to realize a personalized service that provides information and services suitable for users, it is necessary to acquire user situations (situations), preferences, and attributes (profiles).

  For example, a method of automatically acquiring user situations, preferences, and attributes from history information without making the user aware of it is a burdenless method for the user, but the information acquired by this method is There is no subjective judgment. For this reason, there is a possibility that the information automatically acquired from the history information does not reflect the real intention of the user. Profiles cannot be acquired from historical information.

  Here, information that can be acquired by the user's explicit input of information is targeted.

  As means for acquiring user preferences and attributes through the input of explicit information, methods such as questionnaires and user registration are common. There are also many methods of confirming with the user whether the automatically acquired information is correct and establishing the information based on the user's judgment. In these methods that require user involvement, it is important to reduce the burden on the user and increase the opportunity to acquire more information.

  In addition, as a method for reducing the burden on the user's interaction, there are a method for reducing a physical (physical) burden and a method for reducing a psychological burden.

  The method of reducing the physical (physical) burden is a method of reducing the effort of the user. Many of these methods employ a method of reducing the number of keystrokes required for input. For example, predictive conversion (see, for example, Non-Patent Document 1) or a method of reducing input time by presenting input candidates to be selected by the user is employed. In many cases, a method similar to a means for users to communicate with each other is introduced by a method such as voice input.

  On the other hand, in the method of reducing the psychological burden, it is necessary to ask the appropriate content with an appropriate expression at an appropriate timing.

  In human interaction, it is important to identify and start the situation of the interaction partner. If a user imposes an extra task in situations where it is difficult to input or there is no psychological margin for input, not only can the input of the necessary information be expected, but the user is also disgusted with the request for input. However, no input may be made for subsequent inquiries. For this reason, for example, a user's condition is estimated by the environment side and the sensor installed in the device. In this case, the timing for starting the interaction is triggered by a change in the value measured by the sensor.

  Many studies on timing control of Push-type information presentation use position information. There are various levels of position information. For example, it is detected that the user has entered a room, and information about the room is transmitted by e-mail. Further, when the user's face direction is tracked and it is determined that the user is looking at the same place for a certain period of time, information related to the display of the same place is displayed on the portable terminal (for example, see Non-Patent Document 2). .

  In addition, research on controlling the presentation timing of peripheral information in consideration of the task execution state of the user has been conducted. In this research, assuming a document editing task, the task execution state of the user is determined from the number of keyboard hits per hour, and information is presented when the number of key hits is small (for example, see Non-Patent Document 3).

  With regard to the content of information, when a user suddenly displays or asks about a completely irrelevant matter, the user not only feels uncomfortable and suspicious, but also feels annoyed. In other words, it is necessary to listen to the content that the user is interested in and the content that was there. The Office (registered trademark) assistant function of Microsoft (registered trademark) presents necessary help (information) to the user from the user's function use and operation history.

  Also, the secretary agent linked with the mailer or scheduler asks the user what the keyword is currently about. Since the contents to be asked are matters directly related to the input contents such as tasks and operations, the user's interest is considered to be suitable for them, and the psychological burden is small.

Further, regarding the display, it is necessary not to hinder the user's task, but to be easily understood and operated by the user. For example, there is an approach for designing an interface based on ergonomic usability such as button arrangement considering the movement of the line of sight, display size and color scheme that are likely to attract attention (see, for example, Non-Patent Document 4).
Toshiyuki Masui, “High-speed text input method using a pen, Jiro Tanaka (ed.), Interactive System and Software IV: Japan Society for Software Science and Technology WISS'96, pp. 51-60. Yaya, Ishikawa et al., “Pi_book: Exhibition support tool in museum”, Interaction 2003; pp. 51-52, 2003 Ishida, Kuramoto, Shibuya, Kanno: Presentation method of surrounding information that dynamically changes the presentation timing according to the user's condition, HI Symposium 2003, pp. 101-104 (2003) Jef Raskin, “The Humane Interface: New Directions for Designing Interactive Systems”, Addison-Wesley, 29 March 2000 Shun Ishizaki, “System Model of Language and Memory”, Computer Today, Vol. 103, pp. 24-30, 2001.5

  However, the background art described above has the following problems.

  In the background art described above, it is possible to prompt the user to input matters related to the task that are considered to be of interest to the user without disturbing the execution of the user's task. However, the information necessary to perform a personalization service is not limited to matters directly related to the task. Therefore, there is a problem that information necessary for performing a personalization service including matters not directly related to the task cannot be acquired with less psychological burden on the user.

  Therefore, an object of the present invention is to provide an interaction control system, a terminal device, and an interaction control method capable of acquiring information necessary for performing a personalization service.

This system is,
A system having a sensor and a terminal device that acquires information detected by the sensor ,
A situation information acquisition part for acquiring information acquired by the sensor as information of Situation down,
Based on the past history of responses of each information of the acquired situation by the situation information acquisition part, a response possible parameter calculation unit for relative interaction User chromatography The calculated response likelihood parameter indicating the possibility of responding ,
An interaction start determination unit that determines whether or not to perform an interaction based on the response possibility parameter ;
A topic extraction unit that extracts information related to a task from the content of an application used by a user as a topic ;
An item database that stores items to be acquired with respect to user information, a question for acquiring the item, and a question accompanying the question;
A keyword extraction unit that extracts a keyword included in the attribute data of the topic extracted by the topic extraction unit from the concept dictionary as a next topic;
A target setting unit for setting items of user information to be acquired from the items to be acquired stored in the item database ;
Using the keywords extracted by the keyword extracting unit, wherein the extracted topic by topic extraction unit, the route searching unit that creates a route causing the transfer Qian up item of the user information acquiring set by the target setting unit When,
Based on the question stored in the item database, a question regarding a keyword included in the route created by the route search unit is determined, and a question execution unit for asking a question;
Analyzing the response to the question, and a response analyzer unit for acquiring the user information.

  By configuring in this way, it is possible to ask the user a question regarding the user information to be acquired by using a related keyword from a topic directly related to the current task.

The end the end device,
A terminal device that acquires user information based on information detected by a sensor ,
A situation information acquisition unit that acquires information acquired by the sensor as situation information;
A topic extraction unit that extracts information related to a task from the content of an application used by a user as a topic;
An item database that stores items to be acquired with respect to user information, a question for acquiring the item, and a question accompanying the question;
A keyword extraction unit that extracts a keyword included in the attribute data of the topic extracted by the topic extraction unit from the concept dictionary as a next topic;
A target setting unit for setting items of user information to be acquired from the items to be acquired stored in the item database ;
Using the keywords extracted by the keyword extracting unit, wherein the extracted topic by topic extraction unit, the route searching unit that creates a route causing the transfer Qian up item of the user information acquiring set by the target setting unit When,
Based on the question stored in the item database, a question regarding a keyword included in the route created by the route search unit is determined, and a question execution unit for asking a question ;
A response analysis unit that analyzes a response to the question and acquires the user information.

  By configuring in this way, it is possible to ask the user a question regarding the information of the user to be acquired from the topic directly related to the current task via the related topic.

This how is,
A method in a terminal device for acquiring user information based on information detected by a sensor ,
Situation information acquisition step for acquiring information acquired by the sensor as situation information;
A topic extraction step of extracting information related to the task from the content of the application used by the user as a topic;
A keyword extraction step of extracting a keyword included in the attribute data of the topic extracted by the topic extraction step from the concept dictionary as a next topic;
Regarding user information, an item of user information acquired from the item to be acquired stored in an item database that stores an item to be acquired, a question for acquiring the item, and a question accompanying the question is set. A goal setting step to
A route search step for creating a route for transitioning from the topic extracted by the topic extraction step to the item of user information to be acquired set by the goal setting step using the keyword extracted by the keyword extraction step; ,
A question execution step of determining a question regarding a keyword included in the route created by the route search step based on the question stored in the item database, and asking a question;
A response analysis step of analyzing a response to the question and acquiring the user information .

  By doing in this way, it is possible to ask the user a question regarding the user information to be acquired by passing through the related topic from the topic directly related to the current task.

  According to the embodiment of the present invention, it is possible to realize an interaction control system, a terminal device, and an interaction control method capable of acquiring information necessary for performing a personalization service.

Next, embodiments of the present invention will be described with reference to the drawings.
In all the drawings for explaining the embodiments, the same reference numerals are used for those having the same function, and repeated explanation is omitted.

  An interaction control system according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2.

  The interaction control system according to the present embodiment includes a terminal device 100 and a sensor 200. The terminal device 100 includes, for example, a PC and a server, and includes a Web browser, a mailer, a media player, an interaction control device 110 described later, and the like, and performs various task processes. The sensor 200 is configured by a sensor arranged on the environment side. The terminal device 100 and the sensor 200 are connected by a communication network 50. Further, a sensor may be arranged on the terminal device 100.

  As shown in FIG. 2, the interaction control device 110 includes a situation acquisition unit 120, a history management unit 140 connected to the situation acquisition unit 120, a topic extraction unit 130 connected to the history management unit 140, and an interaction control unit 150. And a topic transition control unit 160, a response analysis unit 190, and a user information DB 180 connected to the response analysis unit 190.

  In addition, the situation acquisition unit 120 includes a status acquisition unit 122 to which information from sensors provided on the environment side or the device is input. The topic extraction unit 130 also includes a text appearance frequency calculation unit 132 and a metadata item importance calculation unit 134 to which information from sensors provided on the environment side or device is input. The history management unit 140 is connected to the situation / topic storage unit 142 connected to the situation acquisition unit 122, the text appearance frequency calculation unit 132, and the metadata item importance calculation unit 134, and to the situation / topic storage unit 142. A user response history parameter calculation unit 144 as a response possibility parameter calculation unit.

  The interaction control unit 150 includes a situation match determination unit 152 connected to the situation / topic storage unit 142, an interaction start determination unit 154 connected to the situation match determination unit 152, an interaction start determination unit 154, and a situation topic. And a related topic extraction unit 156 connected to the storage unit 142.

  The topic transition control unit 160 includes a target setting unit 161 as user information setting means connected to the situation / topic storage unit 142 and the interaction start determination unit 154, a route creation unit 164 connected to the target setting unit 161, and The acquisition item DB 162, an optimum route selection unit 168 connected to the route creation unit 164, and a question phrase determination unit 163 as question execution means connected to the acquisition item DB 162 and the optimum route selection unit 168 are provided. The route creation unit 164 includes a concept topic extraction unit 165 as a keyword extraction unit connected to the target setting unit 161, a concept dictionary 166 and a route search unit 167 as attribute data storage units connected to the concept topic extraction unit 165, Is provided. The optimum route selection unit 168 includes a node number calculation unit 169 and a weight calculation unit 170 connected to the route search unit 167, a route evaluation calculation unit 171 connected to the node number calculation unit 169 and the weight calculation unit 170, and a route evaluation An optimal route determination unit 172 connected to the calculation unit 171 and the question phrase determination unit 163 is provided. In addition, the weight calculation unit 170 is connected to the related topic extraction unit 156.

  The interaction control apparatus 110 according to the present embodiment interacts with the user via the interface, and acquires user information through the interaction. In this case, the interaction control device 110 reduces the psychological burden on the user and acquires information necessary for performing the personalization service from information related to the current task to information to be acquired on the spot. Get the information you want to get with as little interaction as possible. Specifically, the interaction control device 110 starts the interaction in consideration of the user situation (situation) in order to realize the interaction for acquiring the user information, and regarding the difference between the current topic and the information to be acquired, Ask questions about what the user is interested in.

  In this way, in order to acquire information that is not directly related to the current task, the situation (situation) is changed from the topic directly related to the current task to the topic that appeared in the previous similar situation. In consideration, the user can be asked about contents that are not different from the user's interest.

  The situation acquisition unit 122 acquires information on the situation where the user is currently placed. Specifically, situation information is acquired using the sensor 200. Further, the terminal device 100 may include a sensor, and situation information may be acquired from the sensor. In addition, the situation acquisition unit 122 stores the acquired situation information in the situation / topic storage unit 142. Situation information is composed of, for example, five items (parameters) of location, time, application used, peripheral object / object, and day of the week.

  The acquisition method of each item and the specific contents of the acquired situation information will be described.

  The location information is acquired by placing a tag having the location ID in an environment, for example, a room, and reading the information of the tag by a sensor 200 configured by a reader. In this case, the level of information indicating the location includes living room, kitchen, meeting room, train, and the like. Further, the information level may be information on a chair that is sitting among chairs arranged indoors as a narrow range, or regional information such as a municipality as a wide range. Moreover, it is good also as position information detected by GPS (Global positioning system).

  The time information is time information by a clock built in the system. Further, for example, it may be divided at a predetermined time interval, for example, a unit of about 2 hours, given meaning, and semantic information corresponding to the divided time range may be acquired as time information. In this case, for example, early morning is 4:00 to 6:00, morning is 6:00 to 8:00, morning is 8:00 to 12:00, noon is 12:00 to 13:00, and afternoon is 13:00. ~ 16: 00, evenings from 16:00 to 18:00, nights from 18:00 to 23:00, and midnight from 23:00 to 4:00, and meaning information corresponding to time information is acquired. . Further, time information may be acquired by dividing into morning and afternoon, or information of day of the week + time may be acquired as time information.

  The information on the application used is obtained from the ID of the application running on the terminal. For example, information on applications such as mailers, browsers, schedulers, media players, and the like is acquired.

  The peripheral object / object information is obtained by embedding a tag having information on the object in an object arranged in the vicinity, and the sensor 200 configured by the reader reads the tag information. For a subject, for example, a person in the vicinity, information on whether or not a person whose name is in the address book exists in the vicinity is acquired. Specifically, a tag having ID information is mounted on the terminal device 100 possessed by each user, and a sensor configured by a reader acquires the information by reading the information of the tag. In addition, regarding mail, information about a transmission partner and a reception partner is acquired.

  The day information is obtained from a calendar built in the terminal device 100.

  The situation information described above is acquired when the situation changes. For example, you may make it re-acquire when one item changes among items constituting situation information, or decide the number of items that triggers re-acquisition of the situation in advance and exceed the number of items If the situation changes, it may be reacquired. Alternatively, a cycle for acquiring the situation may be determined in advance, and acquired according to the cycle.

  The text appearance frequency calculation unit 132 is related to the task being processed from the content of the application used by the user, for example, a web page in the case of web browsing, or a mail text in the case of mail. A possible text is extracted, its appearance frequency is calculated, and a text of a predetermined rank is output as a topic TopicN from the higher frequency. Also, the text appearance frequency calculation unit 132 stores the topic TopicN in the situation / topic storage unit 142.

  For example, the text appearance frequency calculation unit 132 refers to the information on the Web page or mail transmitted from the sensor 200, extracts a character string, for example, a noun from the text included in the body, and ranks the text based on the appearance frequency. Do. The ranking is, for example, converting the number of appearances once into 1 point, and for the nouns included in the page or the title of the email, it directly indicates the content of the web page and email, so it is converted to 2 points, The number of points for each extracted noun is calculated. In this case, it is desirable that character strings having the same meaning, for example, “dock” and “dog” are unified in terms of notation and converted into points as the same. For example, the synonym concept of the concept dictionary is used, and the character strings indicating the same meaning are unified to be the same. Here, the concept dictionary is a dictionary for prescribing concepts referred to from the word dictionary, the bilingual dictionary, and the co-occurrence dictionary (see, for example, Non-Patent Document 5).

  The text appearance frequency calculation unit 132 calculates the ranking of each extracted noun based on the calculated number of points, and sets the noun of a predetermined rank as the topic TopicN in the user's current task. For example, the topic TopicN of the current task of the user is associated with the noun having the top 10 points.

  Note that the ranking method is not limited to the method described above, and other methods may be used.

  The metadata item importance calculation unit 134 refers to the content information of the application transmitted from the sensor 200, calculates the appearance frequency of a keyword that is considered to be related to the task being processed, and The keywords in a predetermined order from the highest one are output as the topic TopicN. For example, when listening to music or watching a moving image, keywords that are considered to be related to the task being processed are extracted from the media file being used. Further, the same processing may be performed with reference to information on application content input from a sensor included in the terminal device 100.

  For example, the appearance frequency of keywords indicating the characteristics of the media such as artist name, file name, company name, etc. described in the metadata is calculated and ranked, and keywords with a predetermined rank from the higher ranking are assigned as topic TopicN. To do. For example, the top 10 keywords with the highest appearance frequency are set as the topic TopicN. The metadata item importance calculation unit 134 stores the topic TopicN in the situation / topic storage unit 142.

  Note that the ranking method is not limited to the method described above, and other methods may be used.

  The user response history parameter calculation unit 144 calculates a response possibility parameter P_int (0 ≦ P_int ≦ 1) indicating a possibility that the user may respond to the dialogue. Since the possibility that the user will respond to the dialogue varies depending on the situation, the user response history parameter calculation unit 144 calculates a response possibility parameter based on the past performance. That is, the response possibility parameter is calculated based on whether or not the user has responded in the same situation in the past. For example, a response possibility parameter is set for a situation. That is, the response possibility parameter is set for a combination of items constituting situation information. For example, a default value is determined in advance, and when a user actually responds to a question in the situation, a predetermined value is added to the default value, and when no response is made, a predetermined value is determined from the default value. The value of is subtracted. For example, the default is set to 0.8, and 0.1 is added when the user actually responds to the question in the situation, and 0.1 is subtracted when the user does not respond.

  The method for calculating the response possibility parameter P_int is not limited to the method described above, and other methods may be used. For example, weighting may be performed according to the response time from the start of the dialogue until the user responds. In this case, the weight is increased when the response time is short.

  The situation / topic storage unit 142 stores the situation information acquired by the situation acquisition unit 122, the topic appearance N calculated by the text appearance frequency calculation unit 132 and the metadata item importance calculation unit 134, and the user response history parameter calculation unit 144. The response possibility parameter calculated in the above is stored. For example, the topic TopicN and the response possibility parameter are stored in association with the situation information.

  The situation match determination unit 152 determines whether or not the user task is blocked, and refers to the situation information stored in the situation topic storage unit 142 when it is determined that the user task is not blocked. Investigate whether the same situation as the current situation is stored.

  The interaction start determination unit 154 determines whether to start the interaction. For example, when the same situation does not exist in the situation match determination unit 152 and when the value of the response possibility parameter P_int in the same situation is a predetermined value or more, it is determined that the interaction is performed. For example, when the value of the response possibility parameter P_int in the same situation is 0.5 or more, it is determined that the interaction is performed.

  When it is determined that the interaction is to be started, the related topic extraction unit 156 refers to the situation topic storage unit 142 and extracts situation information similar to the current situation. For example, a situation in which a predetermined number of parameters match among the parameters (items) constituting the situation information is set as a similar situation. For example, a situation in which three parameters match among five parameters constituting situation information is set as a similar situation. In this case, when there are a plurality of similar situations, priorities are determined in advance for the five parameters constituting the information of the situation, and a situation in which parameters having a high priority coincide with each other is set as a similar situation. Good.

  Also, the related topic extraction unit 156 extracts a topic TopicN composed of text and keywords stored in association with the similar situation, calculates the number of appearances of the topic TopicN, and sets the weight Wn_weight.

  Note that the calculation method of the weight Wn_weight is not limited to the method described above, and other methods may be used. For example, weighting may be performed based on the time when a topic appears. In this case, the weight for the topic that appears recently is increased.

  As shown in FIG. 3, the acquisition item DB 162 records in advance items to be acquired, questions for acquiring the items, and questions associated with the questions regarding user information, for example, profiles and preferences.

  The target setting unit 161 determines the item TopicG of the user information to be acquired. For example, the target setting unit 161 selects, as Topic G, an item of user information that has not yet been acquired from the acquisition item DB 162. For example, information desired by the service information provider side or information close to the current topic is selected.

  The concept dictionary 166 is a dictionary for defining concepts referred to from the word dictionary, the bilingual dictionary, and the co-occurrence dictionary. Further, as the concept dictionary 166, an associative concept dictionary collected from a stimulus word as a superordinate concept, a subordinate concept, an environmental concept, a partial material concept, an attribute concept, a synonym and an action concept may be used (for example, non-patent literature) 5).

  The concept topic extraction unit 165 refers to the concept dictionary 166 and selects keywords included in attribute data such as material concept, synonym concept, and operation concept as well as the superordinate concept, subordinate concept, and peer concept of the current topic. Extract as the next topic. In this way, related topics can be extracted from many relationships. Moreover, since various topics can be extracted by using not only the parent-child relationship but also information from attribute data, it is possible to reduce the number of topics that are routed to the target information.

  The route search unit 167 performs a route search from TopicN to a place including TopicG as a keyword via various keywords randomly extracted from the keywords extracted by the concept topic extraction unit 165. In this way, by creating routes using various concepts, various routes from TopicN to TopicG can be searched. In this case, the limit item number that becomes the limit of the number of items to be routed may be determined in advance. For example, if the number of limit items is 20, when 20 or more items, that is, when Topic G cannot be reached even through 20 keywords from Topic N, the search is stopped and another route is searched. By doing so, the time required for route search processing can be shortened, and since there are many question items, the possibility that the user stops the interaction before reaching the target information can be reduced.

  The node number calculation unit 169 calculates the number of passing keywords as the number of nodes for each route searched by the route search unit 167 and inputs it to the route evaluation calculation unit 171.

  The weight calculation unit 170 calculates the sum of the weights for the keywords that pass through each route searched by the route search unit 167 and inputs the total to the route evaluation calculation unit 171. For example, the related topic extraction unit 156 refers to the weight Wn_weight corresponding to the calculated topic TopicN, and calculates the total weight of keywords passing between routes.

  The route evaluation calculation unit 171 calculates the evaluation value of each route based on the total number of nodes input by the node number calculation unit 169 and the weight input by the weight calculation unit 170. For example, M / N is calculated as a route evaluation value, for example, using the inputted number N of nodes and the total M of topic weights.

  A value obtained by another method may be used as the route evaluation value.

  Based on the route evaluation value calculated by the route evaluation calculation unit 171, the optimum route determination unit 172 changes the topic (keyword) most suitable for the situation (situation), and the number of passages of the topic to the information to be acquired is small. Select a route. For example, a route with a high route evaluation value is selected.

  The question phrase determination unit 163 determines a question phrase to be performed to the user based on the determined route and asks a question. For example, the acquisition item DB 162 is referred to, and a question including a keyword before and after a predetermined keyword is searched according to the determined route. In this case, when there is no question including both front and rear keywords, a question including one keyword may be selected. As for questions that can be answered with YES / NO, questions are continuously asked on topics including similar keywords. In this case, when there is no question including the same keyword, a question accompanying the question of the item of user information desired to be acquired is performed.

  Moreover, in the question memorize | stored in acquisition item DB162, the word contained in them is decomposed | disassembled into a word, and the divided | segmented word can also be changed and used. For example, in “Have you ever been abroad”, the part of “Overseas” can be changed. In addition, the question phrase for the user information item obtained once may be deleted.

  The user information DB 180 stores user information obtained during the conversation, such as date of birth, blood type, hobbies, and the like.

  The response analysis unit 190 analyzes the user response and stores the user information in the user information DB 180. For example, the user's response is divided into words, and a word indicating an item of user information is extracted and stored in the user information DB 180.

  Next, operation | movement of the terminal device 100 provided with the interaction control apparatus 110 concerning a present Example is demonstrated with reference to FIGS.

  In the present embodiment, the interaction that focuses on the reproducibility of the field is performed in consideration of the restriction that the user task is not obstructed in principle.

  In this embodiment, the interaction is performed using the topic (content) that appears in a similar situation. In this way, a topic that appears in a similar situation is not an inappropriate topic in that situation, so the user can be reminded about the topic that he has previously done and is interested in that topic. I think that I can make it disappear.

  First, the situation acquisition unit 122 of the situation acquisition unit 120 acquires situation information from the sensor 200 (acquisition of situation expression items) (step S402). For example, the value of five parameters is acquired by the information from the sensor 200 arranged on the environment side and the function of the terminal device 100, for example.

  In addition, the text appearance frequency calculation unit 132 and the metadata item importance calculation unit 134 of the topic extraction unit 130 acquire a topic TopicN that is considered to be related to the task from the content of the application used by the user. (Step S404).

  Next, the parameter indicating the situation information and the acquired topic are stored in the situation / topic storage unit 142 of the history management unit 140 (step S406). In this case, it is investigated whether the same situation as the situation to be stored is stored in the situation / topic storage unit 142, and if it is stored, the value of the response possibility parameter corresponding to the situation is acquired (step S408). .

  Next, the situation match determination unit 152 determines whether or not the task currently being performed by the user is obstructed (step S410). For example, the determination is made based on the number of keystrokes within a predetermined time and the content of the process. For example, when the number of keystrokes within a predetermined time is large or character conversion is performed, it is determined that the task is disturbed. In addition, when the mail creation screen is active or when the mouse point is moved, it is determined that the task is similarly disturbed.

  If it cannot be determined that the user's task is not hindered (step S410: NO), the process ends without starting the interaction. On the other hand, when it can be determined that the user's task is not hindered (step S410: YES), the interaction start determination unit 154 determines whether or not it is a situation where the interaction may be started (step S412). For example, the determination is made based on a response possibility parameter. That is, it is determined whether or not the value of the response possibility parameter P_int is 0.5 or more.

  When it is determined that the situation is not allowed to start the interaction, that is, when the response possibility parameter P_int is 0.5 or less (step S412: NO), the interaction is terminated without starting. On the other hand, when it is determined that the situation may start the interaction, that is, when the response possibility parameter P_int is 0.5 or more (step S412: YES), the situation match determination unit 152 sets the situation / topic storage unit 142 to It refers to and searches for a past situation, and determines whether there is a situation that matches the current situation (step S414).

  Next, the related topic extraction unit 156 pays out the topic extracted in the situation similar to the current situation from the situation / topic storage unit 142 (step S416). In this case, a weight Wn_weight is attached according to the number of appearances of each topic that has been paid out.

  Next, the target setting unit 161 determines an item TopicG of user information to be acquired (step S418).

  When TopicG is determined, a route that semantically connects TopicG from the current topic TopicN is generated (step S420).

  Next, a question sentence is selected based on the selected optimum route, and a dialogue with the user is performed (step S422). For example, a query including keywords before and after a predetermined keyword is searched from the acquisition item DB 162 based on the optimum route. When there are no questions related to both the front and rear keywords, a question including one keyword is selected. Concerning questions that can be answered with YES / NO, the topic is continued.

  Next, the response analysis unit 190 analyzes the user's response to the question (step S424) and stores it in the user information DB 180 (step S426).

  By doing so, it is possible to acquire user information to be acquired through a question including a keyword of interest to the user while processing the current task.

  Next, a calculation flow of the response possibility parameter P_int in the user response history parameter calculation unit 144 will be described with reference to FIG.

  Since the operation until the optimum route is selected is the same as the operation described with reference to FIG. 4, the description thereof is omitted.

  The question phrase determination unit 163 refers to the acquisition item DB 162, selects a question (step S502), and presents the selected question to the user (step S504).

  Next, it is determined whether or not an input from the user has been made to the presented question (step S506).

  When an input from the user is made (step S506: YES), registration that the determination of the start of interaction is “OK” is performed. For example, a predetermined value, for example, 0.1 is added to the response possibility parameter P_int (step S508).

  On the other hand, when the input from the user is not performed (step S506: NO), registration that the determination of the start of interaction is “impossible” is performed. For example, a predetermined value, for example, 0.1 is subtracted from the response possibility parameter P_int (step S510).

  In this way, by calculating the response possibility parameter and storing it in the situation / topic storage unit 142, it is possible to record a situation in which the user can interact, and to start the interaction start timing for each user.

  Next, a flow for selecting a question item will be described with reference to FIG.

  The operation until the related topic extraction unit 156 pays out the topic extracted in the situation similar to the current situation from the situation / topic storage unit 142 is the same as the flow described with reference to FIG. Description is omitted.

  The goal setting unit 161 determines the item TopicG (goal) of the user information to be acquired (step S602).

  Next, the route creation unit 164 generates a route that semantically connects the current topic TopicN to TopicG (step S604).

  Next, the optimal route selection unit 168 calculates a route evaluation value for evaluating the generated route (step S606), and selects an optimal route based on the calculated route evaluation value (step S608).

  Next, based on the selected optimum route, the acquisition item DB 162 is referred to, a question phrase is determined (step S610), and an inquiry is made to the user (step S612).

  Next, the route selection process (step S604) in the flow for selecting the question item described with reference to FIG. 6 will be described with reference to FIG.

  First, topic transition is performed to randomly change various keywords (step S702), and transition destination data (keyword) is referred to (step S704).

  Next, node tracking, that is, transition from the transition destination data (keyword) to the related keyword (step S706), and keyword setting are performed (step S708). For example, as shown in FIG. 8, semantic upper and lower concepts of words and attribute data associated with each concept are used. The keywords extracted by the concept dictionary 166 include a superordinate concept, a subordinate concept, an environmental concept, a partial material concept, an attribute concept, a synonym and an operation concept. Various keywords are randomly changed from TopicN, and a route search is performed to a location including TopicG as a keyword.

  Next, it is determined whether or not the limit item number, that is, the number of keywords to be changed is within a predetermined value, for example, 20 items (step S710). If the number of keywords to be transitioned is not within 20 items (step S710: NO), the process ends. On the other hand, when the number of topics to be transferred is within 20 items (step S710: YES), it is determined whether or not the goal has been reached (step S712).

  If the goal has not been reached (step S712: NO), the process returns to step S702. On the other hand, when the goal is reached (step S712: YES), the process ends.

  As described above, if the goal cannot be reached even when the limit item number becomes a predetermined value, for example, 20 or more, the search is stopped and another route is searched. By doing so, the number of question items increases and the possibility that the user stops the interaction before reaching the target information can be reduced.

  Next, the route evaluation process (step S606) in the flow for selecting the question item described with reference to FIG. 6 will be described with reference to FIG.

  First, the number N of nodes in the selected route is counted (step S902).

  Next, a sum M of topic weights in the selected route is calculated (step S904). Next, the route evaluation value P is calculated. For example, M / N is calculated and set as a route evaluation value (step S906). Next, it is determined whether or not all route evaluation values have been calculated for the selected route (step S908).

  When all the route evaluation values of the selected route have not been calculated (step S908: NO), the process returns to step S902. On the other hand, when all the route evaluation values of the selected route are calculated (step S908: YES), the calculated route evaluation values P are compared (step S910), and the route with the maximum P is selected (step S910). Step S912).

  For example, as shown in FIG. 10, a case where a route from topic N to topic G is searched will be described.

  In route A, the number of nodes from topic N to topic G is 4, and the total of topic weights is 3. In the route B, the number of nodes from the topic N to the topic G is 6, and the total weight of the topics is 8. In the route C, the number of nodes from the topic N to the topic G is 5, and the total weight of the topics is 7.

  When the route evaluation value (weight / number of nodes) of each route is calculated, the route A, route B, and route C are 3/4, 8/6, and 7/5, respectively. Therefore, the optimum route is the route C having the highest route evaluation value.

  By doing in this way, it is possible to select a route that changes the topic most suitable for the place and has a small number of routed topics to the information to be acquired.

  Next, a specific example will be described with reference to FIG.

  For example, for a user who is watching a breaking sport on a morning commuter train, the topic TopicN is “baseball”, “player”, “soccer”, “Kitajima”. . . The user information TopicG to be acquired is assumed to be “a country to be visited in the future”.

  The route creation unit 164 creates a route from the topic TopicN to the topic TopicG.

  For example, four types of created routes are obtained as shown in FIG.

  The first route (Route (1)) transitions from TopicN (“Baseball”) to “Ichiro” as a famous player, then transitions to the baseball team “Mariners” to which he belongs, Transition to “Seattle” and transition to “USA”, where the city is located. When the country name comes out, the screen shifts to a question for acquiring a user information item of “Country I want to go to” (TopicG) such as “Have you ever been to the United States?”.

  The second route (Route (2)) transitions from TopicN (“Soccer”) to its superordinate concept “Sports”, transitions to “Olympic” as a “Sports” festival, and “Olympics” is held Transition to "Athens" which is the ground, transition from the city to the country name "Greece", and acquire items of user information of "Have you ever been to Greece" and "Country you want to go to" (TopicG) To the question for.

  The third route (Route (3)) transitions from TopicN (“soccer”) to its birthplace “UK”, “Would you have been to England?”, “Country you want to go to” ( Transition to a question for acquiring a user information item of (TopicG).

  The fourth route (Route (4)) transitions from TopicN (“Kitajima”) to “Gold Medal”, transitions to “Medal Acquisition Number”, transitions to “China”, “ ”Or“ Do you want to go in the future ”(TopicG) to make a question for acquiring user information items.

  Choose the best route from these routes.

  As shown in FIG. 11, the weight of each topic routed through each route is 1 for “baseball”, 1 for “Ichiro”, 1 for “soccer”, 1 for “sports”, 6 for “Olympic”, “Athens” "3", "North Island" is 3, "Gold Medal" is 4, "China" is 1. Recently, I watched news about the Olympics in the morning, so there are many occurrences of “Olympic” and “medals”. Other sports were checked at night, so the number of appearances is low.

  Therefore, the route evaluation value of each route calculated based on the total number of nodes and the weight of the nodes is as follows.

Route (1): 5 nodes, weight 2 route evaluation value 2/5 = 0.4
Route (2): 5 nodes, weight 11 Route evaluation value 11/5 = 2.2
Route (3): 2 nodes, weight 1 route evaluation value 1/2 = 0.6
Route (4): 4 nodes, weight 8 route evaluation value 8/4 = 2.0
Thereby, Route (2) is selected.

  In accordance with the selected route, even if the number of times of topic transition is large, a topic suitable for the spot is selected. Also, based on these keywords, the question expression is searched from the question collection.

  Next, another specific example will be described.

  In this example, a case where “birthday” is acquired as user information will be described.

  Currently, the user is searching for restaurants, and “TopicN” such as “Restaurant”, “Italian”, “Map”, “Menu”, “Café”, “Bar”, etc. are acquired.

  The route creation unit 164 creates a route from the topic TopicN to the topic TopicG.

  As a route, TopicN (“Restaurant”) changes to “Birthday Score” as its attribute, and “Birthday” is asked in the flow.

  As another route, TopicN (“menu”) transits to “dessert”, transits to “cake”, transits to “birthday cake”, and asks for “birthday”.

  Also, from TopicN (“Bar”) to “Liquor”, “Wine”, “Cheers”, “Party”, “Birthday Party”, “Birthday Party” Ask "day".

  Choose the best route from these routes. The same evaluation as that described above is performed, and the optimum route is selected.

  As described above, according to the embodiment of the present invention, it is possible to perform topic transition in consideration of the situation, and it is possible to select a topic with less psychological burden on the user. Thereby, the opportunity for the user to acquire information can be increased.

  In the above-described embodiment, the case where the terminal device that acquires user information includes an interaction control device has been described. You may make it perform the interaction for acquiring. By doing in this way, the load with respect to a terminal device can be reduced.

  The interaction control system, the terminal device, and the interaction control method according to the present invention can be applied to a system that performs man-machine interaction for acquiring user information through dialogue.

It is a block diagram which shows the interaction control system concerning one Example of this invention. It is a block diagram which shows the interaction control apparatus concerning one Example of this invention. It is explanatory drawing which shows a user acquisition list and a question collection. It is a flowchart which shows operation | movement of the interaction control system concerning one Example of this invention. It is a flowchart which shows operation | movement of the interaction control system concerning one Example of this invention. It is a flowchart which shows operation | movement of the interaction control system concerning one Example of this invention. It is a flowchart which shows operation | movement of the interaction control system concerning one Example of this invention. It is explanatory drawing which shows the topic transition in a concept dictionary. It is a flowchart which shows operation | movement of the interaction control system concerning one Example of this invention. It is explanatory drawing which shows route evaluation. It is explanatory drawing which shows the specific example of route evaluation.

Explanation of symbols

50 Communication network 100 Terminal device 200 Sensor

Claims (6)

A system having a sensor and a terminal device that acquires information detected by the sensor ,
A situation information acquisition part for acquiring information acquired by the sensor as information of Situation down,
Based on the past history of responses of each information of the acquired situation by the situation information acquisition part, a response possible parameter calculation unit for relative interaction User chromatography The calculated response likelihood parameter indicating the possibility of responding ,
An interaction start determination unit that determines whether or not to perform an interaction based on the response possibility parameter ;
A topic extraction unit that extracts information related to a task from the content of an application used by a user as a topic ;
An item database that stores items to be acquired with respect to user information, a question for acquiring the item, and a question accompanying the question;
A keyword extraction unit that extracts a keyword included in the attribute data of the topic extracted by the topic extraction unit from the concept dictionary as a next topic;
A target setting unit for setting items of user information to be acquired from the items to be acquired stored in the item database ;
Using the keywords extracted by the keyword extracting unit, wherein the extracted topic by topic extraction unit, the route searching unit that creates a route causing the transfer Qian up item of the user information acquiring set by the target setting unit When,
Based on the question stored in the item database, a question regarding a keyword included in the route created by the route search unit is determined, and a question execution unit for asking a question;
Analyzing the response to the question, features and to Resid stem further comprising a response analyzing section for acquiring the user information. In systems according to claim 1,
And the node number calculating unit for calculating a number of keywords to transition about the route created by the route search unit,
Based on the number of keywords that have been calculated by the node number calculating unit calculates the evaluation value for each route, to a; and a best route selection section for selecting an optimal route based on the evaluation value cie Stem. A terminal device that acquires user information based on information detected by a sensor ,
A situation information acquisition unit that acquires information acquired by the sensor as situation information;
A topic extraction unit that extracts information related to a task from the content of an application used by a user as a topic;
An item database that stores items to be acquired with respect to user information, a question for acquiring the item, and a question accompanying the question;
A keyword extraction unit that extracts a keyword included in the attribute data of the topic extracted by the topic extraction unit from the concept dictionary as a next topic;
A target setting unit for setting items of user information to be acquired from the items to be acquired stored in the item database ;
Using the keywords extracted by the keyword extracting unit, wherein the extracted topic by topic extraction unit, the route searching unit that creates a route causing the transfer Qian up item of the user information acquiring set by the target setting unit When,
Based on the question stored in the item database, a question regarding a keyword included in the route created by the route search unit is determined, and a question execution unit for asking a question ;
A terminal device comprising: a response analysis unit that analyzes a response to the question and acquires the user information. The terminal device according to claim 3 ,
And the node number calculating unit for calculating a number of keywords to transition about the route created by the route search unit,
A terminal device comprising: an optimum route selecting unit that calculates an evaluation value of each route based on the number of keywords calculated by the node number calculating unit and selects an optimum route based on the evaluation value . In the terminal device according to claim 3 or 4 ,
Based on the past history of responses of each information of the acquired situation by the situation information acquisition part, a response possible parameter calculation unit for relative interaction User chromatography The calculated response likelihood parameter indicating the possibility of responding ,
An interaction start determining unit that determines whether or not to perform an interaction based on the response possibility parameter. A method in a terminal device for acquiring user information based on information detected by a sensor ,
Situation information acquisition step for acquiring information acquired by the sensor as situation information;
A topic extraction step of extracting information related to the task from the content of the application used by the user as a topic;
A keyword extraction step of extracting a keyword included in the attribute data of the topic extracted by the topic extraction step from the concept dictionary as a next topic;
Regarding user information, an item of user information acquired from the item to be acquired stored in an item database that stores an item to be acquired, a question for acquiring the item, and a question accompanying the question is set. A goal setting step to
A route search step for creating a route for transitioning from the topic extracted by the topic extraction step to the item of user information to be acquired set by the goal setting step using the keyword extracted by the keyword extraction step; ,
A question execution step of determining a question regarding a keyword included in the route created by the route search step based on the question stored in the item database, and asking a question;
How you; and a response analysis step of analyzing the response to the question, to acquire the user information.

Images