JP2020187163A - Voice operation system, voice operation control method and voice operation control program - Google Patents

Abstract

To provide a voice operation system capable of easily correcting recognized or estimated indication details based upon a user's utterance.SOLUTION: An indication candidate determination unit 12 recognizes or estimates, based upon an indication utterance of a user U recognized by an utterance recognition unit 11, details indicated by the user U to determine a first indication candidate. An indication candidate correction unit 13 performs first candidate notification to vocally output the first indication candidate or an execution state of first predetermined processing corresponding to the first indication candidate, determines a second indication candidate, obtained by correcting the first indication candidate according to an indication by a correction indication utterance to indicate corrections of indication details of the first indication candidate when the correction indication utterance is recognized in response to the first candidate notification, and vocally outputs details of the second indication candidate or an execution state of second predetermined processing corresponding to the second indication candidate.SELECTED DRAWING: Figure 1

Description

The present invention relates to a voice operation system, a voice operation control method, and a voice operation control program.

Conventionally, in a navigation device, it is possible to analyze a user's voice input, divide the utterance content into a plurality of components and display them, select each component individually, and modify the component by re-inputting the voice. Such a configuration has been proposed (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2013-11466

In the case of the conventional configuration described above, when a misrecognition of the utterance content occurs due to a mistake or an erroneous analysis, the user must perform an operation of selecting the component in which the erroneous recognition occurs and re-enter the voice. There is the hassle of having to. In addition, when the user utters only a part of the instruction instead of the complete instruction content and estimates the instruction content by AI, the estimated instruction content may be contrary to the user's intention. In this case as well, the user has to perform an operation of selecting an unintended part and re-enter the voice.
The present invention has been made in view of such a background, and a voice operation system, a voice operation control method, and a voice operation control method, in which the user can easily correct the instruction content recognized or estimated based on the utterance of the user. And to provide a voice operation control program.

As the first aspect for achieving the above object, the utterance recognition unit that recognizes the user's utterance and the instruction content by the user are recognized based on the utterance of the user's instruction recognized by the utterance recognition unit. Alternatively, the first candidate notification that outputs by voice the instruction candidate determination unit that determines the first instruction candidate and the execution status of the first predetermined process according to the first instruction candidate or the first instruction candidate by estimating. When the utterance recognition unit recognizes the correction instruction utterance instructing the correction of the instruction content by the first instruction candidate in response to the first candidate notification, the first instruction candidate is used as the correction instruction utterance. The corrected second instruction candidate is determined in response to the instruction given by the above, and the second candidate notification that outputs the content of the second instruction candidate or the execution status of the second predetermined process according to the second instruction candidate by voice is performed. An example is a voice operation system including an instruction candidate correction unit.

In the voice operation system, the instruction candidate correction unit includes a first instruction element of a predetermined genre in the first instruction candidate and a second instruction element of the predetermined genre in the correction instruction utterance. The second instruction candidate may be determined by correcting the first instruction element based on the second instruction element.

In the voice operation system, the voice operation system is used to instruct the search condition of the destination in the navigation device, and the instruction candidate determination unit determines the first search condition of the destination as the first instruction candidate. The instruction candidate correction unit determines, as the second instruction candidate, a second search condition in which the first search condition is corrected according to the instruction by the correction instruction speech, and the predetermined genre is the location and purpose of the destination. The configuration may be one of the departure date and time to the place, the evaluation rank of the facility as the destination, the type of the facility as the destination, and the identification information of the user when there are a plurality of the users.

In the voice operation system, the instruction candidate correction unit uses the first instruction candidate as the user during the period from the recognition of the correction instruction utterance by the utterance recognition unit to the notification of the second candidate. An error confirmation notification may be configured to notify by voice that the instruction content is different from the intended content.

The voice operation system includes a behavior habit estimation unit that estimates the behavior habit of the user, and the instruction candidate determination unit cannot specify the instruction content intended by the user from the instruction utterance. The first instruction candidate may be determined based on the instruction element included in the instruction utterance and the behavior habit of the user estimated by the behavior habit estimation unit.

As a second aspect for achieving the above object, it is a voice operation control method implemented by a single or a plurality of computers having an utterance recognition unit that recognizes a user's utterance, and the use by the utterance recognition unit. The first instruction candidate is determined by recognizing or estimating the instruction content by the user based on the utterance recognition step in which the utterance of the person is recognized and the instruction utterance of the user recognized by the utterance recognition step. The first candidate notification step for outputting the execution status of the first instruction candidate or the first predetermined process corresponding to the first instruction candidate by voice, and the first candidate notification step. Correspondingly, when the correction instruction utterance instructing the correction of the instruction content by the first instruction candidate is recognized by the utterance recognition unit, the second instruction corrected from the first instruction candidate according to the instruction by the correction instruction utterance. The second candidate notification step of determining the instruction candidate and the second candidate notification step of outputting the content of the second instruction candidate or the execution status of the second predetermined process according to the second instruction candidate by voice are included. A voice operation control method can be mentioned.

As a third aspect for achieving the above object, the utterance recognition process installed on a single computer or a plurality of computers and recognizing the user's utterance on the computer, and the utterance recognition process of the user recognized by the utterance recognition process. The first instruction candidate determination process for determining the first instruction candidate by recognizing or estimating the instruction content by the user based on the instruction utterance, and the first instruction candidate or the first instruction candidate corresponding to the first instruction candidate. 1 The first candidate notification process that outputs the execution status of a predetermined process by voice and the correction instruction utterance that instructs the correction of the instruction content by the first instruction candidate according to the first candidate notification process are the utterance recognition processes. When recognized by, the second instruction candidate determination process of determining the second instruction candidate corrected from the first instruction candidate in response to the instruction by the correction instruction utterance, and the content of the second instruction candidate or the second instruction candidate. An example is a voice operation control program that executes a second candidate notification process that outputs the execution status of the second predetermined process according to the instruction candidate by voice.

According to the voice operation system, the instruction candidate determination unit determines the first instruction candidate based on the user's instruction utterance, and the instruction candidate correction unit determines the first instruction candidate or the first instruction candidate according to the first instruction candidate. The first candidate notification that outputs the execution status of the predetermined process by voice is performed. Then, when the correction instruction utterance corresponding to the first candidate notification is recognized, the instruction candidate correction unit determines the second instruction candidate in which the first instruction candidate is corrected according to the instruction by the correction instruction utterance, and the second instruction candidate is determined. The second candidate notification that outputs the content of the instruction candidate or the execution status of the second predetermined process according to the second instruction candidate by voice is performed. As a result, when the first instruction candidate determined by the voice operation system does not have the intended content, the user issues a correction instruction utterance instructing the change of the instruction. Can be confirmed by voice output and the instruction content can be corrected.

FIG. 1 is a configuration diagram of a navigation device including the functions of a voice operation system. FIG. 2 is an explanatory diagram of user data. FIG. 3 is a flowchart of a process of determining the first search condition of the destination based on the user's utterance and behavioral habits. FIG. 4 is a flowchart of a process of modifying the search condition of the destination based on the utterance of the user. It is explanatory drawing of the 1st dialogue example between a user and a voice operation system. It is explanatory drawing of the 2nd dialogue example between a user and a voice operation system. It is explanatory drawing of the 3rd dialogue example between a user and a voice operation system. It is explanatory drawing of the 4th dialogue example between a user and a voice operation system. It is explanatory drawing of the 5th dialogue example between a user and a voice operation system. It is explanatory drawing of the 6th dialogue example between a user and a voice operation system.

[1. Voice operation system configuration]
The configuration of the voice operation system 2 of the present embodiment will be described with reference to FIG. The voice operation system 2 is configured as a part of the function of the navigation device 1 mounted on the vehicle (not shown). Although the navigation device 1 mounted on the vehicle is shown in the present embodiment, it may be a portable type navigation device. Further, it may be a navigation device configured by executing a navigation application (application program) on a mobile terminal such as a smartphone.

The navigation device 1 includes a CPU (Central Processing Unit) 10, a memory 20, a communication unit 30, a microphone 31, a speaker 32, a touch panel 33, a switch 34, and a GPS (Global Positioning System) unit 35. The communication unit 30 communicates with an external system such as the operation support server 110 via the communication network 100. Further, the communication unit 30 communicates with the user terminal 90 used by the user U of the navigation device 1 via the communication network 100 or directly. The user terminal 90 is a portable communication terminal such as a smartphone, a tablet terminal, or a mobile phone.

The microphone 31 inputs the voice of the user U. The speaker 32 outputs voice guidance and the like to the user U. The touch panel 33 is composed of a flat type display such as a liquid crystal panel and a touch switch arranged on the surface of the display. The switch 34 is operated by pressing the user U. The GPS unit 35 detects the current position (current position of the vehicle) of the navigation device 1 by receiving radio waves transmitted from GPS satellites.

The navigation device 1 sets the destination according to the touch operation of the touch panel 33 by the user U or the operation by the user's voice input to the microphone 31. Then, the navigation device 1 determines the destination based on the current position of the navigation device 1 (current position of the vehicle on which the navigation device 1 is mounted) detected by the GPS unit 35 and the map data 23 stored in the memory 20. We will guide you to the route. The map data may be acquired by accessing an external server such as the operation support server 110 by the communication unit 30.

The voice operation system 2 is composed of a CPU 10, a memory 20, and the like. The CPU 10 reads and installs the control program 21 of the voice operation system 2 held in the memory 20, installs the control program 21, and executes the control program 21 to make an utterance. It functions as a recognition unit 11, an instruction candidate determination unit 12, an instruction candidate correction unit 13, a predetermined processing execution unit 14, an action history storage unit 15, and an action habit estimation unit 16. The CPU 10 corresponds to a single computer or a plurality of computers of the present invention, and implements a voice operation control method. The control program 21 includes the voice operation control program of the present invention.

The action history storage unit 15 stores the action history indicating the place and the date and time when the user U has moved so far in the user data 22. As shown in FIG. 2, the user data 22 includes a user ID 22a, biometric data 22b for identifying the user, and an action for each of a plurality of users who use the vehicle equipped with the navigation device 1. The history 22c is recorded. The biometric data 22b stores data for performing biometric authentication such as a face image, voiceprint, iris, and fingerprint. FIG. 2 illustrates the user data 22 for the user U.

The action history storage unit 15 recognizes the place where the user U has moved and acts based on the transition of the current position of the navigation device 1 detected by the GPS unit 35 and the destination set by the user U. Record in history 22c. In addition, the action history storage unit 15 communicates with the user terminal 90 to set the action schedule of the user U set by the schedule application executed on the user terminal 90 and the payment application executed on the user terminal 90. The action status of the user U recognized by the payment history or the like processed by is recorded in the action history 22c.

The utterance recognition unit 11 analyzes the voice of the user U input to the microphone 31 and recognizes the utterance content of the user U. The instruction candidate determination unit 12 is configured by using an AI (Artificial Intelligence) engine, and is based on the utterance content of the user U recognized by the utterance recognition unit 11 and the action history 22c recorded in the user data 22. The first search condition of the destination intended by the user U (corresponding to the first instruction candidate of the present invention) is determined. In the instruction candidate correction unit 13, the utterance (corresponding to the correction instruction utterance of the present invention) instructing the correction of the search condition by the user U is recognized by the utterance recognition unit 11 in response to the voice output of the first search condition. In this case, the second search condition (corresponding to the second instruction candidate of the present invention) obtained by correcting the first search condition is determined.

The predetermined process execution unit 14 performs a first search process for the destination according to the first search condition (corresponding to the first predetermined process of the present invention) and a second search process for the destination according to the second search condition (corresponding to the first predetermined process of the present invention). The second predetermined process of the present invention) is executed.

The behavior habit estimation unit 16 estimates the behavior habit of the user U as follows, for example, based on the behavior history 22c recorded in the user data 22 by the AI engine.
(1) On weekdays, user U often stops at the supermarket XX a-machi store near his home on the way home from work.
(2) Whenever the user U returns to his / her hometown, he / she leaves his / her home after having dinner on Friday, stops at the supermarket XX a town store near his / her home, and then heads for his / her home.
(3) When traveling by car, user U does not have dinner at the hotel, but instead stops at a two-star French restaurant near the hotel, which is listed in the AA guidebook.
(4) User U is a couple, and his wife commute by car to the FF city hall, which is her workplace, every weekday.
(5) User U is a fan of the GG team, and watches the game between the GG team and other teams held at the local Kiyohara baseball stadium about once a year.

[2. Destination search condition determination process]
According to the flowcharts shown in FIGS. 3 to 4, as shown in the first dialogue example of FIG. 5, when the user U makes an utterance V10 (instruction utterance) instructing the destination, the voice operation system 2 is used. The process of determining the search condition of the destination to be executed will be described. In step S1 of FIG. 3, when the utterance recognition unit 11 recognizes the utterance of the user U from the voice input to the microphone 31, the process proceeds to step S2. The process of recognizing the utterance of the user U in step S2 corresponds to the utterance recognition step in the voice operation control method of the present invention, and corresponds to the utterance recognition process in the voice operation control program of the present invention.

In step S2, the utterance recognition unit 11 determines whether or not the search condition for the destination (content of instruction by the user U) can be specified from the utterance content. Then, the utterance recognition unit 11 proceeds to step S20 when the search condition can be specified, and proceeds to step S3 when the search condition cannot be specified. In step S20, the predetermined process execution unit 14 executes the search process of the destination according to the specified search condition, and proceeds to step S13 of FIG.

In the example of FIG. 5, the utterance V10 saying "Tell me the supermarket in the neighborhood" is recognized by the utterance recognition unit, and in step S3, the instruction candidate determination unit 12 extracts the "supermarket in the neighborhood" as an instruction element. In the following step S4, the instruction candidate determination unit 12 recognizes that it is the user U who has performed the utterance V10 by biometric authentication using the voiceprint. In addition, instead of the voiceprint, biometric authentication may be performed using a face image, a fingerprint, an iris, or the like.

In the next step S5, the instruction candidate determination unit 12 determines the first search condition of the destination based on the instruction element "neighborhood supermarket" and the behavior habit of the user U estimated by the behavior habit estimation unit 16. decide. In the example of FIG. 5, for the user U, the behavioral habit estimation unit 16 described the above (1), "User U stops at the supermarket XX a-machi store near his home on the way home from work on weekdays. It is assumed that the behavioral habit of "often" is presumed. Based on the behavioral habit of (1) above, the instruction candidate determination unit 12 estimates that the "neighborhood supermarket" is a supermarket that the user U always stops by when returning home from work, and sets the first search condition. Decide on "the usual supermarket". The process of determining the first search condition in step S5 corresponds to the first instruction candidate determination step in the voice operation control method of the present invention, and corresponds to the first instruction candidate determination process in the voice operation control program of the present invention.

In the following step S6, the predetermined processing execution unit 14 searches for the route to the "super XXa town store" recorded in the action history 22c with reference to the map data 23 according to the first search condition "usual supermarket". The first search process is executed. Subsequent steps S7 to S10 and S12 of FIG. 4 are processes by the instruction candidate correction unit 13.

In step S7, the instruction candidate correction unit 13 outputs the execution status of the first search process by voice from the speaker 32 (corresponding to the first candidate notification of the present invention). In the example of FIG. 5, the instruction candidate correction unit 13 outputs the voice W10 "I am looking for the usual supermarket" from the speaker 32. The process of outputting the execution status of the first search process from the speaker 32 by voice in step S7 corresponds to the first candidate notification step in the voice operation control method of the present invention, and is the first candidate in the voice operation control program of the present invention. Corresponds to notification processing.

In the following step S8 of FIG. 4, the instruction candidate correction unit 13 determines whether or not the utterance (correction instruction utterance) instructing the user U to correct the first search condition has been recognized by the utterance recognition unit 11. Then, when the instruction candidate correction unit 13 recognizes the correction instruction utterance, the process proceeds to step S9. On the other hand, when the correction instruction utterance is not recognized, the instruction candidate correction unit 13 proceeds to step S13, and in this case, the first search condition is not corrected.

In the example of FIG. 5, the correction instruction utterance V11 by the user U, "It is not the usual supermarket, but the supermarket in the neighborhood of the company" is recognized. In step S9, the instruction candidate correction unit 13 recognizes the content of the correction instruction utterance V11, and corresponds to the genre of 5W1H (When, Where, Who, What, Why, How many, How much) (corresponding to the predetermined genre of the present invention). ) Is used to estimate what the user U wants to correct.

Here, the Where genre includes place names, place names, locations, addresses, latitude / longitude information, feature information, and the like. Features include landmarks, tourist areas (mountains, waterfalls, lakes, etc.), buildings (temples, bridges, buildings, houses, stores, etc.), and commercial facilities such as theme parks and shopping malls. In addition, features may include traffic lights, signs, medians, fences, guardrails, poles, utility poles, and other objects. Feature information is information on the names or locations of these features.

In the example of FIG. 5, the "usual supermarket" denied in the correction instruction utterance V11 is the Where genre. Therefore, the instruction candidate correction unit 13 changes the first instruction element "usual super", which is an instruction element of the Where genre in the first search condition, to the second instruction element, which is an instruction element for which correction is instructed by the correction instruction utterance V11. By replacing with "supermarket in the neighborhood of the company", the second search condition obtained by correcting the first search condition is determined. The process of determining the second search condition obtained by correcting the first search condition in step S9 corresponds to the second instruction candidate determination step in the voice operation control method of the present invention, and the second instruction in the voice operation control program of the present invention. Corresponds to the candidate determination process.

In the following step S10, the instruction candidate correction unit 13 outputs a voice for confirming that the first search condition is different from the search condition intended by the user U (corresponding to the error confirmation notification of the present invention). .. In the example of FIG. 5, the instruction candidate correction unit 13 outputs the voice W11 "Eh, did you make a mistake?" From the speaker 32. By listening to the voice W11, the user U can confirm that the voice operation system 2 has recognized the indication of the error by the user U.

In the next step S11, the instruction candidate correction unit 13 outputs a voice for confirming the correction content from the speaker 32 (corresponding to the correction confirmation notification of the present invention). In the example of FIG. 5, the instruction candidate correction unit 13 outputs the voice W12 saying "It is a supermarket in the neighborhood of the company" from the speaker 32. By listening to the voice W12, the user U can confirm that the voice operation system 2 has recognized the correction content instructed by the user.

In the next step S12, the predetermined processing execution unit 14 refers to the map data 23 according to the second search condition “supermarket in the neighborhood of the company”, and from the current location, the work place of the user U recorded in the action history 22c. The second search process for searching the route to the supermarket in the neighborhood of is executed. In the following step S13, the instruction candidate correction unit 13 outputs the voice W13 "search for a supermarket in the neighborhood of the company", which is the execution status of the second search process, from the speaker 32 (to the second candidate notification of the present invention). Equivalent to). The process of outputting the execution status of the second search process by voice in step S12 corresponds to the second candidate notification step in the voice operation control method of the present invention, and also corresponds to the second candidate notification process in the voice operation control program of the present invention. Equivalent to.

By the processing of FIGS. 3 and 4, as shown in FIG. 5, the user U can instruct the first search condition of the destination by a short utterance "Tell me the supermarket in the neighborhood". Further, the user U determines whether or not the first search condition matches the search condition intended by the user U from the voice output W10 of "searching for the usual supermarket" by the voice operation system 2. be able to. Then, when the first search condition is different from the intended one, the user U instructs the correction of the first search condition, "It is not the usual supermarket, but a supermarket in the neighborhood of the company." By making a simple utterance, it is possible to set a second search condition in which the indicator element of the Where genre of the first search condition is corrected.

As a result, in order to correct the search condition, it is not necessary to reset the search condition from the beginning, and only the instruction element "usual super" of the first search condition, which is different from the intention of the user U, is "company". By replacing it with "a supermarket in the neighborhood of", the second search condition obtained by correcting the first search condition can be easily set.

[3. Modifying the indicator elements of the When genre]
With reference to FIG. 6, an embodiment of a second dialogue example for correcting an indicator element of the When genre (including the departure date and time) in the first search condition will be described. In the example of FIG. 6, for the user U, the behavioral habit estimation unit 16 described the above (2), "Whenever the user U returns to his / her parents' home, he / she leaves home after having dinner on Friday. It is estimated that the behavioral habit is "go to the parents' house after stopping by the nearby supermarket XXa-machi store".

In FIG. 6, first, the utterance recognition unit 11 recognizes the utterance V20 (instructed utterance) of the user U, "return to the parents' home in the hometown, what time will arrive". From the behavioral habits of the user U in (2) above, the instruction candidate determination unit 12 estimates that the user U leaves his / her home for his / her parents' house after having dinner. Then, the instruction candidate determination unit 12 sets, as the first search condition, "the arrival time at the parents'home when leaving the home at around 21:00 after dinner and stopping at the a-machi store of Super XX and heading for the parents' home". Set.

The predetermined processing execution unit 14 executes the first search processing according to the first search condition, and calculates the estimated date and time of arrival at the parents' house. The instruction candidate correction unit 13 outputs the voice W20 of “Friday 23:00”, which is the predicted date and time of arrival at the parents' house, from the speaker 32 (first candidate notification). In response to the first candidate notification, the utterance recognition unit 11 recognizes the utterance V21 (correction instruction utterance) by the user U, "It is not the usual departure time, but now." In the instruction candidate correction unit 13, since the correction instruction utterance V21 denies the "usual departure time" of the When genre and instructs the correction to "from now", the instruction candidate correction unit 13 is the first of the When genre in the first search condition. 1 The second search condition in which the first search condition is corrected is determined by replacing the instruction element "usual departure time" with "from now" in the From genre in which the correction is instructed by the correction instruction utterance V21.

The instruction candidate correction unit 13 outputs the voice W21 "Huh, did you make a mistake?" From the speaker 32 (error confirmation notification), and then outputs the voice W22 "It's about now" from the speaker 32 (correction confirmation). Notification). Then, the predetermined process execution unit 14 executes the second search process according to the second search condition, and the instruction candidate correction unit 13 is the execution status (execution result) of the second search process, "From now on, It is 22:00 today "is output as voice W23 (second candidate notification).
[4. How many, correction of indicator elements of How much genre]
With reference to FIG. 7, an embodiment of a third dialogue example in which the indicator elements of the How many and How much genres in the first search condition are corrected will be described. In the example of FIG. 7, the behavioral habit estimation unit 16 describes the user U in the AA guidebook in the above (3), "When the user U travels by car, he / she does not have dinner at the hotel. It is estimated that there is a behavioral habit of "stopping at a two-star French restaurant near the hotel to eat."

In FIG. 7, first, the utterance recognition unit 11 recognizes the utterance V30 (instruction utterance) of the user U, "search for a restaurant in the neighborhood". From the behavioral habit of (3) above, the instruction candidate determination unit 12 estimates that the user U will have dinner at a two-star French restaurant near the hotel where he / she is staying, as usual. As a search condition, "a two-star French restaurant near the hotel where you are staying" is decided.

The predetermined processing execution unit 14 executes the first search processing according to the first search condition to search for a two-star French restaurant. The instruction candidate correction unit 13 outputs a voice W30 from the speaker 32, which is the execution status of the first search process, "There is a BB restaurant of a two-star French restaurant nearby" (first candidate notification). In FIG. 7, the utterance V31 (correction instruction utterance) by the user U, "Look for three stars today, not two stars," is recognized by the utterance recognition unit 11 in response to the voice W30. .. Here, two stars, three stars, etc. correspond to the evaluation rank of the facility.

In the instruction candidate correction unit 13, since the "two stars" denied by the correction instruction utterance V31 are the How many and How much genres, the instruction candidate correction unit 13 is the first instruction element of the How many and How much genres in the first search condition. The second search condition is determined by replacing "two stars" with the second instruction element "three stars" whose correction is instructed by the correction instruction utterance V31. The second search condition is "a three-star French restaurant near the hotel where you are staying".

The instruction candidate correction unit 13 outputs the voice W31 "Huh, did you make a mistake?" From the speaker 32 (error confirmation notification), and then outputs the voice W32 "Today is three stars, isn't it?" From the speaker 32. (Correction confirmation notification). Then, the predetermined processing execution unit 14 executes the second search processing according to the second search condition, and the instruction candidate correction unit 13 is the execution status of the second search processing, "Three stars, at a French restaurant. The voice W33 saying "There is a CC restaurant in the hotel" is output from the speaker 32 (second candidate notification).

[5. Correction of the indicator element of What genre]
An embodiment of the fourth dialogue example for correcting the indicator element of the What genre in the first search condition will be described with reference to FIG. In the example of FIG. 8, the behavioral habit estimation unit 16 indicates that the user U is listed in the AA guidebook in the above (3) "When the user U travels by car, he / she does not have dinner at the hotel. It is estimated that there is a behavioral habit of "stopping at a two-star French restaurant near the hotel to eat."

In FIG. 8, first, the utterance recognition unit 11 recognizes the utterance V40 (instruction utterance) of the user U, "search for a restaurant in the neighborhood". From the behavioral habits of the user in (3) above, the instruction candidate determination unit 12 estimates that the user U will have dinner at a two-star French restaurant near the hotel where he / she will stay, as usual. , As the first search condition, "a two-star French restaurant near the hotel where you are staying" is determined.

The predetermined processing execution unit 14 executes the first search processing according to the first search condition to search for a two-star French restaurant. The instruction candidate correction unit 13 outputs a voice W40 from the speaker 32, which is the execution status of the first search process, "There is a two-star French restaurant DD restaurant nearby" (first candidate notification). In FIG. 8, the utterance V41 (correction instruction utterance) by the user U, "Look for Italian food today, not French food," is recognized by the utterance recognition unit 11 in response to the voice W40.

Since the "French cuisine" denied by the correction instruction utterance V41 is the What genre, the instruction candidate correction unit 13 changes the "French cuisine" which is the first instruction element of the What genre in the first search condition to the correction instruction utterance V41. The second search condition in which the first search condition is corrected is determined by replacing it with "Italian cuisine", which is the second indicator element whose correction is instructed by. The second search condition is "Italian restaurant near the hotel where you are staying".

The instruction candidate correction unit 13 outputs a voice W41 saying "Huh, did you make a mistake?" From the speaker 32 (error confirmation notification), and then outputs a voice W42 saying "Today is Italian food, isn't it?" From the speaker 32. (Correction confirmation notification). Then, the predetermined processing execution unit 14 executes the second search processing according to the second search condition, and the instruction candidate correction unit 13 is the execution status of the second search processing, "Italian food, in the hotel, EE restaurant. The voice W43 saying "There is" is output from the speaker 32 (second candidate notification).

[6. Correction of the indicator elements of the Who genre]
An embodiment of the fifth dialogue example for correcting the indicator element of the Who genre in the first search condition will be described with reference to FIG. In the example of FIG. 9, for the user U, the behavioral habit estimation unit 16 described in (4) above, "The user U is a couple, and the wife commute to work by car to the FF city hall, which is the workplace, every weekday. The behavioral habit of "is." Is presumed.

In FIG. 9, first, the utterance recognition unit 11 recognizes the utterance V50 (instruction utterance) of the user U (here, the husband) who says "search for a workplace". From the behavioral habit of (4) above, the instruction candidate determination unit 12 estimates that the user U heads for the FF city hall, which is the wife's workplace, as usual, and as the first search condition, "FF of the wife's workplace". "City hall" is decided.

The predetermined processing execution unit 14 executes the first search processing according to the first search condition to search for the route to the FF city hall. The instruction candidate correction unit 13 outputs the voice W50 "I am looking for the FF city hall in the workplace", which is the execution content of the first search process, from the speaker 32 (first candidate notification). Hanako is my wife's name. In FIG. 9, the utterance V51 (correction instruction utterance) by the user U (here, the husband) saying "Look for my workplace today, not my wife's workplace" is uttered in response to the output of the voice W50. It is recognized by the recognition unit 11.

Here, the instruction candidate correction unit 13 identifies the utterance by the wife and the utterance by the husband by biometric authentication by the voiceprints of the wife and the husband stored in the user data 22. The wife and husband correspond to the identification information when there are a plurality of users U.

Since the "wife's" denied by the correction instruction utterance V51 is the Who genre, the instruction candidate correction unit 13 corrects the "wife's" which is the instruction element of the Who genre of the first search condition by the correction instruction utterance V51. The second search condition is determined by replacing it with "I (husband)" and correcting it. The second search condition is "husband's workplace".

The instruction candidate correction unit 13 outputs a voice W51 saying "Huh, did you make a mistake?" From the speaker 32 (error confirmation notification), and then a voice W52 saying "Today is Mr. Kazuo's workplace, isn't it?" From the speaker 32. Output (correction confirmation notification). Kazuo is her husband's name. Then, the predetermined processing execution unit 14 executes the second search processing according to the second search condition, and the instruction candidate correction unit 13 makes a voice saying "I was able to find Mr. Kazuo's workplace" which is the execution status of the second search processing. W53 is output from the speaker 32 (second candidate notification).

[7. Correction of the indicator element of the Where genre]
An embodiment of the sixth dialogue example for correcting the indicator element of the Where genre in the first search condition will be described with reference to FIG. In the example of FIG. 10, regarding the user U, the behavioral habit estimation unit 16 described the above (5), "User U is a fan of the GG team, and the GG team is held at the local Kiyohara baseball stadium about once a year. It is estimated that he has a behavioral habit of watching games with other teams every time.

In FIG. 10, first, the utterance recognition unit 11 recognizes the utterance V60 (instruction utterance) of the user U who wants to go to see the game of the GG team. Based on the behavioral habit of (5) above, the instruction candidate determination unit 12 said, "User U will be held at Kiyohara Stadium today, opening at 18:00 and starting at 18:30, a match between the GG team and the JJ team. Since the user U is a fan of the GG team, he will watch the game on the first base side stand on the bench side of the GG team, which is the home team. " Then, the instruction candidate determination unit 12 determines "a parking lot near the first base side of Kiyohara Baseball Stadium" as the first search condition.

The predetermined processing execution unit 14 executes the first search processing according to the first search condition to search for a route from the current location to the parking lot near the first base side of Kiyohara Baseball Stadium. The instruction candidate correction unit 13 outputs a voice W60 from the speaker 32, which is the execution status of the first search process, "Arrive at the parking lot near the 1st base side of Kiyohara Baseball Stadium at 17:30" (No. 1). 1 candidate notification). In FIG. 10, the utterance V61 (correction instruction utterance) by the user U, "No, I watch it on II TV at the HH yakiniku restaurant," is recognized by the utterance recognition unit 11 in response to the output of the voice W60.

In the instruction candidate correction unit 13, since the "HH yakiniku restaurant" corrected by the correction instruction utterance V61 is the Where genre, the first instruction element "the first base side of the Kiyohara stadium" which is the instruction element of the Where genre in the first search condition. By replacing "a nearby parking lot" with "HH yakiniku restaurant" whose correction was instructed by the correction instruction utterance V61, the second search condition in which the first search condition was corrected is determined. The second search condition is "HH yakiniku restaurant".

The instruction candidate correction unit 13 outputs the voice W61 "Huh, did you make a mistake?" From the speaker 32 (error confirmation notification), and then outputs the voice W62 "HH Yakiniku restaurant, right?" From the speaker 32 (correction). Confirmation notification). Then, the predetermined processing execution unit 14 executes the second search processing according to the second search condition, and the instruction candidate correction unit 13 arrives at the “HH yakiniku restaurant” which is the execution status of the second search processing at 17:50. The voice W63 saying "I will do it" is output from the speaker 32 (second candidate notification).

[8. Other embodiments]
In the above embodiment, the voice operation system 2 is configured as a part of the function of the navigation device 1, but the voice operation system 2 may be configured as a part of other types of devices such as home appliances, or a dedicated device. It may be configured as.

In the above embodiment, as the instructional utterance by the user U, the utterance instructing the search condition of the destination with respect to the navigation device 1 is illustrated, but the instructional utterance regarding other contents may be used. For example, when the operating conditions of an air conditioner provided in a vehicle, the operation of home appliances, the operation of security equipment, etc. are instructed by voice operation, the present invention can be applied to facilitate the correction operation of the instruction content. ..

Further, in the above embodiment, the voice operation system 2 is configured as a part of the function of the navigation device 1, but the voice operation system 2 may be configured as, for example, a voice operation unit of the radio receiver. In this case, the voice operation system 2 receives a radio station (depending on the broadcast station name, channel name, etc., the broadcast frequency of FM, AM, satellite, etc.) in response to the utterance of only "turn on the radio" by the user U. (Specified) may be determined as the radio broadcasting station that the user U often listens to during the time when the utterance is made, based on the behavioral habits of the user. Further, in this case, the voice operation system 2 may determine different radio stations on weekdays and holidays based on the behavioral habits of the user.

In the above embodiment, the instruction candidate correction unit 13 performs the error confirmation notification and the correction confirmation notification between the first candidate notification and the second candidate notification. However, the error confirmation notification and the correction confirmation notification are performed. Only one of the above may be performed, and both notifications may be omitted.

Further, the operation support server 110 may be provided with the configuration of the voice operation system 2. In this case, the operation support server 110 receives the utterance data of the user U transmitted from the navigation device 1, recognizes the utterance content, and determines the first instruction candidate based on the instruction utterance and the user's behavioral habit. To do. Further, the operation support server 110 determines the second instruction candidate by correcting the first instruction instruction according to the instruction by the correction instruction utterance. Then, the operation support server 110 is configured to transmit the information of the first candidate instruction and the second candidate instruction to the navigation device 1.

In the above embodiment, the utterance of the user U is input by the microphone 31 provided in the navigation device 1, and the response voice by the voice operation system 2 is output from the speaker 32 provided in the navigation device 1. As another configuration, the utterance of the user U may be input by a microphone (not shown) provided in the user terminal 90, and the utterance data may be transmitted from the user terminal 90 to the navigation device 1. Further, the response voice data may be transmitted from the navigation device 1 to the user terminal 90, and the response voice may be output from the speaker (not shown) of the user terminal 90.

In the above embodiment, the behavior habit estimation unit 16 is provided, and the instruction candidate determination unit 12 determines the first search condition based on the behavior habit of the user U estimated by the behavior habit estimation unit 16, but the behavior The habit estimation unit 16 may be omitted.

In the above embodiment, the instruction candidate determination unit 12 is configured by using the AI engine, but it may be configured not by using the AI engine.

Note that FIG. 1 is a schematic view showing the functional configuration of the voice operation system 2 divided according to the main processing contents in order to facilitate understanding of the present invention. It may be configured according to the classification of. Further, the processing of each component may be executed by one hardware unit, or may be executed by a plurality of hardware units. Further, the processing of each component may be executed by one program or may be executed by a plurality of programs.

1 ... Navigation device, 2 ... Voice operation system, 10 ... CPU, 11 ... Speaking recognition unit, 12 ... Instruction candidate determination unit, 13 ... Instruction candidate correction unit, 14 ... Predetermined processing execution unit, 15 ... Action history storage unit, 16 ... Behavior habit estimation unit, 20 ... Memory, 21 ... Control program, 22 ... User data, 23 ... Map data, 30 ... Communication unit, 31 ... Mike, 32 ... Speaker, 33 ... Touch panel, 34 ... Switch, 35 ... GPS unit, 90 ... user terminal, 100 ... communication network, 110 ... operation support server, U ... user.

Claims (8)

The utterance recognition unit that recognizes the user's utterance,
An instruction candidate determination unit that determines a first instruction candidate by recognizing or estimating the instruction content by the user based on the instruction utterance of the user recognized by the utterance recognition unit.
The first candidate notification that outputs the execution status of the first instruction candidate or the first predetermined process according to the first instruction candidate by voice is performed, and the instruction by the first instruction candidate is performed in response to the first candidate notification. When the correction instruction utterance instructing the correction of the content is recognized by the utterance recognition unit, the second instruction candidate in which the first instruction candidate is corrected according to the instruction by the correction instruction utterance is determined, and the second instruction is determined. A voice operation system including an instruction candidate correction unit that outputs a second candidate notification by voice output of the content of the candidate or the execution status of the second predetermined process according to the second instruction candidate. When the first instruction candidate includes the first instruction element of the predetermined genre and the correction instruction utterance includes the second instruction element of the predetermined genre, the instruction candidate correction unit sets the first instruction element. The voice operation system according to claim 1, wherein the second instruction candidate is determined by making corrections based on the second instruction element. The voice operation system is used to instruct the search condition of the destination in the navigation device.
The instruction candidate determination unit determines the first search condition of the destination as the first instruction candidate.
The instruction candidate correction unit determines, as the second instruction candidate, a second search condition in which the first search condition is corrected according to the instruction by the correction instruction utterance, and the predetermined genre is the location and purpose of the destination. The description in claim 2, which is one of the departure date and time to the place, the evaluation rank of the facility as the destination, the type of the facility as the destination, and the identification information of the user when there are a plurality of users. Voice control system. In the instruction candidate correction unit, the first instruction candidate is different from the instruction content intended by the user between the time when the correction instruction utterance is recognized by the utterance recognition unit and the time when the second candidate notification is performed. The voice operation system according to any one of claims 1 to 3, wherein an error confirmation notification is performed to notify the fact that the system has been used by voice. The instruction candidate correction unit is a correction confirmation notification that notifies the correction contents of the first instruction candidate by voice between the time when the correction instruction utterance is recognized by the utterance recognition unit and the time when the second candidate notification is performed. The voice operation system according to any one of claims 1 to 4. It is equipped with a behavior habit estimation unit that estimates the behavior habits of the user.
When the instruction candidate determination unit cannot specify the instruction content intended by the user from the instruction utterance, the instruction element included in the instruction utterance and the user estimated by the behavior habit estimation unit. The voice operation system according to any one of claims 1 to 5, which determines the first instruction candidate based on the behavioral habits of the above. A voice operation control method performed by a single computer or a plurality of computers having an utterance recognition unit that recognizes a user's utterance.
An utterance recognition step in which the user's utterance is recognized by the utterance recognition unit, and
The first instruction candidate determination step for determining the first instruction candidate by recognizing or estimating the instruction content by the user based on the instruction utterance of the user recognized by the utterance recognition step.
A first candidate notification step that outputs the execution status of the first instruction candidate or the first predetermined process corresponding to the first instruction candidate by voice, and
When the utterance recognition unit recognizes the correction instruction utterance instructing the correction of the instruction content by the first instruction candidate in response to the first candidate notification step, the first instruction is given in response to the instruction by the correction instruction utterance. The second instruction candidate determination step for determining the second instruction candidate corrected from the candidates, and
A voice operation control method including a second candidate notification step that outputs the content of the second instruction candidate or the execution status of the second predetermined process according to the second instruction candidate by voice. Installed on one or more computers,
Utterance recognition processing that recognizes the user's utterance,
The first instruction candidate determination process for determining the first instruction candidate by recognizing or estimating the instruction content by the user based on the instruction utterance of the user recognized by the utterance recognition process.
The first candidate notification process that outputs the execution status of the first instruction candidate or the first predetermined process according to the first instruction candidate by voice, and
When the correction instruction utterance instructing the correction of the instruction content by the first instruction candidate is recognized by the utterance recognition process in response to the first candidate notification process, the first instruction is given in response to the instruction by the correction instruction utterance. The second instruction candidate determination process for determining the second instruction candidate corrected from the candidates, and
A voice operation control program that executes a second candidate notification process that outputs the content of the second instruction candidate or the execution status of the second predetermined process according to the second instruction candidate by voice.

Images