JP2018155896A - Control method and controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To change an execution of the sequence data, when there is an omission in the case of execution along the sequence data, according to the omission.SOLUTION: An acquisition unit 11 acquires input data from the outside. A storage unit 14 stores sequence data having a branching structure which branches at least between data having an omission portion and data having no omission portion. A detection unit 12 compares the sequence data stored in the storage unit 14 with the acquired input data and detects an omission state in the input data. A selection unit 13 refers to the stored contents of the storage unit 14 and selects one of the data among data groups branched in the sequence data according to the detected omission state. A control unit 15 controls to cause a voice interactive robot 20 to perform an operation corresponding to the selected data.SELECTED DRAWING: Figure 5

Description

  The present invention relates to a control method and a control apparatus for controlling execution along sequence data.

  As a method in which the voice interaction device performs a voice interaction with a human, there is a method in which the conversation is advanced by the voice interaction device sequentially speaking according to a predetermined scenario. For example, by repeating the process in which the voice interaction device asks the user a question and the voice interaction device further responds in response to the user's response to the question, the dialogue between the two is advanced. For example, in Patent Document 1, based on the result of voice recognition of the user's utterance, the dialogue with the user is controlled according to a scenario given in advance, and a response sentence according to the content of the user's utterance is generated as necessary. A mechanism is disclosed in which a synthesized sentence of a reproduced scenario or a generated response sentence is subjected to speech synthesis processing.

JP 2004-287016 A

  By the way, in a dialogue between humans, for example, a subject, an object, or a sentence ending may be omitted as appropriate. A dialogue that returns a response with no omission part to an utterance containing such an omission part is unnatural and unnatural except in special cases.

  Therefore, the present invention, for example, when there is an omission in the case where the execution is performed along sequence data such that the utterance contents are continuous in time series, the execution of the sequence data is changed according to the omission. Objective.

  In order to solve the above problems, the present invention compares input data acquired from the outside with sequence data having a branch structure that branches at least between data having an omitted portion and data having no omitted portion. A control method comprising: a detection step for detecting an omission state in data; and a selection step for selecting any data in a data group branched in the sequence data according to the detected omission state. I will provide a.

  In the detection step, the position of the omitted portion in the input data may be detected, and in the selection step, the data may be selected according to the detected position of the omitted portion.

  In the detecting step, an amount of the omitted portion in the input data may be detected, and in the selecting step, the data may be selected according to the detected amount of the omitted portion.

  The sequence data may include data corresponding to human operations and data corresponding to device operations.

  The selection step includes a control step of selecting data corresponding to the operation of the device in accordance with the detected omission state, and further controlling the device to perform an operation corresponding to the selected data. It may be.

  Further, the present invention compares the input data acquired from the outside with the sequence data having a branch structure that branches at least between the data with the omitted portion and the data without the omitted portion, and determines the omitted state in the input data. There is provided a control device comprising: a detection unit that detects; and a selection unit that selects any one of the data groups branched in the sequence data according to the detected omission state.

  In the present invention, the computer compares the input data acquired from the outside with the sequence data having a branch structure that branches at least between the data with the omitted portion and the data without the omitted portion. A detection unit for detecting an omission state and a program for functioning as a selection unit for selecting any one of the data groups branched in the sequence data according to the detected omission state are provided.

  According to the present invention, when execution is performed in accordance with sequence data, when there is an omission, the execution of the sequence data can be changed according to the omission.

1 is a block diagram illustrating an overall configuration of a dialog control system according to an embodiment of the present invention. It is a block diagram which shows the electric hardware constitutions of a voice dialogue robot. It is a block diagram which shows the electrical hardware constitutions of a control apparatus. It is a figure which shows an example of the sequence data which the control apparatus has memorize | stored. It is a block diagram which shows the function structure of a control apparatus. It is a flowchart which shows operation | movement of a control apparatus.

[Constitution]
FIG. 1 is a block diagram showing the overall configuration of a dialog control system according to an embodiment of the present invention. The dialogue control system 1 includes a control device 10, a voice dialogue robot 20, and a communication network 90 that connects the control device 10 and the voice dialogue robot 20 so that they can communicate with each other. The communication network 90 is a network including at least one of a wireless communication section according to a wireless communication standard and a wired communication section according to a wired communication standard. The voice interaction robot 20 is an audio interaction device that has an external appearance imitating, for example, the shape of a human being or an animal, and performs a dialogue with a human according to control by the control device 10.

  FIG. 2 is a diagram illustrating a hardware configuration of the voice interactive robot 20. The voice interactive robot 20 is a computer device having a CPU 201 (Central Processing Unit), a ROM (Read Only Memory) 202, a RAM (Random Access Memory) 203, an auxiliary storage device 204, a communication IF 205, a speaker 206, and a microphone 207. The CPU 201 is a processor that performs various calculations. A RAM 203 is a volatile memory that functions as a work area when the CPU 201 executes a program. The ROM 202 is a non-volatile memory that stores a program and data used for starting the voice interactive robot 20a, for example. The auxiliary storage device 204 is a non-volatile storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and stores programs and data used in the voice interactive robot 20. The CPU 201 controls the operation of the voice interactive robot 20 by executing this program. The communication IF 205 communicates with the control device 10 via the communication network 90. The speaker 206 is an utterance unit that emits sound to a human in accordance with control data transmitted from the control device 10. The microphone 207 is a sound collection unit that collects the voice of a person who interacts with the voice dialogue robot 20.

  FIG. 3 is a diagram illustrating a hardware configuration of the control device 10. The control device 10 is a computer device having a CPU 101, a ROM 102, a RAM 103, an auxiliary storage device 104, and a communication IF 105. The CPU 101 is a processor that performs various calculations. The RAM 103 is a volatile memory that functions as a work area when the CPU 101 executes a program. The ROM 102 is a non-volatile memory that stores a program and data used for starting up the control device 10, for example. The auxiliary storage device 104 is a non-volatile storage device such as an HDD or an SSD, and stores programs and data used in the control device 10. When the CPU 101 executes this program, the functions shown in FIG. 5 described later are realized. The communication IF 105 communicates with the voice interactive robot 20 via the communication network 90.

  In the present embodiment, an interactive method is adopted in which the dialogue with the human being is advanced by the voice dialogue robot 20 sequentially speaking according to a predetermined scenario. For example, when the voice dialogue robot 20 asks a human question and repeats the process in which the voice dialogue robot 20 further responds in response to a human response to the question, the dialogue between the two is advanced.

  It is conceivable that a part of a human utterance is omitted as compared with an utterance assumed in a scenario created in advance. In this way, when a part of human utterance is omitted, if only the voice dialogue robot utters without omission as per the scenario, the conversation between the human and the voice dialogue robot will become strange and unnatural. . Therefore, in this embodiment, the omitted state is described in the human utterance part specified in the scenario. Specifically, using a hidden Markov model, we assume a finite state automaton that probabilistically changes the state of continuous utterances between a human and a spoken dialogue robot. Let us consider state transitions in which the utterance of

  On the premise of the above, the auxiliary storage device 104 stores sequence data in which utterances of the voice interactive robot 20 and human utterances are alternately repeated in time series. This sequence data has a branching structure in which the speech of the voice interactive robot branches in accordance with a human speech. This branching structure branches at least between data with an omitted part and data without an omitted part.

  FIG. 4 is a diagram illustrating an example of sequence data stored in the auxiliary storage device 104 of the control device 10. This sequence data is a set of utterance data of the human and the voice interactive robot. This sequence data has a so-called link structure in which human speech data and speech dialogue robot speech data are alternately associated in time series, and speech control proceeds according to the link order. In Figure 4, it is assumed example the interaction begins with the people in the direction of travel to the right from the speech of, for example, by voice dialogue robot 20 "Hello". Further, in the example of FIG. 4, there is a dialogue scenario concerning greetings between a human and a voice dialogue robot first, followed by a dialogue scenario about each other's hobbies, and so on. A case is assumed in which a plurality of scenarios transition in time series. For example, how each scenario changes in time series, such as scenario A branching to scenario B and scenario C and then returning to scenario D. Is possible.

  In the example of FIG. 4, for the utterance of the voice interactive robot 20 that the first "Hello", "Hello", "Continent is", "blood" ... "Oh!" ... "there? Long time no see." -・ ・ Multiple types of human utterances are specified. Here, in the speech group referred to as "Hello", "Continent is", "blood", it is not omitted in the utterance of "Hello", "Conti is", "blood" omitted there in the speech that. Also, the amount of omission in “Chiha” is greater than the amount of omission in “Konha”. And, for the human speech there is no omission of "Hello", is associated with the utterance of the voice interactive robot 20 without omission of "How are you?", "Conti is" omitted the little human beings The utterance is associated with the utterance of the voice dialogue robot 20, which is less abbreviated as “How are you?”, And the abbreviation “How are you?” For the utterance of a person who is often abbreviated as “Chi!”. The utterances of the voice dialogue robot 20 having a large number of voices are associated.

  As described above, in the sequence data, utterances having different omission states of the voice interactive robot 20 are branched according to the omission state of the human utterance. The meaning of the term “abbreviated state” includes, in addition to whether or not there is an omitted part, the meaning of the amount of the omitted part when there is an omitted part. The amount of the omitted portion in the speech of the voice interactive robot 20 is defined according to the amount of the omitted portion detected in step. For example, when there is a large amount of omission in a human utterance, an utterance with a large amount of omission is also defined in the utterance of the voice dialogue robot 20. Utterances with a small amount of omission are specified.

  Here, for the utterance of the first voice interactive robot 20 of "Hello", when a person makes a speech of "Hello", since there is no omitted in the speech of this man, in the utterance of the voice interactive robot 20 Is also associated with an utterance that is not abbreviated as "How are you?" In addition to the utterances of the voice dialogue robot 20 "How are you?", There are further defined a plurality of human utterances such as "I'm fine", "I'm fine" and "I'm fine!". In these utterances of “I'm fine”, “I'm fine” and “I'm fine!”, There is no omission in the utterance “I'm fine”, and there are omissions in the utterances “I'm fine” and “I'm fine!”. Further, the amount of omission in “Genki!” Is larger than the amount of omission in “I'm fine”.

  In addition, for the utterance of the first voice interactive robot 20 of "Hello", when a person makes a speech that "Conch is", because the omission is present in the speech of this man, in the utterance of the voice interactive robot 20 Is also associated with an utterance with the abbreviation "How are you?" For the utterance of the voice dialogue robot 20 “How are you?”, A plurality of human utterances such as “I'm fine”, “I'm fine” and “I'm fine!” Are defined. In these utterances of “I'm fine”, “I'm fine” and “I'm fine!”, There is no omission in the utterance “I'm fine”, and there is an omission in the utterances “I'm fine” and “I'm fine!”.

  Thus, the sequence data has a branching structure that branches when there is an omitted part and when there is no omitted part in human speech. Therefore, by controlling the speech of the voice interactive robot 20 according to such sequence data, different sequence data is executed according to the omitted state in human speech.

  FIG. 5 is a block diagram illustrating a functional configuration of the control device 10. The control device 10 realizes the functions of an acquisition unit 11, a detection unit 12, a selection unit 13, a storage unit 14, and a control unit 15. The acquisition unit 11 is realized by the communication IF 105 of the control device 10, the detection unit 12 is realized by the CPU 101 of the control device 10, the selection unit 13 is realized by the CPU 101 of the control device 10, and the storage unit 14 is an auxiliary storage of the control device 10. The control unit 15 is realized by the CPU 104 and the communication IF 105 of the control device 10.

  The acquiring unit 11 acquires human voice data transmitted from the outside (here, the voice interactive robot 20) as input data. The storage unit 14 stores sequence data having a branch structure that branches at least between data having an omitted portion as exemplified in FIG. 4 and data having no omitted portion. The detection unit 12 compares the sequence data stored in the storage unit 14 with input data acquired from the outside, and detects an omitted state in the input data. The selection unit 13 refers to the storage contents of the storage unit 14 and selects any utterance data from the utterance data group branched in the sequence data according to the detected omission state. The control unit 15 controls the voice interaction robot 20 to perform an utterance operation according to the selected data.

[Operation]
FIG. 6 is a flowchart showing the operation of the control device 10. The process of FIG. 6 is started when the voice dialogue robot 20 performs the first utterance defined by the sequence data. That is, the first utterance data defined by the sequence data is transmitted from the control device 10 to the voice dialogue robot 20, and the voice dialogue robot 20 utters by emitting sound from the speaker 206 according to the utterance data. When an utterance in which a human responds to the utterance of the voice interactive robot 20 is made, the microphone 207 of the voice interactive robot 20 picks up the sound and transmits it to the control apparatus 10 as voice data.

  The communication IF 105 (acquisition unit 11) of the control device 10 acquires voice data transmitted from the voice interaction robot 20 as input data (step S11).

  The CPU 101 (detection unit 12) of the control device 10 recognizes the input data acquired from the outside by a predetermined speech recognition algorithm, the speech recognition result, and the sequence data stored in the auxiliary storage device 104 (storage unit 14). And the omission state in the input data is detected (step S12). In the example of FIG. 4, for example, "the Conch" human speech if were, CPU 101 of the control unit 10 (detecting unit 12) is "on" for a human speech "Hello" in the sequence data It is specified that the omitted “Kon-ha” is a human utterance.

  The CPU 101 (selection unit 13) of the control device 10 refers to the stored contents of the auxiliary storage device 104 (storage unit 14), and the speech data group of the voice interactive robot 20 branched in the sequence data according to the detected omission state. Is selected (step S13). For example, in the illustration of FIG. 4, when human speech is "Conch" for the because the omission of the letter "a" is present for the "Hello", the "on" is omitted "conche is" The utterance data “How are you?” That is the corresponding utterance is selected.

  The CPU 101 and the communication ID 105 (control unit 15) of the control device 10 control the voice dialogue robot 20 to perform an operation corresponding to the selected utterance data (step S14). That is, the CPU 101 and the communication IF 105 (control unit 15) of the control device 10 transmit the selected utterance data to the voice dialogue robot 20, and the voice dialogue robot 20 utters by emitting sound from the speaker 206 according to the utterance data. . Further, when an utterance in which a human responds to the utterance of the voice dialogue robot 20 is performed, the microphone 207 of the voice dialogue robot 20 picks up the sound and transmits it as voice data to the control device 10. Thereafter, the process returns to step S11 again, and the processes of steps S12 to S14 are repeated.

  In the embodiment described above, the human speech and the speech defined in the sequence data are compared, and the speech of the voice interactive robot 20 is selected according to the comparison result. More specifically, in the human utterance, how much of the utterance specified in the sequence data is omitted is analyzed, and the utterance corresponding to the omitted state is selected and output from the voice interactive robot 20. Thereby, the utterance of the voice dialogue robot 20 can be changed according to the omission state of the human utterance, and a natural dialogue is realized between the two.

[Modification]
The embodiment described above can be modified as follows. The following modifications may be implemented in combination with each other.
[Modification 1]
The sequence data in the present invention only needs to include a branch structure in which data having an omitted portion and data having no omitted portion logically branch. For example, the sequence data stored in the control device 10 is sequence data having no omitted portion, and the control device 10 logically branches based on this sequence data between data having the omitted portion and data having no omitted portion. A branch structure to be generated may be generated each time. In this case, the control device 10 generates sequence data without an omitted portion, and algorithm data that generates a branch structure that logically branches between data having an omitted portion and data without the omitted portion based on the sequence data. Therefore, the stored data can be said to be data including a branch structure that logically branches between data having an omitted portion and data having no omitted portion. The control device 10 detects the omission state in the input data by comparing the input data with the sequence data having a branch structure that branches at least between the data with the omission part and the data without the omission part. Depending on the omitted state, one of the data groups branched in the sequence data is selected.

[Modification 2]
The sequence data in the present invention is not limited to the data indicating the utterance content of the human and the voice interactive robot 20 as exemplified in the embodiment, but may be data corresponding to human motion and data corresponding to device operation. . For example, when humans and robots perform together on the stage to perform a performance or play, motion data indicating their respective movements such as movements, postures, facial expressions or positions on the stage of humans and robots may be used as sequence data. In this case, the sequence data has a branching structure that branches at least between operation data having an omission part as human action and operation data having no omission part in human action, and the human action omission state. In response to this, any operation data is selected from the data group branched in the sequence data. In this case, the device includes an operation driving device that drives the device itself instead of the speaker 206 of the embodiment, and a camera for imaging and recognizing human motion instead of the microphone 207 of the embodiment. .

[Modification 3]
The meaning of the term “omitted state” includes the position of the omitted portion in the input data, and the utterance of the voice interactive robot 20 may be selected according to the position of the omitted portion detected in the human speech. For example, the speech of the voice interactive robot 20 is omitted at a position where the attribute is the same as the omitted position in human speech. For example, for a human utterance in which the position corresponding to the subject is omitted, a response is made with the utterance of the voice interactive robot in which the position corresponding to the subject is omitted. For example, for a human utterance in which the position corresponding to the object is omitted, a response is made with the utterance of the voice interactive robot in which the position corresponding to the object is omitted. For example, for a human utterance in which the position corresponding to the end of the sentence is omitted, a response is made with the utterance of the voice interactive robot in which the position corresponding to the end of the sentence is omitted.

[Modification 4]
The block diagram of FIG. 5 used for description of the said embodiment has shown the block of the functional unit. Each of these functional blocks is realized by any combination of hardware and / or software. Further, the means for realizing each functional block is not particularly limited. That is, each functional block may be realized by one device physically and / or logically coupled, and two or more devices physically and / or logically separated may be directly and / or indirectly. (For example, wired and / or wireless) and may be realized by these plural devices. Therefore, as described in the embodiment, the control device according to the present invention can be realized by a device that is provided with all of the functions, or a system in which each function is distributed and implemented in a plurality of devices. There may be. Further, the order of the processing procedures described in the above embodiments may be changed as long as there is no contradiction. About the method demonstrated by embodiment, the element of each step is shown in the example order, and is not limited to the specific order shown.

  The present invention can also be implemented in the form of a control method performed by the control device or a program for causing a computer to function as the control device. Such a program may be provided in a form recorded on a recording medium such as an optical disc, or may be provided in a form such that the program is downloaded to a computer via a communication network such as the Internet, and is installed and made available. Is possible.

DESCRIPTION OF SYMBOLS 1 ... Dialog control system, 10 ... Control apparatus, 11 ... Acquisition part, 12 ... Detection part, 13 ... Selection part, 14 ... Memory | storage part, 15 ... Control part, 20 ... Voice dialogue robot, 90 ... Communication network, 101 ... CPU , 102 ... RAM, 103 ... ROM, 104 ... auxiliary storage device, 105 ... communication IF, 201 ... CPU, 202 ... RAM, 203 ... ROM, 204 ... auxiliary storage device, 205 ... communication IF, 206 ... speaker, 207 ... microphone .

Claims (6)

A detection step of detecting an omission state in the input data by comparing input data acquired from the outside with sequence data having a branch structure that branches at least between data having an omission part and data without the omission part;
And a selection step of selecting any one of the data groups branched in the sequence data according to the detected omission state. In the detection step, the position of the omitted portion in the input data is detected,
The control method according to claim 1, wherein, in the selection step, the data is selected in accordance with the position of the detected omitted part. In the detection step, the amount of omitted parts in the input data is detected,
The control method according to claim 1 or 2, wherein, in the selection step, the data is selected according to the amount of the omitted portion detected. The control method according to any one of claims 1 to 3, wherein the sequence data includes data corresponding to a human operation and data corresponding to a device operation. In the selection step, data corresponding to the operation of the device is selected according to the detected omission state,
The control method according to claim 4, further comprising a control step of controlling the apparatus to perform an operation according to the selected data. A detection unit that detects input data acquired from the outside, sequence data having a branch structure that branches at least between data having an omission part and data having no omission part, and detecting an omission state in the input data;
A control unit comprising: a selection unit that selects any one of the data groups branched in the sequence data according to the detected omission state.

Images