JPWO2019044534A1 - Information processing device and information processing method - Google Patents

Abstract

The present technology relates to an information processing device and an information processing method that enable more human speech by the Shichigocho constraint. By providing an information processing device provided with a processing unit that converts the input text information into the 75-tone and outputs it, it becomes possible to perform more human utterance by the 75-tone constraint. This technology can be applied to, for example, a system such as a dialogue system that generates a response to a user's utterance, or a system such as a news program production system or a digital signage system that reads out text information by voice synthesis.

Description

The present technology relates to an information processing device and an information processing method, and more particularly to an information processing device and an information processing method that enable more human utterances due to the restrictions of the 75th tone.

In recent years, products and services that engage in voice dialogue have become widespread. For example, Patent Document 1 discloses an electronic toy (domestic robot) that performs an arbitrary operation in response to a voice from the outside.

Japanese Unexamined Patent Publication No. 2002-307354

By the way, in the case of voice dialogue, there are many cases where the response is monotonous, such as "Tomorrow's weather in Yokohama is sunny", which is clearly different from the dialogue between humans.

Then, depending on the user, for example, it is assumed that it is not interesting, the dialogue cannot be enjoyed, it is difficult to remember, and the user does not want to continue using it, so that he / she can speak more humanly. There was a need for technology to make it.

This technology was made in view of such a situation, and it is possible to make more human utterances by the Shichigocho constraint.

The information processing device on one aspect of the present technology is an information processing device including a processing unit that converts input text information into 75-tone and outputs it.

The information processing method of one aspect of the present technology is the information processing method of the information processing apparatus, in which the information processing apparatus converts the input text information into seventy-five tones and outputs the information processing method.

In the information processing device and the information processing method of one aspect of the present technology, the input text information is converted into 75 tones and output.

The information processing device on one aspect of the present technology may be an independent device or an internal block constituting one device.

According to one aspect of the present technology, more human utterances can be made by the Shichigocho constraint.

The effects described here are not necessarily limited, and may be any of the effects described in the present disclosure.

It is a block diagram which shows the example of the hardware composition of the information processing apparatus to which this technology is applied. It is a block diagram which shows the example of the software configuration of the information processing apparatus to which this technology is applied. It is a flowchart which shows the flow of an interactive process. It is a flowchart which shows the flow of an interactive process. It is a figure which shows the example of the context information DB. It is a figure which shows the example of the user feedback information DB. It is a figure which shows the example of the ending list. It is a figure which shows the example of the meaning invariant word list. It is a figure which shows an example of an onomatopoeia. It is a figure which shows the example of the synonym dictionary. It is a figure which shows the example of the news program production system which applied this technology. It is a figure which shows the example of the digital signage system to which this technology is applied. It is a block diagram which shows the example of the structure of the dialogue system.

Hereinafter, embodiments of the present technology will be described with reference to the drawings. The explanations will be given in the following order.

1. 1. First Embodiment: Dialogue system 2. Second embodiment: News program production system 3. Third embodiment: Digital signage system 4. Modification example

<1. First Embodiment>

(Hardware configuration example)
FIG. 1 is a block diagram showing an example of a hardware configuration of an information processing apparatus to which the present technology is applied.

The information processing device 10 of FIG. 1 is, for example, a speaker that can be connected to a network, and is also called a so-called smart speaker or a home agent. In addition to playing music, this type of speaker can perform voice dialogue with a user, voice operation on equipment such as lighting equipment and air conditioning equipment, and the like.

The information processing device 10 is not limited to a speaker, but is configured as, for example, a mobile device such as a smartphone or a mobile phone, or an electronic device such as a tablet computer, a personal computer, a television receiver, or a game machine. May be good.

In FIG. 1, the information processing device 10 includes a CPU 101, ROM 102, RAM 103, an information access unit 104, a hard disk 105, an operation I / F 106, an operation unit 107, a voice input I / F 108, a microphone 109, a video input I / F 110, and a camera 111. , Audio output I / F 112, speaker 113, video output I / F 114, display 115, communication I / F 116, and bus 117.

The CPU (Central Processing Unit) 101, the ROM (Read Only Memory) 102, and the RAM (Random Access Memory) 103 are configured as the control unit 100. The control unit 100 operates as a central processing device in the information processing device 10, such as various arithmetic processes and operation control of each unit.

The information access unit 104 is composed of, for example, a data writing / reading circuit. The information access unit 104 writes various data to the hard disk 105 and reads various data recorded on the hard disk 105 according to the control from the control unit 100 via the bus 117. The hard disk 105 is an HDD (Hard Disk Drive) and is configured as a large-capacity recording device.

The operation I / F 106 is composed of, for example, an operation interface circuit or the like. The operation unit 107 is composed of, for example, a button, a keyboard, a mouse, and the like. The operation I / F 106 supplies an operation signal corresponding to the user's operation to the operation unit 107 to the control unit 100 via the bus 117.

The voice input I / F 108 is composed of, for example, a voice input interface circuit or the like. The microphone 109 is a device (sound collector) that converts sound from the outside into an electric signal. The voice input I / F 108 supplies a voice signal corresponding to the sound picked up by the microphone 109 to the control unit 100, the voice output I / F 112, and the like via the bus 117.

The video input I / F 110 is composed of, for example, a video input interface circuit or the like. The camera 111 has an image sensor, a signal processing circuit, and the like, and images a subject to generate an captured image. The video input I / F 110 supplies the image data of the captured image generated by the camera 111 to the control unit 100, the information access unit 104, the video output I / F 114, and the like via the bus 117.

The audio output I / F 112 is composed of, for example, an audio output interface circuit or the like. The speaker 113 is a device that converts an electric signal into physical vibration to produce sound. The voice output I / F 112 outputs a sound corresponding to the voice signal from the speaker 113 according to the control from the control unit 100 via the bus 117.

The video output I / F 114 is composed of, for example, a video output interface circuit or the like. The display 115 is composed of, for example, a liquid crystal display or an organic EL display. The video output I / F 114 displays various information (for example, characters, images, etc.) according to the video signal on the display 115 according to the control from the control unit 100 via the bus 117.

The communication I / F 116 is composed of, for example, a communication interface circuit or the like. The communication I / F 116 accesses a server (not shown) connected to the Internet 30 in accordance with control from the control unit 100 via the bus 117, and exchanges various data.

In the configuration shown in FIG. 1, the image sensor included in the camera 111 is illustrated as the sensor, but another sensor may be provided. The sensor information obtained as a result of sensing by various sensors is supplied to the control unit 100 via the bus 117 and processed.

Here, for example, as sensors, a magnetic sensor that detects the magnitude and direction of a magnetic field (magnetic field), an acceleration sensor that detects acceleration, a gyro sensor that detects an angle (attitude), an angular velocity, and an angular acceleration, and a gyro sensor that detects an adjacent object are detected. Various sensors can be included, such as a proximity sensor or a biosensor that detects biometric information such as fingerprints, irises, and pulses.

Further, the sensor can include a sensor for measuring the surrounding environment such as a temperature sensor for detecting temperature, a humidity sensor for detecting humidity, and an ambient light sensor for detecting ambient brightness. In addition to the sensor information obtained from the above-mentioned sensor, the sensor information includes various information such as position information calculated from GPS (Global Positioning System) signals and time information measured by a timekeeping means. Can be done.

(Software configuration example)
FIG. 2 is a block diagram showing an example of a software configuration of an information processing apparatus to which the present technology is applied.

That is, in the information processing device 10 shown in FIG. 1, the CPU 101 loads the program recorded in the recording device such as the ROM 102 or the hard disk 105 into the RAM 103 and executes the information processing shown in FIG. The function of unit 150 is realized, and various processes are executed.

In FIG. 2, the information processing unit 150 includes a voice recognition unit 151, an utterance intention understanding unit 152, an application / service execution unit 153, a response generation unit 154, a context acquisition unit 155, a 75-tone conversion unit 156, a voice synthesis unit 157, and a user. It is composed of a feedback collecting unit 158.

In the following description, among the response sentences to the user, the response sentence before being converted to the 75th tone by the 75th tone conversion unit 156 is referred to as "response sentence (before conversion)" and is converted to the 75th tone by the 75th tone conversion unit 156. The converted response statement is referred to as "response statement (after conversion)" to distinguish it. Further, among the response sentences, the response sentence generated as a candidate for the response sentence (after conversion) is referred to as "response sentence (candidate after conversion)".

In addition, the seven-five tone refers to a form in which five tones and seven tones are repeated, such as "5,7,5" and "5,7,5,7,7". As the unit here, for example, a mora (phonology) can be used.

The syllable is the number of vowels when written in Roman letters, and is different from the mora. That is, the mora includes the number of "tsu", "n", and "-" in addition to the number of syllables. For example, the word "Zurich" has three syllables and is also a five mora. However, the definitions of these syllables and mora are examples, and other definitions may be adopted.

The voice recognition unit 151 executes a voice recognition process of replacing the voice signal corresponding to the user's utterance input to the microphone 109 with text information, and understands the utterance intention of the voice recognition result obtained as a result. Supply to unit 152. In this voice recognition process, a database for voice text conversion or the like is used.

The utterance intention understanding unit 152 executes an analysis process (process for understanding the user's intention) on the text information according to the user's utterance based on the voice recognition result supplied from the voice recognition unit 151. The utterance intention understanding unit 152 supplies the intention understanding result obtained as a result of the analysis process to the application / service execution unit 153.

Here, as the user's intention, for example, the intention such as "weather confirmation" or "schedule confirmation" is estimated, and the "date and time" or "location" is analyzed as the parameter (slot).

The application / service execution unit 153 executes an application or service that matches the user's intention based on the intention understanding result supplied from the utterance intention understanding unit 152, and supplies the execution result to the response generation unit 154. ..

Here, for example, when "weather confirmation" is estimated as the user's intention and "date and time" or "location" is analyzed as its parameters, the weather confirmation API provided by an external service ( By passing the parameters of "date and time" and "location" as arguments to Application Programming Interface), it is possible to obtain information about the weather of the target date and time and location.

The response generation unit 154 generates a response statement (before conversion) for the user based on the execution result of the application or service supplied from the application / service execution unit 153, and supplies it to the Shichigocho conversion unit 156.

The context acquisition unit 155 acquires the context information based on the time information, the position information, the analysis result of the captured image captured by the camera 111, and the like, and supplies the context information to the Shichigocho conversion unit 156.

The 75-tone conversion unit 156 executes the 75-tone conversion process for the response sentence (before conversion) supplied from the response generation unit 154, and supplies the response sentence (after conversion) obtained as a result to the speech synthesis unit 157. At this time, the 75-tone conversion unit 156 more appropriately converts the response sentence (before conversion) into the 75-tone based on the context information supplied from the context acquisition unit 155 and the user feedback information supplied from the user feedback collection unit 158. Then, the response sentence (after conversion) can be obtained.

The voice synthesis unit 157 synthesizes the voice based on the response sentence (after conversion) supplied from the 75-tone conversion unit 156, and outputs the voice from the speaker 113. That is, the speech synthesis unit 157 has a function (TTS: Text To Speech) of reading out the text information of the response sentence (after conversion).

The user feedback collecting unit 158 collects information on the user's reaction when the response sentence (after conversion) is output to the user, and supplies the user feedback information to the Shichigocho conversion unit 156.

The configuration of the information processing device 10 has been described above.

(Flow of dialogue processing)
Next, the overall flow of the dialogue processing executed by the information processing unit 150 shown in FIG. 2 will be described with reference to the flowcharts of FIGS. 3 to 4.

Here, as the input to the information processing unit 150, there is a voice signal corresponding to the user's utterance supplied from the microphone 109 or text information corresponding to the operation signal supplied from the operation unit 107 which is a keyboard.

The information processing unit 150 executes the process of step S11 when a voice signal corresponding to the user's utterance is input. In step S11, the voice recognition unit 151 executes a voice recognition process in which the voice signal corresponding to the user's utterance is replaced with text information.

In this way, the text information obtained as a result of the process (speech recognition process) in step S11 or the text information corresponding to the user's input operation is input to the utterance intention understanding unit 152.

In step S12, the utterance intention understanding unit 152 understands the user's intention by executing an analysis process for the text information input therein.

For example, when the user utters "Tell me the weather in Yokohama tomorrow?", The intention of the user is "weather confirmation", and the parameter is "tomorrow". "Date and time" and "Yokohama" "place" can be obtained.

In step S13, the application / service execution unit 153 executes the application or service that meets the user's intention based on the intention understanding result obtained in the process of step S12.

Here, for example, when "weather confirmation" is estimated as the user's intention and the "date and time" which is "tomorrow" and the "location" which is "Yokohama" are analyzed as the parameters, an external service is provided. By passing the parameters of "date and time" which is "tomorrow" and "location" which is "Yokohama" as arguments to the weather confirmation API provided by, you can get information about "weather in Yokohama tomorrow". be able to.

In step S14, the response generation unit 154 generates a response statement (before conversion) for the user based on the execution result of the application or service obtained in the process of step S13.

Here, for example, a response sentence (before conversion) that says "Tomorrow's weather in Yokohama is sunny" is generated based on the information about "Tomorrow's Yokohama weather" obtained from the weather confirmation API.

In step S15, the context acquisition unit 155 acquires context information based on the time information, the position information, the analysis result of the captured image captured by the camera 111, and the like.

Here, the context information includes, for example, the current environmental information related to the utterance such as the time zone and place of the user's utterance, the speaker, the attendees, and the atmosphere of the place, and is temporarily recorded in the context information DB 201. Can be done. However, the current environment information may be processed directly without being recorded in the context information DB201.

In step S16, the 75-tone conversion unit 156 converts the response sentence (before conversion) into the 75-tone by executing the 75-tone conversion process for the response sentence (before conversion) generated in the process of step S14, and the response sentence (before conversion). After conversion) is generated.

In this 75-tone conversion process, the 75-tone conversion process suitable for the context information acquired in the process of step S15 is selected with reference to the user feedback information collected in the process of step S18 described later, and the response statement (before conversion). On the other hand, the selected 75-tone conversion process is executed.

Specifically, for example, the user feedback information includes information obtained by scoring the past environmental information and the user's reaction for each candidate generation pattern that generates the candidates of the 75th tone, and the context information is the current environmental information. The 75-tone conversion unit 156 can select a candidate generation pattern in which the score of the past environmental information that is the same as or similar to the current environmental information is equal to or higher than the threshold value among the past environmental information.

Further, at this time, when a plurality of candidate generation patterns can be selected, for example, one candidate generation pattern can be randomly selected. That is, in the 75-tone conversion process, the selected candidate generation patterns (75-tone candidate generation process (combination)) are executed in order, and each process is executed so that the response statement (before conversion) becomes the 75-tone. ..

The details of the 75-tone conversion process will be described after explaining the overall flow of this dialogue process, but the candidate generation pattern can include, for example, the following 75-tone candidate generation process.

(A) Remove auxiliary words to generate candidates that become 75-tone (B) Remove semantically unnecessary parts to generate candidates that become 75-tone Remove unnecessary parts (C) Add endings to become 75-tone Addition of endings to generate candidates (D) Add words that do not change in meaning to generate candidates that become 75-tone Addition of meaning-invariant words (E) Repeat addition of words to generate candidates that become 75-tone (F) Add onomatope to generate candidates that become 75-tone Add onomatope (G) Add synonyms to generate candidates that become 75-tone by replacing with synonyms (H) Combination of (A) to (G) above

The ending list 211 is used when generating the 75-tone candidate for adding the ending in (C). In addition, the meaning-invariant word list 212 is used when generating the 75-tone candidate for adding the meaning-invariant word in (D). Further, the onomatopoeia list 213 is used when the onomatopoeia addition 75-tone candidate is generated in (F), and the synonym dictionary 214 is used when the onomatopoeia addition 75-tone candidate is generated in (G).

In this way, the response sentence (before conversion) is converted to Shichigocho, and the response sentence (after conversion) is generated. Then, when the 75-tone conversion result obtained in the process of step S16 is output as voice, the response sentence (after conversion) is output to the voice synthesis unit 157.

In step S17, the voice synthesis unit 157 synthesizes the voice based on the response sentence (after conversion) obtained in the process of step S16, and outputs the voice from the speaker 113 via the voice output I / F 112. As a result, the speaker 113 outputs the sound corresponding to the response of the 75th tone.

On the other hand, when the 75-tone conversion result obtained in the process of step S16 is output as a video, the response sentence (after conversion) is output to the display 115 via the video output I / F 114. As a result, the text corresponding to the response of the Shichigocho is displayed on the screen of the display 115.

More specifically, here, for example, the above-mentioned response sentence "Tomorrow's weather in Yokohama is sunny" (before conversion) is "Yokohama's tomorrow's weather is sunny." Like (after conversion), it will be converted to 75-tone and output.

The user feedback collecting unit 158 collects information on the user's reaction when the response sentence (after conversion) is output to the user (S18), and the user feedback information obtained at that time is used as user feedback. It is recorded in the information DB 202. This user feedback information is used in the above-mentioned Shichigocho conversion process.

The overall flow of dialogue processing has been described above.

(Flow of Shichigocho conversion process)
Next, among the above-mentioned dialogue processes shown in FIGS. 3 to 4, the details of the process (75-tone conversion process) in step S16 of FIG. 4 will be described.

In step S111, the Shichigocho conversion unit 156 executes the language analysis process.

In this language analysis process, the morphological analysis process is performed on the response sentence (before conversion) input therein, and the morphological analysis result of the response sentence (before conversion) is obtained. However, in this morphological analysis process, the response sentence (before conversion) is divided into columns of morphemes (words), and the part of speech of each morpheme is determined, but a reading kana is also given to each morpheme. It shall be.

Contents of language analysis processing (S111 in FIG. 4) Input (IN): Response sentence (before conversion)
Output (OUT): Morphological analysis result processing: Morphological analysis processing

In step S112, the Shichigocho conversion unit 156 executes the candidate generation selection process.

In this candidate generation selection process, for example, the user feedback information stored in the user feedback information DB 202 is referred to to identify a case close to the current context information, and the candidate generation process in which the feedback score is equal to or higher than the threshold value. One candidate generation process can be randomly selected from the combinations.

Further, in this candidate generation selection process, for example, it is possible to randomly select a combination of candidate generation processes that does not exist in the user feedback information, and it is also possible to randomly select whether or not it is selected. it can. If there is no case where the feedback score is equal to or higher than the threshold value, the response sentence may be output as it is without executing the 75-tone conversion process for the response sentence (before conversion).

FIG. 5 is a diagram showing an example of the context information DB 201.

In FIG. 5, the context information DB 201 stores a value for each item of context information. In the example of FIG. 5, the context information is the time zone of "Sunday 21:00", the place of "home", the speaker of "Hiroshi Yamada", the attendees of "family", and "fun". The atmosphere etc. are stored.

Here, for the "time zone", for example, time information measured by a time measuring means built in the information processing apparatus 10 may be used. Further, as the "location", position information calculated from a GPS (Global Positioning System) signal or the like may be used.

Further, the "speaker", "accompanying person", and "atmosphere" may be determined by analyzing the image data of the subject captured by the camera 111 and based on the analysis result. For example, "atmosphere" is to judge "fun" or "sad" from the facial expressions of speakers and attendees obtained from the image analysis result, and to determine the atmosphere of the place from the sum of the judgment results. Can be done.

Various studies have already been conducted on technologies for reading human facial expressions and emotions. Here, for example, APIs (Application Programming Interfaces) for recognizing faces, images, and voices using those known technologies have been conducted. ) Can be used. For example, by using a service that provides this kind of API and sending an image containing the faces of the speaker and attendants, for example, joy, surprise, anger, sadness, contempt, disgust, etc. You can get information about the facial expressions and emotions of the attendees.

Further, here, the context information is managed in the database as the context information DB201, but it is not always necessary to manage it in the database. However, for example, among the current environmental information, information such as a location that does not change easily over time should be managed as a database.

FIG. 6 is a diagram showing an example of the user feedback information DB 202.

In FIG. 6, the user feedback information DB 202 stores the context information and the user's reaction as score values for each candidate generation pattern. Here, the score value is calculated. For example, if the user's response to the output of the response sentence (after conversion) generated using the target candidate generation pattern is good, "+1". On the other hand, if the user's reaction is bad, set it to "-1".

For example, "when the user's reaction is good" means when the speaker or the attendant can recognize that the speaker or the attendant has said "interesting" as a voice recognition result, or when the image analysis result shows the speaker or the like. It is assumed that the attendees can recognize that they "laughed". On the other hand, "when the user's reaction is bad" means when the speaker or attendant can recognize what he said, or as an image analysis result, the speaker or attendant, etc. However, it is assumed that it is possible to recognize that it is "angry".

If it is determined that there is no reaction from the user based on the voice recognition result or the image analysis result, the score value can be set to, for example, "0".

In the example of Fig. 6, for the candidate generation pattern of "addition of ending & addition of onomatopoeia", context information of "time zone = Sunday 21:00", "location = home", "speaker = Hiroshi Yamada", The score that is "+129" is stored.

In addition, in the example of FIG. 6, various context information is also applied to candidate generation patterns such as "addition of meaning invariant words", "without particles & repeated addition & replacement of synonyms", and "without particles & removal of unnecessary parts". And scores such as "+86", "+29", and "-42" are given.

Here, for example, when "+80" is set as the threshold value to be compared with the score, "addition of ending & addition of onomatopoeia" and "addition of meaning-invariant word" can be selected as candidate generation patterns. , Here, for example, one candidate generation pattern among "addition of ending & addition of onomatopoeia" and "addition of meaning invariant word" can be randomly selected.

In this way, when a plurality of candidate generation patterns can be selected, one candidate generation pattern is randomly selected, so that one candidate generation pattern satisfying a certain condition is selected. The 75-tone conversion is performed according to the candidate generation pattern, and various patterns of 75-tone response sentences can be presented to the user.

For example, if the candidate generation pattern with the highest score is selected, there is a high possibility that only a specific candidate generation pattern will be repeatedly selected. Therefore, the candidate generation patterns are randomly selected as described above. However, depending on the operation of the dialogue system, it is not limited to random, and for example, the one with the highest score may be selected. Further, in this example, an example of selecting from candidate generation patterns having a threshold value or more is shown, but for example, a candidate generation pattern having a threshold value or less may be randomly selected.

Contents of candidate generation selection process (S112 in FIG. 4) Input (IN): Response sentence (before conversion), morphological analysis result (S111 in FIG. 4), context information (S15 in FIG. 4), user feedback information (in FIG. 4). S18)
Output (OUT): Selection result of candidate generation processing (which candidate generation processing to apply)
Processing: Candidate generation selection processing

In step S113, the 75-tone conversion unit 156 executes the 75-tone candidate generation process.

In this 75-tone candidate generation process, for example, among the candidate generation processes for generating 75-tone candidates by removing particles, removing unnecessary parts, adding endings, adding meaning-invariant words, adding repeatedly, adding onomatopoeia, or replacing synonyms, step S112 One or more candidate generation processes are executed according to the selection result obtained in the process of.

For details of the candidate generation processing such as particle removal, unnecessary part removal, ending addition, meaning invariant addition, repeated addition, onomatopoeia addition, or synonym replacement listed here, refer to the first candidate generation processing to the seventh candidate generation. The process will be described later.

Contents of the Shichigocho candidate generation process (S113 in FIG. 4) Input (IN): Response sentence (before conversion), morphological analysis result (S111 in FIG. 4), selection result of candidate generation process (S112 in FIG. 4)
Output (OUT): Response statement (candidate after conversion)
Processing: Shichigocho candidate generation processing according to the selection result of candidate generation processing

(A) First candidate generation process

In the first candidate generation process, by removing the particles included in the response sentence (before conversion), a response sentence (post-conversion candidate) having a 75-tone tone is generated (S113A in FIG. 4).

The following is an example of the response sentences of the pre-conversion and post-conversion candidates when the 75-tone candidate without particles is generated. However, the response sentence (before conversion) is written in Japanese, Romaji, and English, and the response sentence (after conversion) is written in two types, Japanese and Romaji.

At that time, Japanese is abbreviated as "Japanese", Romaji is abbreviated as "Ro", and English is abbreviated as "English". The notation of examples of these response statements is the same for other response statements described below.

Example of generating 75-tone candidates without particles

Response statement (before conversion)
(Sun): A popular izakaya in Matsuyama is Saba Saba Tei.
(B): matsuyama de ninki no izakaya ha sabasabatei
(English): Popular bar in Matsuyama is Sabasabatei.

Response statement (candidate after conversion)
(Sun): Saba Saba Tei, a popular izakaya in Matsuyama.
(B): matsuyama de ninki no izakaya sabasabatei

In this example of generating Shichigocho candidates without particles, the response sentence (candidate after conversion) is made into Shichigocho by removing the particle "ha" that connects "Izakaya" and "Saba Saba Tei" in the response sentence (before conversion). I am trying to be.

(B) Second candidate generation process

In the second candidate generation process, a semantically unnecessary part included in the response sentence (before conversion) is removed to generate a response sentence (post-conversion candidate) having a 75-tone tone (S113B in FIG. 4).

The following is an example of the response sentences of the candidates before and after conversion when the 75-tone candidate for removing unnecessary parts is generated.

Example of generating 75-tone candidates for removing unnecessary parts

Response statement (before conversion)
(Sun): The current time in London is 8 pm.
(B): rondon no genzai zikoku ha yoru hatizi ni narimasita
(UK): The current time in London is 8 o'clock in the evening.

Response statement (candidate after conversion)
(Sun): The current time in London is 8:00 pm.
(B): rondon no genzai zikoku ha yoru hatizi

In this example of Shichigocho candidate generation for removing unnecessary parts, the response sentence (before conversion) is followed by "8 pm" and the response sentence (before conversion) is removed by removing the semantically unnecessary part that is "become". (Candidate after conversion) is set to 75 tone.

(C) Third candidate generation process

In the third candidate generation process, the ending list 211 is used to add the ending to the response sentence (before conversion) to generate a response sentence (candidate after conversion) having a 75-tone tone (S113C in FIG. 4).

FIG. 7 is a diagram showing an example of the ending list 211. In FIG. 7, "*" (asterisk) represents an arbitrary character (character string) or part of speech.

In FIG. 7, "this notation" is a word added to the end of a word, and its part of speech is represented as "this part of speech". Further, in a certain response sentence, the notation before the "main notation" is the "pre-notation", and the notation after the "main notation" is the "post-notation". Further, the part of speech of "pre-notation" is represented as "pre-part of speech", and the part of speech of "post-notation" is represented as "post-part of speech".

In the example of FIG. 7, this notation of "yo" is a particle (final particle), and an arbitrary character consisting of an auxiliary verb (termination form) is specified as a pre-notation, and an arbitrary part of speech is specified as a post-notation. Any character is specified. Further, in the example of FIG. 7, this notation of "ne" is a particle (final particle), and an arbitrary character string consisting of a particle (case particle) is specified as a pre-particle, and an arbitrary character string consisting of a particle (case particle) is specified as a postpositional notation. Any character consisting of the part of particles of is specified.

At this time, when the pre-morpheme, the present morpheme, and the post-morpheme match the conditions, the meaning does not change even if the present morpheme is added, so that the present morpheme can be added to the response sentence (before conversion). ..

The meaning of "*" shown in FIG. 7 and the relationship between the "main notation" and the "pre-notation" and "post-notation" are the same in other figures (FIGS. 8 and 9) described later. It is said that.

The following is an example of the response sentences of the candidates before and after conversion when the 75-tone candidate with the ending added is generated.

First example of generating Shichigocho candidates with additional endings

Response statement (before conversion)
(Sun): An email was sent to Mr. Kishi.
(B): kisi san ni me-ru wo sousin simasita
(English): I sent an email to Kishi.

Response statement (candidate after conversion)
(Sun): I sent an email to Mr. Kishi.
(B): kisi san ni me-ru wo sousin simasita yo

In the first example of generating the Shichi-Go-Chinese candidate with this ending added, the response sentence (post-conversion candidate) is created by adding the ending that is "yo" after "sent" in the response sentence (before conversion). I try to be in the 75th tone.

Second example of generating Shichigocho candidates with additional endings

Response statement (before conversion)
(Sun): Tomorrow's weather in Yokohama will be fine.
(B): yokohama no asita no tenki ha hare desu
(English): The weather in Yokohama is sunny tomorrow.

Response statement (candidate after conversion)
(Sun): Tomorrow's weather in Yokohama will be sunny.
(B): yokohama no asita no tenki ha hare nanosa

In the second example of generating the 75-tone candidate with the addition of the ending, the response sentence is the 75-tone (post-conversion candidate) by adding the ending of "Nanosa" after "Sunny" in the response sentence (before conversion). I am trying to be. However, in this second example, the ending of "desu" in the response sentence (before conversion) is deleted.

As another example of generating the 75-tone candidate with the addition of the ending, for example, add the ending of "ne" to the response sentence (before conversion) of "Japan is" and use "Japan is". It is assumed that a certain response statement (candidate after conversion) is generated.

(D) Fourth candidate generation process

In the fourth candidate generation process, by using the meaning-invariant word list 212 and adding a word (meaning-invariant word) whose meaning does not change to the response sentence (before conversion), a response sentence (post-conversion candidate) that becomes 75-tone is obtained. Generate (S113D in FIG. 4).

FIG. 8 is a diagram showing an example of the meaning immutable word list 212.

In the example of FIG. 8, this notation, which is "after all", is an adverb, and any character consisting of an arbitrary part of speech is specified as the pre-notation and the post-notation. Further, in the example of FIG. 8, this notation of "by the way" is a conjunction, and an arbitrary character consisting of an arbitrary part of speech is specified as a pre-notation and a post-notation.

At this time, when the pre-morpheme, the present morpheme, and the post-morpheme match the conditions, the meaning does not change even if the present morpheme is added, so that the present morpheme can be added to the response sentence (before conversion). ..

The following is an example of the response sentences of the candidates before and after conversion when the 75-tone candidate with the addition of meaning-invariant words is generated.

Example of generating 75-tone candidates with additional meaning-invariant words

Response statement (before conversion)
(Sun): Shows a map of Kawasaki.
(B): kawasaki no tizu wo utusi masu
(English): I will display a map of Kawasaki.

Response statement (candidate after conversion)
(Sun): After all, the map of Kawasaki will be projected.
(B): kawasaki no tizu wo yappari utusi masu

In the example of generating the Shichigocho candidate for adding the meaning invariant word, the response sentence is added by adding the meaning invariant word that is "after all" between "map" and "project" in the response sentence (before conversion). (Candidate after conversion) is set to 75 tones.

(E) Fifth candidate generation process

In the fifth candidate generation process, the words included in the response sentence (before conversion) are repeated to generate a response sentence (post-conversion candidate) having a 75-tone tone (S113E in FIG. 4).

The following is an example of the response sentences of the pre-conversion and post-conversion candidates when the 75-tone candidates are repeatedly added.

Example of repeating additional 75-tone candidate generation

Response statement (before conversion)
(Sun): The weather forecast for the day after tomorrow will be fine.
(B): asatte no tenkiyohou ha hare desu
(English): The weather forecast for the day after tomorrow is sunny.

Response statement (candidate after conversion)
(Sun): The weather forecast for the day after tomorrow will be sunny and sunny.
(B): asatte no tenkiyohou ha hare hare hare

In the example of generating the 75-tone candidate added repeatedly, the response sentence (post-conversion candidate) is made to be 75-tone by repeating the word "sunny" in the response sentence (before conversion) three times. However, in this example of Shichigocho candidate generation, the ending of "desu" in the response sentence (before conversion) is removed.

(F) 6th candidate generation process

In the sixth candidate generation process, the onomatopoeia 213 is used to add the onomatopoeia to the response sentence (before conversion) to generate a response sentence (post-conversion candidate) having a 75-tone tone (S113F in FIG. 4).

FIG. 9 is a diagram showing an example of onomatopoeia 213.

In the example of FIG. 9, this notation, which is "Janjan", is an adverb, and an arbitrary character consisting of an arbitrary part of speech is specified as a prenotation, and an arbitrary character consisting of a verb is specified as a postnotation. There is. Further, in the example of FIG. 9, this notation of "gingin" is an adjective verb, and an arbitrary character consisting of an arbitrary part of speech is specified as a prenotation, and an arbitrary character consisting of a verb is specified as a postnotation. Is specified.

At this time, when the pre-morpheme, the present morpheme, and the post-morpheme match the conditions, the meaning does not change even if the present morpheme is added, so that the present morpheme can be added to the response sentence (before conversion). ..

The following is an example of the response sentences of the pre-conversion and post-conversion candidates when the 75-tone candidate is added to the onomatopoeia.

First example of generating 75-tone candidates for onomatopoeia

Response statement (before conversion)
(Sun): I have received an email. what should I do?
(B): me-ru ga kite imasu dou simasuka
(English): E-mail is coming. What will you do?

Response statement (candidate after conversion)
(Sun): I'm getting an email. What do you do
(B): me-ru gane zyanzyan kite imasu dou simasu

In the first example of generating the 75-tone candidate for adding onomatopoeia, the response sentence (candidate after conversion) is created by adding the onomatopoeia that is "janjan" before "coming" in the response sentence (before conversion). I try to be in the 75th tone. Further, in this first example, the ending of "ne (final particle)" is added after "mail is", and the particle of "ka (final particle)" is omitted.

Here, since "Janjan" has 4 mora, it has 5 sounds, 8 sounds, and 5 sounds, but it is said that it is within the permissible range. Further, the concept of morphemes before and after may be a condition. For example, in the case of this first example, there are morphemes representing quantities before and after "Janjan".

Second example of generating 75-tone candidates with onomatopoeia

Response statement (before conversion)
(Sun): The timer was set after 2 minutes.
(B): ni hungo ni taima setto simasita
(English): Timer set after 2 minutes.

Response statement (candidate after conversion)
(Sun): Timer set after 2 minutes.
(B): ni hungo ni taima setto tikutakku

In the second example of generating the 75-tone candidate for adding onomatopoeia, the response sentence (post-conversion candidate) is 75-tone by adding the onomatopoeia that is "ticktack" after the "timer set" in the response sentence (before conversion). I am trying to be. However, in this second example, unnecessary parts are removed for "done" in the response sentence (before conversion).

Third example of generating 75-tone candidates with onomatopoeia

Response statement (before conversion)
(Sun): The weather forecast for the day after tomorrow will be fine.
(B): asatte no tenkiyohou ha hare desu
(English): The weather forecast for the day after tomorrow is sunny.

Response statement (candidate after conversion)
(Sun): The weather forecast for the day after tomorrow will be fine (b): asate no tenkiyohou ha hare re re re

In the third example of this onomatopoeia-added 75-tone candidate generation, when converting the response sentence (before conversion) to 75-tone, it will never be 75-tone, so "re" included in the word "sunny" is repeated three times. By doing so, the response sentence (candidate after conversion) is made to be 75-tone. However, in this third example, unnecessary parts are removed for "desu" in the response sentence (before conversion). If the tone does not change to 75, you may add a word such as "Dadada" to fill it.

(G) 7th candidate generation process

In the seventh candidate generation process, the synonym dictionary 214 is used to replace the words included in the response sentence (before conversion) with synonyms to generate a response sentence (post-conversion candidate) that has a tone of 75 (Fig.). 4 S113G).

FIG. 10 is a diagram showing an example of the synonym dictionary 214.

In FIG. 10, "notation 1" and "notation 2" have the same meaning, the part of speech of "notation 1" is represented as "part of speech 1", and the part of speech of "notation 2" is referred to as "part of speech 2". It is expressed as.

In the example of FIG. 10, the noun "ramen" and the noun "Chinese soba" are designated as synonyms. Further, in the example of FIG. 10, the adjective "happy" and the noun "happy" are designated as synonyms.

At this time, the first morpheme can be replaced with the second morpheme. On the contrary, the second morpheme may be replaced with the first morpheme.

The following is an example of the response sentences of the pre-conversion and post-conversion candidates when the 75-tone candidate for synonymous word substitution is generated.

First example of generating 75-tone candidates for synonymous substitution

Response statement (before conversion)
(Sun): Chinese soba is recommended tonight.
(B): konban ha tyuuka soba ga osusume desu
(English): Chinese noodle is recommended tonight

Response statement (candidate after conversion)
(Sun): I recommend ramen tonight.
(B): konban ha ra-men ga osusume yo

In the first example of generating the Shichi-Go-Chinese candidate for synonymous word substitution, the word "Chinese soba" in the response sentence (before conversion) is replaced with the synonym "ramen" to replace the response sentence (candidate after conversion). ) Is set to 75. However, in this first example, unnecessary parts are removed from "desu" in the response sentence (before conversion), and the ending "yo (final particle)" is added.

Second example of generating 75-tone candidates for synonymous substitution

Response statement (before conversion)
(Sun): I plan to shop the day after tomorrow.
(B): asatte ha kaimono no yotei desu
(English): I will go shopping the day after tomorrow.

Response statement (candidate after conversion)
(Sun): The day after tomorrow will be shopping.
(B): asatte ha syoppingu no yotei desu

In the second example of generating the Shichi-Go-Chinese candidate for synonymous word substitution, the word "shopping" in the response sentence (before conversion) is replaced with the synonym for "shopping", so that the response sentence (candidate after conversion) Is trying to be in the 75th tone.

(H) Combination of 1st candidate generation processing to 7th candidate generation processing

The first candidate generation process to the seventh candidate generation process described above can be used alone as a candidate generation pattern, or a combination of them to generate a response sentence (candidate after conversion) that has a tone of 75. You may try to do it.

Note that (A) particle removal and (B) generation of 75-tone candidates for removing unnecessary parts can be classified as the first case of excluding words included in the response sentence (before conversion). In addition, (C) addition of endings, (D) addition of meaning-invariant words, (E) addition of repetitions, and (F) addition of onomatopoeia are classified as the second case of adding words to the response sentence (before conversion). be able to. Further, (G) synonymous word substitution can be classified as a third case of replacing a word contained in a response sentence (before conversion).

Below, as an example of the response sentence at the time of generating the 75th tone candidate when any of the 1st candidate generation processing and the 7th candidate generation processing is combined, before the conversion at the time of generating the 75th tone candidate by combining the addition of the ending and the synonym substitution. An example of the response statement of the candidate after conversion is shown.

Example of adding endings & generating 75-tone candidates for synonymous replacement

Response statement (before conversion)
(Sun): Today's fortune-telling result is very good.
(B): kyou no uranai kekka ha totemo ii desu
(English): Today's fortunetelling results are very good.

Response statement (candidate after conversion)
(Sun): Today's fortune-telling result is good.
(B): kyou no ne uranai kekka ha guddo desu

In the example of this addition of ending & generation of 75-tone candidates for synonymous replacement, the ending of "ne" is added after "today" in the response sentence (before conversion), and the word "very good". Is replaced with a synonym for "good" so that the response sentence (candidate after conversion) becomes 75-tone.

(I) Examples of languages other than Japanese

In the above explanation, an example of converting a Japanese response sentence into a 75-tone conversion has been described, but the 75-tone conversion can be performed in the same manner for languages other than Japanese. Here, as another language, English is taken as an example, and a case where an English response sentence is converted to Shichigocho is shown.

The following is an example of the response sentences of the candidates before and after conversion at the time of generating the Shichigocho candidate for the English response sentence.

First example of generating 75-tone candidates for English response sentences

Response statement (before conversion)
(English): Today's weather in Tokyo is rainy.

Response statement (candidate after conversion)
(English): In Tokyo today's weather is rainy.

In the first example of generating the Shichigo-Chinese candidate of this English response sentence, the word order of "today's weather" and "in Tokyo" is exchanged between the response sentence (before conversion) and the response sentence (candidate after conversion). The response sentence is made to be 75-tone.

Second example of generating 75-tone candidates for English response sentences

Response statement (before conversion)
(English): These are maps you want.

Response statement (candidate after conversion)
(English): These are maps which you want.

In the second example of generating the 75-tone candidate of the English response sentence, the response sentence is made to be the 75-tone by inserting "which" after "maps" in the response sentence (before conversion).

Third example of generating 75-tone candidates for English response sentences

Response statement (before conversion)
(English): This is recommendation for you.

(Candidate after conversion)
(English): This is a recommended restaurant.

In the third example of generating the 75-tone candidate of this English response sentence, the response sentence is changed to the 75-tone by deleting "for you" in the response sentence (before conversion) and inserting "a" and "restaurant". I am trying to be.

Fourth example of generating 75-tone candidates for English response sentences

Response statement (before conversion)
(English): You got a mail.

Response statement (candidate after conversion)
(English): Just you got a mail you're happy?

In the fourth example of generating 75-tone candidates for this English response sentence, by inserting "Just" and "you're happy?" Into the response sentence (before conversion), the response sentence becomes 75-tone. I have to.

In addition, although there are seventy-five tones in languages other than Japanese, such as English mentioned above, there are languages that are not as familiar as Japanese, so user feedback is required when processing for such languages. It is also possible to take measures such as setting the threshold value in the threshold value processing using information higher than the threshold value in the case of Japanese.

The first embodiment has been described above.

In this first embodiment, the response (text information) to the user's utterance is converted into 75-tone and output so that more human utterance (for example, comfortable utterance) can be performed by the 75-tone constraint. I have to.

In other words, in recent years, products and services that engage in voice dialogue have become widespread, but when conducting voice dialogue, the response is simple, for example, "Tomorrow's weather in Yokohama is sunny." Many of them are clearly different from the dialogues that they carry out. Then, depending on the user, for example, it is assumed that it is not interesting, the dialogue cannot be enjoyed, it is difficult to remember, and the user does not want to continue using it, so that he / she can speak more humanly. Was required to be.

Therefore, in the first embodiment, by returning the response to the user's utterance in the 75th tone, a human touch comes out and the dialogue with the system can be enjoyed. For example, by changing the above-mentioned "Tomorrow's weather in Yokohama is sunny" to "Tomorrow's weather in Yokohama is sunny", it will be 5 sounds, 7 sounds, and 5 sounds. Can have a human dialogue.

Further, in the first embodiment, by using the user feedback information and the context information, it is possible to more appropriately convert to the 75-tone when performing the 75-tone constraint.

In other words, by responding in the Shichigo-cho, some users may feel annoyed or annoyed and feel uncomfortable.Therefore, as a result of the system's utterance, in a past situation similar to the current situation, Collect how the user's reaction was (for example, whether it was good or bad), and preferentially select the 75-tone conversion process that the user's reaction was good. As a result, for example, it is possible to perform a response in the 75th tone that knows the time, place, and case (TPO: Time Place Occasion).

In addition, by using this technology, the agent of the dialogue system can characterize (characterize) the agent by making a more human utterance due to the Shichigocho constraint.

In addition, by using this technology, even if the accuracy of speech synthesis (TTS) by the speech synthesis unit 157 is low (for example, there is instability of intonation), the accuracy is low by performing the 75-tone constraint. Can be hidden. Furthermore, by using this technology, even if the utterance intention understanding unit 152 cannot analyze the user's intention, the user allows the system to not understand the intention by returning a response in the 75th tone. The possibility can be increased.

The above-mentioned 75-tone candidate generation process is an example, and any process may be adopted as long as the response sentence (before conversion) can be converted to the 75-tone. For example, if the user says "play a classical song", the system would normally respond with "no classical song", but here is a case frame dictionary. You may use to respond with something like "Mayonnaise that you can call."

In other words, the word "kake" has the meaning of "playing music" as well as "playing seasonings" such as "mayonnaise", so here, "kake" is intentionally different. By using it in a sense, we are trying to have a human dialogue. The case frame is a collection of nouns and related nouns according to the usage of the words. Further, when converting the response sentence (before conversion) to the 75-tone, the verbal tone may be given priority as a whole.

<2. Second Embodiment>

By the way, in the above-described first embodiment, when the dialogue with the user, that is, the content of the user's utterance is analyzed and the response to the utterance content is output, a more human response is performed due to the Shichigocho constraint. I tried to be.

This technology is not limited to such a dialogue system, and can be applied to various voice utterance scenes such as a scene in which a character such as Avata reads out a manuscript (text) such as news or a weather forecast. Therefore, a case where the present technology is applied to a news program production system will be described below as a second embodiment.

(Example of news program production system)
FIG. 11 shows an example of a news program production system to which the present technology is applied.

In FIG. 11, the information processing device 10 is configured as a part of a news program production system. For example, in a large-scale vision installed in the city or in front of a station, a female character can send text information of news and weather forecast. , Read aloud by voice synthesis.

At this time, for example, at the timing between news or at the end of the program, the female character applies the present technology to read out the text information converted into the 75-tone.

In this way, by allowing more human speech to be performed by the Shichigocho constraint, for example, by directing a blurred character and impressing the user that there is an unexpected side, a woman's It is possible to increase the number of users who are interested in the character, and as a result, increase the number of viewers of the news program.

Here, as user feedback information when this technology is applied to a news program production system, for example, an image captured by a camera 111 provided in the information processing device 10 is analyzed and installed in a city or the like. The score for each candidate generation pattern can be calculated from the facial expressions of passers-by who are watching the large-scale vision.

Also here, by using the service that provides the API for recognizing faces and images described above and sending captured images including the faces of a large number of passers-by, for example, facial expressions of passers-by such as joy and surprise. You can get information about and feelings.

For example, if it is recognized that many passers-by have a happy look, the score will be added, while many passers-by will have a boring look, for example. If recognized, the score is subtracted.

For example, when the user is watching the news program at home using a TV receiver or a smartphone, the user can operate the remote controller or the application started on the smartphone. Voting (eg, good or bad) for the female character's 75-tone utterances may be made.

Also, here, as the context information, in addition to the time information and the position information, for example, information obtained from the analysis result of the captured image captured by the camera 111 provided in the information processing apparatus 10 can be used.

In the news program production system to which this technology is applied, it is not necessary to recognize the user's utterance or generate a response. Therefore, the functions of the information processing unit 150 (speech recognition unit 151 to the user) shown in FIG. Among the feedback collecting units 158), for example, it may be configured to include functions corresponding to the context acquisition unit 155, the 75-tone conversion unit 156, the speech synthesis unit 157, and the user feedback collecting unit 158.

In addition, in the above explanation, voice utterances by female characters have been described as an example, but not limited to voice utterances, the contents of news and weather forecasts, and the contents of Shichigocho utterances can be used as text information for large-scale visions, displays, etc. It may be displayed on the screen.

The second embodiment has been described above.

In this second embodiment, text information intended to be read aloud by voice synthesis (for example, utterances between news) is converted into 75-tone and output, so that it is more human-like due to the 75-tone constraint. It makes it possible to make utterances (for example, utterances that give a feeling of familiarity). Further, also in the second embodiment, by using the user feedback information and the context information, it is possible to more appropriately convert to the 75-tone when performing the 75-tone constraint.

<3. Third Embodiment>

Further, as a configuration other than the dialogue system, the news program production system has been described in the above-described second embodiment, but as another configuration, for example, it may be applied to a scene in which a commercial is played in digital signage. Can be done. Therefore, a case where the present technology is applied to a digital signage system will be described below as a third embodiment.

(Example of digital signage system)
FIG. 12 shows an example of a digital signage system to which the present technology is applied.

In FIG. 12, the information processing device 10 is configured as a part of a digital signage system, and is installed indoors such as a station or a commercial facility to display information such as advertisements and guidance.

At this time, for example, at a timing such as between CMs, the present technology is applied to read out the text information converted into the Shichigocho. More specifically, as shown in FIG. 12, a passerby walking near the digital signage will be interested between the commercial of the car flowing at a certain time and the commercial of the detergent flowing at the subsequent time. For example, the voice of the 75-tone utterance with blurred content is output.

In this way, by allowing more human speech to be made by the Shichigocho constraint, for example, it is possible to impress passers-by with a product that is the target of a commercial and make the product interested. ..

Although the voice utterance has been described here as an example, not only the voice utterance but also the text information converted into the 75-tone may be displayed on the screen of the digital signage.

Here, the user feedback information when this technology is applied to a digital signage system is, for example, analyzed by an image captured by a camera 111 provided in the information processing apparatus 10 and used indoors such as a station. The score for each candidate generation pattern can be calculated from the facial expressions of passers-by who are watching the installed digital signage.

Also here, by using the service that provides the API for recognizing faces and images described above and sending captured images including the faces of a large number of passers-by, for example, facial expressions of passers-by such as joy and surprise. You can get information about and feelings.

For example, if it is recognized that many passers-by have a happy look, the score will be added, while many passers-by will have a boring look, for example. If recognized, the score is subtracted.

Also, here, as the context information, in addition to the time information and the position information, for example, information obtained from the analysis result of the captured image captured by the camera 111 provided in the information processing apparatus 10 can be used.

Even in the digital signage system to which this technology is applied, it is not necessary to recognize the user's utterance or generate a response, so that the function of the information processing unit 150 (speech recognition unit 151 to the user) shown in FIG. Among the feedback collecting units 158), for example, it may be configured to include functions corresponding to the context acquisition unit 155, the 75-tone conversion unit 156, the speech synthesis unit 157, and the user feedback collecting unit 158.

The third embodiment has been described above.

In this third embodiment, text information intended to be read aloud by voice synthesis (for example, an utterance output between CMs) is converted into 75-tone and output, thereby being subject to the 75-tone constraint. , Makes it possible to make more human-like utterances (for example, utterances that make the product interesting). Further, also in the third embodiment, by using the user feedback information and the context information, it is possible to more appropriately convert to the 75-tone when performing the 75-tone constraint.

<4. Modification example>

(Example of interactive system configuration)
In the above description, the case where the dialogue service is realized by executing the dialogue processing by the information processing device 10 (information processing unit 150) has been illustrated, but as a configuration for realizing such a dialogue service, For example, the configuration shown in FIG. 13 can be adopted.

In FIG. 13, the dialogue system 1 is installed on the local side such as a user's house and functions as a user interface of the dialogue service, and is installed on the cloud side such as a data center to realize the dialogue function of the dialogue service. It is composed of a server 20 that performs processing for the purpose of processing.

In the dialogue system 1, the information processing device 10 and the server 20 are connected to each other via the Internet 30.

The information processing device 10 is, for example, a speaker that can be connected to a network such as a home LAN, and is also called a so-called smart speaker or a home agent. In addition to playing music, this type of speaker can perform voice dialogue with a user, voice operation on equipment such as lighting equipment and air conditioning equipment, and the like.

The information processing device 10 is not limited to a speaker, but is configured as, for example, a mobile device such as a smartphone or a mobile phone, or an electronic device such as a tablet computer, a personal computer, a television receiver, or a game machine. May be good.

The information processing device 10 can provide a dialogue service (user interface) to a user by cooperating with the server 20 via the Internet 30.

That is, the information processing device 10 picks up the voice (user's utterance) emitted from the user and transmits the voice data to the server 20 via the Internet 30. Further, the information processing device 10 receives the processing data transmitted from the server 20 via the Internet, and outputs the voice corresponding to the processing data.

The server 20 is a server that provides a cloud-based dialogue service. The server 20 performs voice recognition processing for converting voice data transmitted from the information processing device 10 via the Internet 30 into text information. Further, the server 20 performs processing such as interactive processing according to the user's intention on the text information, and transmits the processing data obtained as a result of the processing to the information processing device 10 via the Internet 30.

In the configuration composed of the local side and the cloud side shown in FIG. 13, a system for generating a response to a user's speech was described as the dialogue system 1, but as described above, a news program production system, a digital signage system, etc. It may be configured as a system that reads out the text information of the above by voice synthesis.

As described above, in the above description, the functions of the information processing unit 150 of FIG. 2 (voice recognition unit 151 to the user feedback collecting unit 158) have been described as being incorporated in the information processing device 10, but the information processing unit of FIG. 2 has been described. The 150 functions may be incorporated as the functions of the server 20. That is, each of the functions of the information processing unit 150 (voice recognition unit 151 to the user feedback collecting unit 158) of FIG. 2 may be incorporated in any of the information processing device 10 and the server 20.

For example, among the functions of the information processing unit 150 in FIG. 2, the voice recognition unit 151 to the response generation unit 154 are incorporated in the server 20 on the cloud side, and the context acquisition unit 155 to the user feedback collection unit 158 provide information on the local side. It can be incorporated into the processing device 10.

Regardless of which configuration is adopted, the database such as the user feedback information DB 202, the ending list 211, the meaning-invariant word list 212, the onomatopoeia 213, and the synonym dictionary 214 shall be managed by the server 20 on the Internet 30. Can be done.

As described above, the present technology can have a cloud computing configuration in which one function is shared by a plurality of devices via a network and jointly processed.

(Computer configuration)
The series of processes described above (for example, the 75-tone conversion output process shown in FIGS. 3 to 4) can be executed by hardware or software. As shown in the configurations of FIGS. 1 and 2, when a series of processes are executed by software, a program constituting the software is installed in the information processing apparatus 10 (computer).

The program executed by the information processing device 10 (CPU101) of FIG. 1 can be recorded and provided on a removable recording medium such as a package medium. The removable recording medium is composed of, for example, a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like.

In addition to the Internet 30, the program can be provided via a wired or wireless transmission medium such as a local area network or digital satellite broadcasting.

In the information processing device 10 of FIG. 1, the program can be installed in a recording device such as a hard disk 105 via the information access unit 104 by mounting a removable recording medium on a drive (not shown).

Further, the program can be received by the communication I / F 116 via a wired or wireless transmission medium and installed in a recording device such as a hard disk 105. In addition, the program can be installed in advance in the ROM 102, the recording device, or the like.

Here, in the present specification, the processing performed by the information processing apparatus 10 (CPU101) of FIG. 1 according to the program does not necessarily have to be performed in chronological order in the order described as the flowcharts shown in FIGS. 3 to 4. Absent. That is, the processing performed by the information processing apparatus 10 (CPU101) of FIG. 1 according to the program includes processing executed in parallel or individually (for example, parallel processing or processing by an object).

Further, the program may be processed by one computer (processor) or may be distributed by a plurality of computers. That is, each step of the 75-tone conversion output process shown in FIGS. 3 to 4 can be executed by one device or shared by a plurality of devices. Further, when a plurality of processes are included in one step, the plurality of processes included in the one step can be executed by one device or shared by a plurality of devices.

The embodiment of the present technology is not limited to the above-described embodiment, and various changes can be made without departing from the gist of the present technology.

In addition, the present technology can have the following configurations.

(1)
An information processing device equipped with a processing unit that converts the input text information into 75-tone and outputs it.
(2)
The information processing device according to (1) above, wherein the processing unit converts the text information into 75-tone based on the user feedback information obtained by the feedback from the user.
(3)
The information processing device according to (1) or (2) above, wherein the processing unit converts the text information into a seven-five tone based on context information for the text information.
(4)
The processing unit
With reference to the user feedback information, the 75-tone conversion process suitable for the context information is selected.
The information processing device according to (3) above, which executes the selected 75-tone conversion process on the text information.
(5)
The user feedback information includes information obtained by scoring the past environmental information and the reaction of the user for each candidate generation pattern that generates the candidates of the Shichigocho.
The context information includes current environmental information.
The information processing unit according to (4) above selects a candidate generation pattern in which the score of the past environmental information that is the same as or similar to the current environmental information is equal to or higher than the threshold value among the past environmental information. apparatus.
(6)
The information processing apparatus according to (5) above, wherein the processing unit randomly selects one candidate generation pattern when a plurality of candidate generation patterns can be selected.
(7)
The candidate generation pattern is
Without particles, to generate candidates that will be in the 75th tone without particles,
Remove unnecessary parts to generate candidates that become 75-tone by removing semantically unnecessary parts,
Add endings to generate candidates for Shichigocho
Add meaning-invariant words to generate candidates that become 75-tone by adding words that do not change in meaning,
Repeated addition to generate candidates that become 75-tone by repeating a certain word,
Add onomatopoeia to generate candidates for 75-tone, and add onomatopoeia to generate candidates for 75-tone, and add synonyms to generate one-75-tone candidate or multiple 75-tone candidates. The information processing apparatus according to (5) or (6) above, which includes a combination of generations.
(8)
The information processing device according to any one of (5) to (7) above, wherein the context information includes at least one piece of information indicating a time zone, a place, a speaker, an attendant, and an atmosphere of the place.
(9)
The information processing device according to any one of (1) to (8) above, wherein the text information is a response to a user's utterance.
(10)
The information processing device according to any one of (1) to (8) above, wherein the text information is information intended to be read aloud by voice synthesis.
(11)
In the information processing method of the information processing device
The information processing device
An information processing method that converts the input text information into 75-tone and outputs it.

1 Dialogue system, 10 Information processing device, 20 Server, 30 Internet, 100 Control unit, 101 CPU, 102 ROM, 103 RAM, 150 Information processing unit, 151 Speech recognition unit, 152 Speech intention understanding unit, 153 Application / service execution unit , 154 Response generation unit, 155 Context acquisition unit, 156 75-tone conversion unit, 157 Speech synthesis unit, 158 User feedback collection unit, 201 Context information DB, 202 User feedback information DB, 211 Ending list, 212 Semantic invariant list, 213 Onomatope List, 214 Synonyms list

Claims (11)

An information processing device equipped with a processing unit that converts the input text information into 75-tone and outputs it. The information processing device according to claim 1, wherein the processing unit converts the text information into 75-tone based on the user feedback information obtained by the feedback from the user. The information processing apparatus according to claim 2, wherein the processing unit converts the text information into Shichigo-Chine based on context information for the text information. The processing unit
With reference to the user feedback information, the 75-tone conversion process suitable for the context information is selected.
The information processing device according to claim 3, which executes the selected 75-tone conversion process on the text information. The user feedback information includes information obtained by scoring the past environmental information and the reaction of the user for each candidate generation pattern that generates the candidates of the Shichigocho.
The context information includes current environmental information.
The information processing apparatus according to claim 4, wherein the processing unit selects a candidate generation pattern in which the score of the past environmental information that is the same as or similar to the current environmental information is equal to or higher than the threshold value among the past environmental information. .. The information processing apparatus according to claim 5, wherein the processing unit randomly selects one candidate generation pattern when a plurality of candidate generation patterns can be selected. The candidate generation pattern is
Without particles, to generate candidates that will be in the 75th tone without particles,
Remove unnecessary parts to generate candidates that become 75-tone by removing semantically unnecessary parts,
Add endings to generate candidates for Shichigocho
Add meaning-invariant words to generate candidates that become 75-tone by adding words that do not change in meaning,
Repeated addition to generate candidates that become 75-tone by repeating a certain word,
Add onomatopoeia to generate candidates for 75-tone, and add onomatopoeia to generate candidates for 75-tone, and add synonyms to generate one-75-tone candidate or multiple 75-tone candidates. The information processing apparatus according to claim 5, which includes a combination of generations. The information processing device according to claim 5, wherein the context information includes at least one piece of information indicating a time zone, a place, a speaker, an attendant, and an atmosphere of the place. The information processing device according to claim 1, wherein the text information is a response to a user's utterance. The information processing device according to claim 1, wherein the text information is information intended to be read aloud by voice synthesis. In the information processing method of the information processing device
The information processing device
An information processing method that converts the input text information into 75-tone and outputs it.

Images