JP2002041411A - Text-reading robot, its control method and recording medium recorded with program for controlling text recording robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a text-reading robot, capable of performing a text reading advanced in an emotional information transmission action without increasing capacity, to provide a method for controlling the robot and to provide a recording medium, in which a program for controlling the robot is recorded. SOLUTION: The text-reading robot for reading a text stored in a memory of a personal computer(PC), connectable to a computer system, analyzes the contents of sentences by morpheme analysis for dividing the text into words, extracts information concerned with emotions and generating voice, gesture or an effective sound corresponding to the extracted emotion information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text-to-speech robot for automatically reading text such as e-mails and homepages distributed on the Internet by voice synthesis, a control method thereof, and a recording medium on which a text-to-speech robot control program is recorded.

[0002]

2. Description of the Related Art There are existing robots that automatically read an e-mail by voice synthesis. However, there is a problem that emotional expression is poor, difficult to understand, and tired because only synthetic sounds are generated. Also, JP-A-11-3
In a method of storing an operation program and phoneme information in an e-mail text in advance, transmitting an e-mail, and executing a corresponding program at a destination as in Japanese Patent No. 277872 and Japanese Patent Application Laid-Open No. 11-65964, an operation is performed on an e-mail. Because the program and phoneme information are attached, the capacity increases,
Alternatively, there is a problem that the same mailer needs to be installed at the destination of the child mail.

[0003]

The conventional e-mail reading robot has problems such as poor emotional expression and large capacity.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and a text-to-speech robot capable of reading out a text-to-speech evolved into an emotional information transmission action accompanied by emotional expression without increasing the capacity, a control method thereof, and a control method thereof. An object of the present invention is to provide a recording medium on which a text-to-speech robot control program is recorded.

[0005]

In order to achieve the above object, the present invention provides a PC which can be connected to a computer system.
A text-to-speech robot that reads text stored in a memory in a (personal computer), wherein the text-to-speech robot analyzes sentence contents by morphological analysis that divides the text into words, and extracts emotion-related information,
It is characterized by comprising voice / motion expression means for expressing voices, gestures and sound effects corresponding to the emotion information extracted by the emotion information extraction means.

In the text-reading robot, when the text is analyzed by the emotion information extracting means, pictographic information in the text is extracted, and the intention of the pictographic information and the text in the text are extracted by the voice / motion expressing means. Voices, gestures, and sound effects according to the location of the character.

The present invention is also characterized in that in the text-to-speech robot, when the gesture is expressed by the voice / motion expression means, the size of opening / closing of the robot's mouth changes according to the speaker output. is there.

[0008] Further, the present invention provides the text-to-speech robot, wherein the text is text transmitted / received as an e-mail, and stores sender data from the received mail and the number of replies to the received mail. The feature is that the voice reading and the speech of the mail change for many specific mail senders.

Further, according to the present invention, in the text-to-speech robot, the text is text transmitted / received as an e-mail, and the content of the sentence is analyzed by morphological analysis that divides the e-mail text into words to obtain information related to emotion. Then, the intimacy between the e-mail sender and the e-mail is analyzed, and the read-out voice and dialogue of the e-mail are changed according to the intimacy.

The present invention also relates to a method of controlling a text-to-speech robot that reads text stored in a memory in a PC connectable to a computer system.
The sentence is analyzed by morphological analysis to divide the text into words, and an emotion information extraction step of extracting information related to emotion, and a voice, gesture, and sound effect corresponding to the emotion information extracted in the emotion information extraction step are generated. Voice / motion expression step.

Further, the present invention is a computer-readable recording medium recording a control program for a text-to-speech robot that reads text stored in a memory in a PC connectable to a computer system,
The sentence is analyzed by morphological analysis that divides the text into words, and an emotion information extraction procedure for extracting information related to emotions, and voices, gestures, and sound effects corresponding to the emotion information extracted in the emotion information extraction procedure are expressed. This is for causing a computer to execute a voice / motion expression procedure.

According to the present invention, a function of extracting an expression relating to an emotion from the contents of a text and adding a function of expressing a line or a gesture appropriate to the emotion during the reading is added to the normal reading function. In addition, attention is paid to the fact that a pictogram generally called a smiley is an emotional expression of an e-mail sender, and a robot action is performed at the timing when the pictogram exists, thereby aiming for a more timely effect.

[0013]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a functional block diagram of a robot when reading out text of an electronic mail according to an embodiment of the present invention. In the figure, 11 is an emotion information extraction device, 1
2 is a mail sender DB (database), 13 is a serif DB, 14 is a voice synthesizer, 15 is a voice controller, 16
Is a motion control device, 17 is a gesture DB, 18 is a drive mechanism, and 19 is a speaker.

A mail text is given as an input, and the text is basically uttered from a speaker 19 through the voice synthesizer 14. On the other hand, emotion expressions such as emotions and emotions of the mail content are extracted. The extracted emotion information such as emotions and emotions is transmitted to the exercise control device 16 and the voice control device 15. In each device, effective voices and gestures suitable for the emotion information are selected from the respective DBs and transmitted to the respective output devices.

The gesture is expressed by the drive mechanism 18. At the same time, words suitable for sound effects and emotions are uttered from the speaker 19.

Further, when the gesture is expressed, the size of the opening and closing of the mouth of the robot changes according to the speaker output. That is, an audio signal is input from the speaker 19 to the exercise control device 16, the loudness of the sound input to the speaker 19, that is, the amplitude value of the electric signal is obtained, and the opening and closing of the mouth is driven according to the loudness of the sound. The pulse width of the input PWM signal to the motor to be controlled (servo motor controllable by the PWM signal) is changed.

Specifically, the motor is controlled by, for example, the following processing.

(1) Vsp: electric signal value to the speaker (mv) αv-max: maximum input signal value to the speaker (mv) αv-min: minimum input signal value to the speaker (mv) αv-min≪αv −low≪αv-high≪αv-
max (2) Pw: pulse width of the PWM signal (μsec) αp-max: effective maximum pulse width (μsec) αp-min: effective minimum pulse width (μsec) αp-min≪αp-low≪αp-high≪αp-
max (3) When the sound is stationary αv-low <= Vsp <= αv-high When the pulse width of the PWM signal to be output is αp-low <= Pw <= αp-high (4) When the sound is loud In the case of αv-high <Vsp <= αv-max, the pulse width of the PWM signal to be output is αp-min <= Pw <= αp-max (5) When the sound is low αv-min <= Vsp <αv- In the case of low, the pulse width of the PWM signal to be output is αv−low≪Pw≪αv−high Next, each module in FIG. 1 will be described.

[Emotion Information Extraction Apparatus 11] Morphological analysis that divides e-mail text into words and tags parts of speech (56th Annual Convention of IPSJ: "E-mail ranking based on information extraction and user's action history") "Hasegawa, Takagi, pp. 2-259-260, 1998 IPSJ research report Natural language processing:" Extraction of schedule information in e-mail communication ", Hasegawa, Takagi, 123-10, 1998:" Support Vector Machi
Text classification by ne ", Taira, Mukouchi, Haruno, 128-2
4 IPSJ 56th National Convention: "Japanese Morphological Analysis System Considering Conservativeness", Fuchi, Matsuoka, Takagi, 11
7-9, 1997: "A method for speeding up the search for intermediate partial matches using morphological analysis", Oku, Noda, Hayashi, 121-9, 19
97), the contents of the sentence are analyzed, and tag information relating to emotions is obtained (see the example below). The output information is an information tag, a text file containing the text information of the word serving as a clue of the tag, and an attribute of the mail based on the calculation of the appearance frequency of the information tag.

<Output information> E-mail attribute Pleasure tag ＠Pleasure Clued word ｛Waii｝ <Example of morphological analysis> (input mail text) I'm sorry. Miho. Invite a home party at Satchan's house. This time, bring your own specialties.

(After processing) {Greetings} Excuse me.

It is Miho.

Inviting a “fun” home party at Satchan's house.

[0025] This time, let's bring in your own specialty dish.

From “＠” to “｛” → tag information “｛” to “｝” → word as a clue of tag.
Prosodic parameters (pitch frequency, power,
Pre-stored as a voice file that modulates the phonetic duration and adds inflection.

<Examples of emotion information tags and clues of tags> Tags: joy Words: success, congratulations, (＾＾) V, etc. Tags: anger Words: phrases, effort, (｀ ′), etc. Tags: Sorrow Words: failure, pain, (T_T), etc. Tags: fun Words: lol, pleasure, p (＾＾) q, etc. Tags: Chen Xie Words: sorry, sorry, m (_) m, etc. Tags: surprise word : Eh, ah, (* _ *), etc. Tag: Emphasis Word:! , Urgent, o (＾ o ＾) o, etc. <Example of attribute determination of mail content> Number of appearances of tags in the text Tag 1 times Appreciation 4 Sorrow 2 Surprise 1 ↓ ↓ Attribute of mail = Mail appreciation [Mail sender DB12] Stores the sender list and the frequency of mail reception with the sender from the mail header information. Also,
The relationship between the sender and the recipient is extracted based on the attribute information of the mail from the emotion information extraction device and the number of times the mail has been transmitted and received. Outputs the sender name, Subject, and the relationship with the sender of the mail. FIG. 2 is an explanatory diagram showing a record table of the mail sender DB 12 according to the embodiment of the present invention.

[Serif DB13] Prosody parameters (pitch frequency, power, phoneme duration) are changed for words that are clues to the emotion information tag, and the inflection voice file and the contents of the mail are uttered. The audio file of the comment sentence to be added is stored before and after. The comment text is
Polite sentences, broken sentences, sentences with intonation, and sentences without intonation are stored as audio files.

<Example of a comment sentence> XX "E-mail from Chan!" XX "That's it!" XX "An e-mail has arrived from XXX." XX "It's about."Sender's name XX → Contents of subject [Speech synthesizer 14] Converts a text file containing text information of information tags and words used as clues of tags into speech files. At this time, the voice file of the word that became the clue of the tag is obtained from the dialogue DB 13 and synthesized. Further, the prosodic parameters (pitch frequency, power, phoneme duration) are changed based on the attribute of the mail acquired from the emotion information extraction device 11. Regarding the generation of the comment sentence, reference is made to the relationship with the sender, the sender name, and the subject from the mail sender DB 12, and the dialog DB
13 is synthesized with the voice file of the dialogue acquired from the dialogue. When selecting a line from the line DB, a sentence with a broken intonation is selected as the intimacy increases, in consideration of the relationship with the sender.

As the output information, a voice file of the generated mail text and comment text is output to the voice control device 15. Also, the motion control device 16 has information tags, information on the words that have become clues to the information tags, reproduction time information of the generated audio files, and time information on the positions where the words that have become clues to the information tags are embedded. , And information including the reproduction time of the word.

<Output Information to Exercise Control Device 16> Total number of tags: 11 (pieces) Total playback time of audio files (ms): 3600 First tag information: ＠Chan Xie Clue words: ｛Sorry め Word playback Start time (ms): 600 (ms from beginning of synthesized speech) End time of word playback (ms): 900 (ms from beginning of synthesized speech) Word playback time (ms): 300 (ms) Until next tag Time: 500 (ms) Second tag information: ＠Greeting Clues: ｛Good morning｝ Start time of word playback (ms): 1400 (ms from beginning of synthesized speech) End time of word playback (ms): 2200 (Ms from head of synthesized speech) Word playback time (ms): 800 (ms) Time to next tag: 400 (ms) Third tag information:: 11th Tag information: [Gesture-DB17] Stores time-series information of command values (motor number, position, speed, time) of the joint driving motor of the robot, which are associated with the emotion information tag.

[Motion control device 16] Based on the output information from the speech synthesis device 14, the gesture DB 17
The command value of the motor for driving the joint of the robot corresponding to the emotion information tag is called and transmitted to the drive mechanism 18 as a PWM signal. The drive timing is in accordance with the reproduction time information of the audio file, the time information of the position where the word serving as the clue of the information tag is embedded, and the reproduction time of the word, which are the output information of the voice synthesis device 14.

[Sound control device 15] The sound file obtained from the sound synthesis device 14 is reproduced by the speaker 19.

The method for controlling a text-to-speech robot according to the present invention is, more specifically, a control program for a text-to-speech robot that is recorded in advance on a computer-readable recording medium by a computer such as a personal computer (PC). It is executed based on.

That is, a computer-readable recording medium recording a control program of a text-to-speech robot that reads text stored in a memory in a PC connectable to a computer system, wherein the text is divided into words. A sentence information extraction procedure for analyzing the contents of a sentence and extracting information relating to emotions, and a voice / motion expression procedure for expressing voices, gestures, and sound effects corresponding to the emotion information extracted in the emotion information extraction procedure To run.

[0036]

As described above, according to the present invention, the robot assists the expression of the emotion with sound and action in accordance with the emotions such as emotions, emotions, and so on related to the contents of the mail. As a result, e-mail reading, which tends to be tasteless and dry, has evolved into an emotional communication with emotional expression, enabling detailed distribution of e-mail information that cannot be expressed using only letters and their synthesized sounds. Becomes

Further, the timing of the robot operation manifestation is
By associating with the position of the pictogram in the mail, it is possible to express emotions by reading the writer's feelings more deeply.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing a text-to-speech robot when reading out text of an e-mail according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a record table of a mail sender DB according to the embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Emotion information extraction device 12 Mail sender DB (database) 13 Dialog DB 14 Voice synthesis device 15 Voice control device 16 Motion control device 17 Gesture DB 18 Drive mechanism 19 Speaker

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Takao Kakizaki 2-3-1 Otemachi, Chiyoda-ku, Tokyo Within Nippon Telegraph and Telephone Corporation (72) Inventor Tamotsu Machino 2-3-3 Otemachi, Chiyoda-ku, Tokyo No. 1 F-term in Nippon Telegraph and Telephone Corporation (reference) 5D045 AA07 AB11

Claims (7)

[Claims] 1. A PC connectable to a computer system
A text-to-speech robot that reads text stored on a memory in a text-to-speech robot, wherein the sentiment information is analyzed by morphological analysis that divides the text into words, and emotion-related information is extracted from the sentence. A text-to-speech robot comprising: voice / motion expression means for expressing voices, gestures, and sound effects corresponding to the emotion information extracted by the extraction means. 2. The text-to-speech robot according to claim 1, wherein when the text is analyzed by the emotion information extraction means, pictograph information in the text is extracted, and the intention of the pictograph information and the text are extracted by voice / motion expression means. A text-to-speech robot that produces voices, gestures, and sound effects according to its location in the space. 3. The text-to-speech robot according to claim 1, wherein when the gesture is expressed by the voice / motion expression means, the size of the opening / closing of the mouth of the robot changes according to the speaker output. A reading robot. 4. The text-to-speech robot according to claim 1, wherein the text is text transmitted / received as an e-mail, stores sender data from a received mail, and the number of replies to the received mail, and sends / receives the e-mail. For specific email senders who send many emails,
A text-to-speech robot characterized by changing dialogue. 5. The text-to-speech robot according to claim 1, wherein the text is text transmitted / received as an e-mail, and the content of the sentence is analyzed by morphological analysis that divides the e-mail text into words, and information relating to emotions is analyzed. A text-to-speech robot that obtains and analyzes the intimacy between an e-mail sender and a recipient, and changes the voice and speech of an e-mail according to the intimacy. 6. A PC connectable to a computer system
A text-to-speech robot control method for reading a text stored on a memory in a text-to-speech robot, wherein the sentence is analyzed by morphological analysis that divides the text into words, and an emotion information extraction step of extracting information related to emotion is provided. A method for controlling a text-to-speech robot, comprising: a voice / motion expression step of expressing voices, gestures, and sound effects corresponding to the emotion information extracted in the emotion information extraction step. 7. A PC connectable to a computer system
A control program for a text-to-speech robot that reads text stored on a memory in the memory, wherein the content of the sentence is analyzed by morphological analysis for dividing the text into words, and emotion-related information is extracted. A recording medium for recording a text-to-speech robot control program for causing a computer to execute a voice / gesture / sound effect expressing procedure corresponding to the emotion information extracted in the emotion information extracting step.