JP2004341293A - Device and method for converting speech into character data, and program and interaction device therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide technology which makes it possible to expect accurate conversion of speech into character data corresponding to the speech that a user utters. <P>SOLUTION: Disclosed is a device which converts the speech that the user utters into the character data. This character data converting device is equipped with a means of inputting the speech, a means of converting the inputted speech into the character data, a correcting means of enabling the user to correct the character data converted by the converting means, a means of storing the character data converted by the converting means and corrected character data of the character data while making them correspond to each other, and a means of specifying the "corrected character data" by retrieval from the storage means making the character data converted by the converting means as a key. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for converting a human voice into character data.
[0002]
2. Description of the Related Art In recent years, devices that can interact with humans (users) and devices that operate in response to voices uttered by humans (voice-compatible robots) have been developed. These devices input a voice uttered by a human, convert the input voice into character data, and execute a predetermined process according to the converted character data. For example, some interactive devices execute a process of outputting a response corresponding to the converted character data. Some voice-enabled robots execute an operation corresponding to the converted character data.
The above-described interactive device, voice-capable robot, and the like need to accurately convert the voice uttered by the user into character data. Patent Literature 1 discloses one technique for converting a voice uttered by a user into character data. According to the technique of Patent Literature 1, it is possible to convert words that the user utters slowly while separating sounds into character data.
[0003]
[Patent Document 1]
JP-A-2000-242295
[0004]
The technique disclosed in Patent Document 1 is based on the premise that each sound uttered by a user can be accurately converted into character data. However, in practice, it may not be converted into character data corresponding to the voice uttered by the user. For example, even though the user utters the word X, it may be converted to character data of Y. One of the reasons is that it is difficult to generalize the method of converting the input voice to character data so that it can be applied to all users because there are individual differences in voice quality and pronunciation. It is believed that there is.
[0005]
The present invention has been made in view of the above circumstances, and has as its object to provide a technology that can be expected to be accurately converted to character data corresponding to a voice uttered by a user.
[0006]
Means for Solving the Problems, Functions and Effects The invention of claim 1 created to solve the above-mentioned problem is an apparatus for converting a voice uttered by a user into character data. The character data conversion device includes means for inputting a voice, means for converting the input voice to character data, correction means for allowing a user to correct the character data converted by the conversion means, and conversion means for the user. Means for storing the character data converted by the above and character data corrected from the character data in association with each other, and searching the storage means using the character data converted by the conversion means as a key, Means for specifying “character data obtained”.
According to the character data conversion device of the first aspect, when the voice uttered by the user is converted into erroneous character data by the conversion means, the erroneously converted character data can be corrected by the correction means. As this correction method, any method may be adopted, for example, the correction may be made by the user inputting character data using a keyboard, or the user may make a correction by re-uttering. . For example, if the conversion unit erroneously converts the character data to Y when the user utters X, the user can correct the converted character data Y to X. At this time, information associating X and Y is stored in the storage means. By storing this information, if X uttered by the user is incorrectly converted to Y again by the conversion unit, X can be specified by searching the storage unit using that Y as a key. . According to the present invention, even when the conversion unit converts the character data to a character data different from the voice uttered by the user, the error can be covered and the character data corresponding to the voice uttered by the user can be specified.
[0007]
The following phenomena may occur due to differences in voice quality, pronunciation, etc. among users. For example, the conversion unit may convert the utterance of the user A to Y into the character data of X while the conversion unit converts the utterance of the user B to the character data of X. When both are corrected, the storage means stores X (converted character data) and Y (corrected character data) in association with each other, and stores X (converted character data) and Z (converted character data). (Corrected character data) are stored in association with each other. In this case, when the conversion unit converts the utterance of the user A to Y again to character data of X, there is a possibility that Z is specified. In order to deal with such a problem, the above-described character data conversion device may have the following configuration. That is, a user specifying means for specifying a user is added to the above character data conversion device. In this case, the storage unit stores the user specified by the user specifying unit, the character data converted by the conversion unit, and the character data corrected from the character data in association with each other. The specifying unit searches the storage unit using the user specified by the user specifying unit and the character data converted by the conversion unit as a key, and specifies “corrected character data”.
With such a configuration, in the above example, user A and X (converted character data) and Y (corrected character data) can be stored in association with each other, and user B and X (converted character data) can be stored. Character data) and Z (corrected character data) can be stored in association with each other. In this case, when the conversion unit converts the utterance of the user A to Y again to character data of X, the storage unit is searched using the users A and X as keys to specify Y. According to the present invention, it can be expected that accurate conversion to character data corresponding to the voice uttered by each user can be expected.
[0008]
The storage means may be capable of storing a plurality of "corrected character data" in a weighted manner in association with the character data converted by the conversion means. In this case, when a plurality of "corrected character data" are stored in association with the "character data converted by the converting means" as a key, the specifying means sets the weight to be the maximum. "Corrected character data" is specified.
The above-mentioned “weight” is, for example, the “corrected character data”, which is associated with the “character data converted by the conversion unit”, and the “corrected character data” for the total number of conversions. Means the ratio of the number of times the user has uttered the “character data”.
The present invention will be described by taking the following events as examples. (1) Although the user utters X, the conversion means converts the Z to Z, and the user corrects Z to X. (2) Even though the user uttered X, the conversion means converted again to Z, and the storage means was searched from Z to identify X. (3) Although the user utters Y, the conversion means converts it to Z, and the user corrects Z to Y.
When the above events (1) to (3) occur, the storage means associates Z (converted character data), X (corrected character data), and 2/3 (weight) with each other. Z (converted character data), Y (corrected character data), and 1/3 are stored in association with each other. The denominator of the weight here is the total number of times that the conversion unit has converted to Z. The numerator of the weight is X or Y and the number of times the user has spoken. In this state, assume that the following event occurs. (4) Although the user uttered X, the conversion means converted it to Z. When the event (4) occurs, the specifying unit searches the storage unit using Z as a key, and specifies X having a large weight. As described above, in the present invention, in order to select “corrected character data” which is likely to be erroneously converted, even if one character data is corrected to a plurality of character data, the present invention can cope well. can do.
[0009]
The present invention can be embodied in a device that interacts with a user by outputting a response corresponding to a voice uttered by the user. This interactive device includes means for inputting voice, means for converting the input voice to character data, correction means for allowing a user to correct the character data converted by the conversion means, and conversion by the conversion means. A first storage unit that stores the corrected character data and the character data corrected from the character data in association with each other, and stores a plurality of character data and stores a reply for each character data. 2 storage means, and when the character data converted by the conversion means is not stored in the second storage means, the "character data converted by the conversion means" is used as a key in the first storage means. And a means for specifying “corrected character data” and the character data converted by the conversion means or the “corrected character data” specified by the specification means are stored in the second storage means. If it is, and means for voice output a response corresponding to the identified in the converted text data or the specific means by said conversion means "corrected character data".
According to this interactive device, it can be expected that a reply corresponding to the voice uttered by the user can be output even when the voice data is converted into character data different from the voice uttered by the user.
[0010]
Further, the present invention can be defined as a method in which a computer converts voice uttered by a user into character data. This character data conversion method includes a step of inputting a voice, a step of converting the input voice to character data, and a step of, when a user corrects the character data converted in the conversion step, performing the “ Storing the character data converted in the step and character data corrected from the character data in association with each other, and storing the character data converted in the conversion step as a key in the storage step. Searching the stored contents to specify the "corrected character data".
By using this method, it can be expected that the computer can accurately convert the voice uttered by the user into character data.
[0011]
Further, the present invention can be defined as a program for a computer to convert voice uttered by a user into character data. This program provides the following processing to the computer, namely, processing for inputting voice, processing for converting the input voice to character data, and processing when the user corrects the character data converted by the conversion processing. A process of storing the “character data converted by the conversion process” and the character data corrected from the character data in association with each other; and using the character data converted by the conversion process as a key, And a process of searching the storage contents stored in the storage process to specify “corrected character data”.
By using this program, it can be expected that the computer can accurately convert the voice uttered by the user into character data.
[0012]
DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention described in each of the above claims can be suitably implemented in the following modes.
(Mode 1) The character data conversion device has a microphone for inputting a voice (word) uttered by a user. The microphone converts the input sound wave into an electric signal. The conversion means converts the sound converted into an electric signal by the microphone into character data in a text format.
(Mode 2) The character data conversion device has a keyboard. The user can correct the character data converted by the conversion means by inputting characters on the keyboard.
(Mode 3) The above-mentioned second storage means stores a question for the user. Further, a plurality of answers to the question are stored, and a reply is stored for each answer.
(Mode 4) The user specifying means specifies a user by inputting a user name using a keyboard.
[0013]
Embodiment (First Embodiment) An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a schematic configuration of an interactive device (interactive computer) 10 according to the present embodiment. The interactive device 10 asks a question to a user, specifies an answer spoken by the user in response to the question, and replies to the specified answer.
The interactive device 10 includes a microphone 20, a voice identification unit 30, a control unit 40, a voice synthesis unit 50, a speaker 60, a display 70, a keyboard 80, a first database 90, a second database 100, a user name-user ID storage unit 110, and the like. Consists of
The microphone 20 inputs a word (voice wave) uttered by the user. Then, the input words are converted into electric signals and sent to the voice specifying unit 30.
The voice identification unit 30 converts the words converted into electric signals into character data in a text format. The voice specifying unit 30 sends the character data to the control unit 40.
The control unit 40 is connected with the voice identification unit 30, the voice synthesis unit 50, the display 70, the keyboard 80, the first database 90, the second database 100, and the user name-user ID storage unit 110. The control unit 40 inputs the character data sent from the voice specifying unit 30. Then, a process of outputting a response based on the input character data is executed. The control unit 40 also executes a process for displaying information on the display 70, a process for changing storage contents of the second database 100 and the user name-user ID storage unit 110, and the like. Each process executed by the control unit 40 will be described later in detail.
[0014]
The display 70 displays various images. The displayed image will be described later. The user can input character data using the keyboard 80. The character data input to the keyboard 80 is sent to the control unit 40.
The user name-user ID storage unit 110 stores a user name and an ID for specifying the user in association with each other. FIG. 2 shows an example of the storage contents of the user name-user ID storage unit 110. The user of the interactive device 10 registers his / her name in the interactive device 10 in advance. The registration of the user name is performed by inputting the name using the keyboard 80. The input user name is sent to the control unit 40. The control unit 40 acquires the user ID when the user name is sent, and stores the user ID in the user name-user ID storage unit 110 in association with the user name.
[0015]
In the first database 90, a question to the user, an answer group assumed as an answer to the question, and a response to each answer are stored in association with each other. FIG. 3 shows an example of the storage contents of the first database 90. For example, answers (apples, pears, tangerines, etc.) are stored in association with the question "What is your favorite fruit?" Further, a reply is stored for each answer of the user. For example, in the case of an apple, a reply that "apples are delicious" is stored.
[0016]
FIG. 4 shows an example of the storage contents of the second database 100. The second database 100 stores user IDs, specific words, true words, and weights in association with each other.
The “specific word” is a word (converted character data) specified by the voice specifying unit 30 and is not stored in the first database 90. For example, when converted to a word different from the word spoken by the user, the converted word may be a specific word. It should be noted that a “word” in this specification means a single character or a group of characters, and is a broader concept than a generally used concept of a word. In addition, as described in detail later, when the voice specifying unit 30 converts the erroneously converted word into a word different from the word uttered by the user, the user can correct the erroneously converted word (using the keyboard 80). You can enter words).
The “true word” means a word input by the user using the keyboard 80. For example, a word corrected by the user from a word converted by the voice specifying unit 30 into character data different from the word spoken by the user (in short, a word corrected from a word incorrectly specified by the voice specifying unit 30) means. In the example of FIG. 4, the true word “none” is stored in association with the specific word “palm”. This is because the voice identification unit 30 outputs “none” despite the user saying “none”. It is converted to "palm", which means that the "palm" has been corrected to "none". Further, the true word “palm” is stored in association with the specific word “palm”. This is because the voice specifying unit 30 converts the utterance of the user “palm” into “palm”. However, since “Pay” is not stored in the first database 90, when the user inputs a word using the keyboard 80, it means that “Pay” has been input.
“Weight” means the ratio of the number of times the user has uttered the true word associated with the weight to the total number of times the voice identification unit 30 has converted the specific word associated with the weight ( The unit is percent). In the example of FIG. 4, the specific word “palm”, the true word “none”, and the weight “80” are stored in association with each other. This means that the ratio of the number of times the user uttered “none” to the number of times is 80%. Also, a fraction is stored together with the weight (shown in parentheses in the weight column in FIG. 4), which indicates the total number of times that the voice identification unit 30 has converted the specific word associated with the weight. Is the denominator, and the number of times the user utters the true word associated with the weight is the numerator. In the example of FIG. 4, 16/20 is described in parentheses of the weights associated with the specific word “palm” and the true word “none”. This means that there are 20 conversions to "palm" and that the user uttered "none" 16 times out of the 20 times. In addition, although 1/1 is described in parentheses of the weights associated with the specific word “ashi” and the true word “none”, this is converted into the specific word “ashi” by the voice identification unit 30. Means that it was only once and was corrected to "none" (i.e., the user uttered "none").
[0017]
The speech synthesizer 50 shown in FIG. 1 converts the character data output from the controller 40 into an electric signal, and sends the electric signal to the speaker 60. Thereby, sound is output from the speaker 60.
[0018]
Next, a process executed by the interactive device 10 will be described. 5 to 7 show flowcharts of processing executed by the interactive device 10.
The user enters his name using the keyboard 80 prior to using the interactive device 10. The interactive device 10 reads the user name input by the user (Step S2). When the user name is read in step S2, the user ID is specified by searching the user name / user ID storage unit 110 from the read user name (step S4).
Subsequently, the interactive device 10 asks a question to the user (step S6). This processing is specifically executed as follows. First, the control unit 40 randomly selects one question from a plurality of questions stored in the first database 90. Then, the character data of the selected question is sent to the speech synthesizer 50. The voice synthesis unit 50 converts the transmitted character data (question) into an electric signal and transmits the electric signal to the speaker 60. Thereby, the question is output from the speaker 60 by voice.
[0019]
When a question is asked to the user, the user's answer to the question is specified (step S8). This processing is executed by the voice specifying unit 30 converting the user's answer input to the microphone 20 into character data. The voice specifying unit 30 sends the specified answer (converted character data) to the control unit 40. For example, the user may answer "Is an apple" or "I like an apple" to the question "What is your favorite fruit?" In some cases, only the word "apple" is answered. In this step S8, the control unit 40 removes phrases ("I like", "I like", "Is", etc.) attached before and after the answer word (apple in the above example) and returns the answer. Only words that are
Next, it is determined whether the specified answer is stored in the first database 90 (step S10). This process is executed by checking whether or not the answer group stored in association with the question selected in step S6 includes the answer specified in step S8. For example, since “pear” is stored in association with the question “What is your favorite fruit?”, If the answer specified in step S8 is “none”, “YES” is determined in step S10. You. On the other hand, since “palm” is not stored in association with the question “What is your favorite fruit?”, If the answer specified in step S8 is “palm”, the process proceeds to step S10. NO is determined.
If YES is determined in the step S10, a reply associated with the answer specified in the step S8 is output (step S12). This process is performed as follows. First, the first database 90 is searched from the answer specified in step S8, and a reply associated with the answer is specified. Next, the specified response is sent to the speech synthesizer 50. The voice synthesizer 50 converts the response of the character data into an electric signal, and sends the electric signal to the speaker 60, whereby the response is output as a voice.
[0020]
FIG. 8 shows a conversation example between the interactive device 10 and a user. When the processing is performed in the flow of steps S6, S8, S10, and S12 described above, the pattern is as shown in pattern 1 in FIG. In Pattern 1, when the user asks “What is your favorite fruit?” And the user answers “Pear”, the voice specifying unit 30 specifies “None”. He replied, "It's delicious."
[0021]
On the other hand, if NO is determined in the step S10, the process proceeds to a step S20 in FIG. In step S20, it is determined whether or not the answer specified in step S8 is stored as a specific word in the second database 100. This process is executed by searching the second database 100 from the answer specified in step S8. However, in this process, not all the specific words stored in the second database 100 are searched for, but the specific word group associated with the user ID specified in step S4 is included in the specific word group in step S8. It is determined whether there is an identified answer. For example, if the user ID is “XXX1” and the answer specified in step S8 is “palm”, the specific word “palm” is stored in association with the user ID “XXX1”. (See FIG. 4), YES is determined in step S20. If “YES” is determined in the step S20, the process proceeds to a step S22. If “NO” is determined in the step S20, the process proceeds to a step S40 in FIG.
In step S22, the true word having the largest weight is specified. For example, when the user ID is “XXX1” and the answer specified in step S8 is “palm”, the true word “none” having the largest weight is specified.
When the true word is specified in step S22, it is determined whether or not the true word is stored in the first database 90 (step S24). If YES is determined in this step S24, the process proceeds to step S26, and if NO is determined, the process proceeds to step S34.
[0022]
In step S34, a voice is output as "I do not understand your word. I will change the question." Then, the process returns to step S6 to output another question. That is, since it is impossible to interact any more, another question is asked.
[0023]
In step S26, it is confirmed with the user whether or not the true word specified in step S22 has been uttered. For example, when the true word specified in step S22 is "none", the voice is output as "Did you say pear?" At the same time as the output of the voice, a process of displaying “Please answer yes or no” on the screen of the display 70 is also executed. This display processing is performed by the control unit 40 sending display data to the display 70.
Subsequently, it is monitored whether the user uttered “Yes” or “No” (step S28). This process is executed when the voice specifying unit 30 specifies “Yes” or “No” and the control unit 40 inputs “Yes” or “No”. When the user utters “yes”, the process proceeds to step S30. When the user utters “No”, the process proceeds to step S40 in FIG.
In step S30, a response corresponding to the true word specified in step S22 is output as voice. For example, when the true word specified in step S22 is “none”, “slippery feeling is delicious” stored in the first database 90 in association with “none” is output.
After the processing of step S30 is completed, subsequently, processing for changing the storage content of the second database 100 is executed (step S32). For example, when the specific word is “palm” and the true word “none” is specified and the processing up to step S30 is executed, the specific word “palm” is associated with the true word “none”. And the weight associated with the specific word "palm" and another true word (for example, "palm" or "palm") is reduced. In the example of FIG. 4, the weight “80 (16/20)” associated with the specific word “palm” and the true word “none” is changed to “81 (17/21)”. Then, the weight “15 (3/20)” associated with the specific word “palm” and the true word “palm” is changed to “14 (3/21)”, and the specific word “palm” is changed. And the weight “5 (1/20)” associated with the true word “Yagi” is changed to “5 (1/21)”. For example, if the word specified in step S8 is “foot” and the true word “none” is specified in step S22 and the processing up to step S30 is executed, the weight “100 (1/1)” "To" 100 (2/2) ".
[0024]
When the processing is performed in the flow of steps S6 to S10 and S20 to S32 described above, the pattern is as shown in pattern 2 in FIG. In Pattern 2, the user answers "Pear" to the question "What is your favorite fruit?", But the voice specifying unit 30 incorrectly specifies "Palm". This “palm” is not in the first database 90, but is stored in the second database 100. At this time, since the weight associated with the specific word "none" and the true word "none" is the largest, the user is asked "Did you say pear?" Since the user answered "yes", he replied, "It's delicious." Finally, the weight of the second database 100 is changed.
[0025]
Subsequently, the processing from step S40 will be described with reference to FIG. In step S40, a voice output "Please enter the spoken word into the keyboard" is output. Next, it is monitored whether or not the user has made an input on the keyboard 80 (step S42). In this process, if the voice is input to the keyboard 80 before 30 seconds elapse from the voice output in step S40, YES is set, and if the voice is not input to the keyboard 80 until 30 seconds elapse, NO is set. If the determination is YES here, the process proceeds to step S44, and if the determination is NO, the process proceeds to step S50. By inputting character data to the keyboard 80, the user can correct character data erroneously converted by the voice identification unit 30. For example, if no specific word is stored in association with the user IDXXX1, and if the user utters “none” and is specified as “palm” in step S8, the user operates the keyboard 80. To enter "none". As a result, “palm” is corrected to “none”.
[0026]
In step S50, a voice is output as "I do not understand your word. I will change the question." In this case, the process returns to step S6 (FIG. 5) to output another question.
On the other hand, in step S44, it is determined whether or not the character data input from the keyboard is stored in first database 90. That is, it is determined whether or not there is character data input by the keyboard in step S42 in the answer group associated with the question selected in step S6. If the determination is YES here, the process proceeds to step S46, and if the determination is NO, the process proceeds to step S48.
In step S48, a voice is output as "I do not understand your word. I will change the question." In this case, the process returns to step S6 (FIG. 5) to output another question. In step S46, a response corresponding to the word input from the keyboard is output. When the processing in step S46 or S48 is completed, the storage contents of the second database 100 are changed (step S52). How the stored contents are changed in step S52 will be described in detail below.
[0027]
If it is determined as NO in step S20 and then the processing is performed in the flow of steps S40, S42, S44, S46, and S52, the pattern is as shown in pattern 3 in FIG. In Pattern 3, the user answers “Pear” to the question “What is your favorite fruit?”, But the voice specifying unit 30 incorrectly specifies “Palm”. The specified “palm” is not stored in the first database 90 and is not stored in the second database 100 as the specific word “palm”. When prompting for keyboard input, "none" is entered (that is, "palm" is corrected to "none"). He replied, "It's delicious." In this case, the specific word “palm” and the true word “none” are stored in the second database 100 in association with each other. The weight in this case is 100 (1/1).
Further, when it is determined NO in step S20 and the processing is performed in the flow of steps S40, S42, S44, S48, and S52, the pattern is as shown in pattern 4 in FIG. In Pattern 4, the user specifies “palm” to the question “What is your favorite fruit?” And the voice specifying unit 30 specifies “palm”. The specified “palm” is not stored in the first database 90 and is not stored in the second database 100 as the specific word “palm”. When the keyboard input is prompted, “Pay” is input to the keyboard 80. And he said, "I don't understand your word. I'll change my question." In this case, the specific word “palm” and the true word “palm” are stored in the second database 100 in association with each other. The weight in this case is 100 (1/1).
Further, when it is determined as NO in step S28 and the processing is performed in the flow of steps S40, S42, S44, S48, and S52, the pattern becomes the pattern 5 in FIG. In Pattern 5, the user answers “Pay” to the question “What is your favorite fruit?” And the voice specifying unit 30 specifies “Pay”. This “palm” is not stored in the first database 90, but is stored in the second database 100. Since the weight associated with the specific word “palm” and the true word “none” is the largest, the user is asked “Did you say pear?” At this time, when the user answered "No" and prompted for keyboard input, "Pay" was input to the keyboard 80. And he said, "I don't understand your word. I'll change my question." In this case, all weights associated with the specific word "palm" are changed. In the example of FIG. 4, the weight “15 (3/20)” associated with the specific word “palm” and the true word “palm” is changed to “19 (4/21)”, and the specific word “palm” is changed. The weight “80 (16/20)” associated with “Yashi” and the true word “None” is changed to “76 (16/21)” to correspond to the specific word “Yashi” and the true word “Yagi”. The assigned weight “5 (1/20)” is changed to “5 (1/21)”.
[0028]
According to the above-described interactive device 10 according to the present embodiment, even when the voice identification unit 30 erroneously converts character data different from the word uttered by the user, the word uttered by the user is accurately identified by covering the error. it can. Further, since the specific word and the true word are stored for each user in the second database 100, the true word can be specified for each user having different pronunciation, voice quality, and the like.
[0029]
As mentioned above, although the specific example of this invention was demonstrated in detail, these are only illustrations and do not limit a claim. The technology described in the claims includes various modifications and alterations of the specific examples illustrated above.
In the above-described embodiment, a method of inputting a name using the keyboard 80 has been adopted as a method of specifying a user (step S2). You can also. Further, the user can be specified by photographing the user using a camera. When this camera is used, there is no need to register a user name, and it is only necessary to store the facial features of the user and the user ID in association with each other.
In the above-described embodiment, as a method of correcting a specific word into a true word by the user, a method of inputting a true word to the keyboard 80 is adopted, but the following method may be adopted. For example, a true word can be corrected by having the user write a true word, photographing it with a camera or the like, and reading the image. It can also be corrected to a true word by having the user re-speak.
Further, the technical elements described in the present specification or the drawings exhibit technical utility singly or in various combinations, and are not limited to the combinations described in the claims at the time of filing. The technology illustrated in the present specification or the drawings achieves a plurality of objects at the same time, and has technical utility by achieving one of the objects.
[Brief description of the drawings]
FIG. 1 is a schematic configuration of a dialogue device according to an embodiment.
FIG. 2 shows an example of contents stored in a user name-user ID storage unit.
FIG. 3 shows an example of storage contents of a first database.
FIG. 4 shows an example of storage contents of a second database.
FIG. 5 is a flowchart of a process executed by the interactive device.
FIG. 6 is a flowchart (continuation of FIG. 5) of a process executed by the interactive device.
FIG. 7 is a flowchart of a process executed by the interactive device (continuation of FIG. 6).
FIG. 8 shows an example of a process executed by the interactive device (patterns 1 to 3).
FIG. 9 shows an example of a process executed by the interactive device (patterns 4 and 5).
[Explanation of symbols]
10. Dialogue device
20 ・ ・ Mike
30 ... voice identification unit
40 Control unit
50 ... Speech synthesis unit
60 speaker
70 Display
80 Keyboard
90-1st database
100 ... second database
110 user name-user ID storage unit

Claims (6)

It is a device that converts the voice uttered by the user into character data,
Means for inputting voice,
Means for converting the input voice to character data,
Correction means by which the user can correct the character data converted by the conversion means,
Means for storing the character data converted by the conversion means and the character data corrected from the character data in association with each other;
Means for searching the storage means using the character data converted by the conversion means as a key to specify "corrected character data". The character data conversion device according to claim 1, further comprising a user identification unit that identifies a user,
The storage unit stores the user specified by the user specifying unit, the character data converted by the conversion unit, and the character data corrected from the character data in association with each other,
The character data, wherein the specifying means searches the storage means using the user specified by the user specifying means and the character data converted by the converting means as a key to specify "corrected character data". Conversion device. The storage means is capable of storing a plurality of `` corrected character data '' in a weighted manner in association with the character data converted by the conversion means,
When a plurality of "corrected character data" are stored in association with the "character data converted by the converting means" as a key, the specifying means may select the "corrected character data" having the largest weight. 3. The character data conversion device according to claim 1, wherein data is specified. A device that interacts with the user by outputting a response corresponding to the voice emitted by the user,
Means for inputting voice,
Means for converting the input voice to character data,
Correction means by which the user can correct the character data converted by the conversion means,
First storage means for storing the character data converted by the conversion means and character data corrected from the character data in association with each other;
A second storage unit that stores a plurality of character data and stores a reply for each character data;
When the character data converted by the conversion means is not stored in the second storage means, the "character data converted by the conversion means" is used as a key to search the first storage means and the "corrected character data" Means for identifying "character data",
When the character data converted by the conversion unit or the “corrected character data” specified by the specification unit is stored in the second storage unit, the character data converted by the conversion unit or the specification unit A voice output of a response corresponding to the “corrected character data” specified in (1). A method in which a computer converts speech made by a user into character data,
Inputting a voice;
Converting the input voice into character data;
When the user has corrected the character data converted in the conversion step, the `` character data converted in the conversion step '' and a step of storing the character data corrected from the character data in association with each other,
Searching the storage contents stored in the storage step using the character data converted in the conversion step as a key to specify “corrected character data”. A program that allows a computer to convert user-generated speech into character data.
The computer performs the following processing:
The process of inputting voice,
A process of converting the input voice to character data;
When the user has corrected the character data converted in the conversion process, the process of storing the `` character data converted in the conversion process '' and the character data corrected from the character data in association with each other,
And performing a process of searching the storage content stored in the storage process using the character data converted in the conversion process as a key to specify “corrected character data”.

Images