JP2003259013A - Inquiring method and inquiring system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that a high recognition accuracy of the voice recognition and a low communication cost are required for an inquiring system. <P>SOLUTION: An information terminal makes a question in a voice file to prepare a question mail attached to an electronic mail, and sends the question mail to a mail answer server. The mail answer server prepares and sends an answer mail, utilizing the voice recognition, and the information terminal receives the answer mail from the mail answer server, thus obtaining an inquiring system having a high recognition accuracy of the voice recognition and a low communication cost. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to an inquiry method and an inquiry system using voice recognition.

[0002]

2. Description of the Related Art Conventionally, an inquiry system is disclosed in Japanese Patent Laid-Open No. 05-2.
The one described in Japanese Patent No. 60184 is known.

FIG. 6 shows the configuration of a conventional inquiry system. In FIG. 6, a telephone 41 is wirelessly connected to a relay 42, and the relay 42 is connected to a telephone response server 43 via a telephone line. The database 44 is directly connected to the telephone response server 43.

The telephone 41 is for transmitting a question voice and receiving a response voice. The repeater 42 relays a signal from the telephone 41 to the telephone response server 43 and a signal from the telephone response server 43 to the telephone 41. The telephone response server 43 searches the database 44 according to the content of the question voice, creates a response sentence, and returns the response voice. Database 4
Reference numeral 4 holds data such as telephone directory information in which correspondences between individual / group names, addresses, and telephone numbers are described.

FIG. 7 is a flowchart showing an operation procedure when the telephone response server 43 processes a question voice and returns a response voice.

When the user makes a call to the telephone response server 43, the telephone response server 43 connects the line (step 51). Next, the user utters a question voice (for example, a voice "Taro Matsushita's telephone number?") (Step 52). Next, the telephone response server 43
The question voice is subjected to voice recognition processing to obtain a voice recognition result (step 53). As a recognition result, for example, a text result "Taro Matsushita's telephone number is" is obtained.

Next, the telephone response server 43 analyzes from the recognition result that "Taro Matsushita" is the name and the inquiry is "telephone number". Then, the telephone response server 43 searches the database 44 for the telephone number of "Taro Matsushita". For example, the result is 1234-567
Number 8 Then, the response sentence "Taro Matsushita's telephone number is 1234-5678" is generated (step 54).

Next, the telephone response server 43 performs voice synthesis processing on the text of the response sentence to generate a response voice and outputs it to the telephone line (step 55). Next, the user hears the response voice and hangs up, and the telephone response server 43 disconnects the telephone line (step 56). With the above processing procedure, the response processing to the inquiry is completed.

[0009]

In this inquiry system, it is required that the recognition accuracy of voice recognition is high and the communication cost is low.

It is an object of the present invention to realize an inquiry system with high recognition accuracy of voice recognition and low communication cost.

[0011]

In order to solve this problem, the present invention makes a question a voice file in an information terminal,
Attach it to an e-mail as a question e-mail, send the question e-mail to the e-mail response server, use the e-mail response server to create and send a response e-mail using voice recognition, and use the information terminal to respond from the e-mail response server It receives mail and is configured to go through voice files and email instead of using voice recognition over the phone.

As a result, the recognition accuracy of voice recognition is high,
An inquiry system with low communication costs can be obtained.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION In the invention described in claim 1 of the present invention, a question is converted into a voice file in an information terminal, attached to an electronic mail to form a question mail, and the question mail is transmitted to a mail response server, The mail response server uses voice recognition to create and send a reply mail, and the information terminal receives the reply mail from the mail reply server. It has the effect that it only takes time.

The invention according to claim 2 is characterized in that, in the inquiry method according to claim 1, the voice file is created without limiting the voice band of the telephone. Since there is no restriction, it has the effect of ensuring the accuracy of voice recognition.

The invention according to claim 3 is characterized in that, in the inquiry method according to claim 1 or 2, the response mail is in text format, and since the response text is in text format, information Is easy to reread
In addition, the communication cost is lower than that of the voice format.

The invention according to claim 4 is an inquiry system comprising an information terminal and a mail response server,
In an inquiry system including a plurality of information terminals and a mail response server, a voice file creation unit for converting the content of a question into a voice file, attaching the voice file to an email as a question mail, and transmitting the question mail to a mail response server An information terminal equipped with a mail sending unit and a reply mail receiving unit for receiving a reply mail from a mail reply server, a question mail receiving unit for receiving a question mail from the information terminal, and a question from the received voice file A voice recognition unit that converts the content into text data by voice recognition, a search unit that analyzes the text data and searches a database from the analyzed question content, and a response that creates a response sentence from the question content and the search result. A sentence creation unit, a voice synthesis unit that converts the response sentence into a voice file, and the voice file The inquiry system is characterized by comprising a mail response server having a sending unit for sending and attaching as a reply mail, and the communication cost only affects the time for sending and receiving mail. Have.

The invention according to claim 5 is the inquiry system according to claim 4, characterized in that the voice file is created without limiting the voice band of the telephone. The effect of ensuring the accuracy of speech recognition can be ensured because it is not restricted by.

The invention according to claim 6 is characterized in that, in the inquiry system according to claim 4 or claim 5, the response mail is in a text format. Easy to reread,
In addition, the communication cost is lower than that of the voice format.

According to the invention of claim 7, the question content is converted into a voice file, the question mail attached to the electronic mail is received, and the voice file of the received question mail is converted into a text sentence by using voice recognition. Characterized by analyzing the question contents converted into text, searching a database in which the answer contents are accumulated in advance from the analysis result, creating a response sentence from the search result and the analysis result, and transmitting it as a response mail. Since the response text is in the text format, it has the effect of making it easier to reread the information and reducing the communication cost compared to the voice format.

According to the invention described in claim 8, in the response method to the inquiry according to claim 7, the response mail is
It is characterized in that the response sentence is attached as a voice file by voice synthesis, and it has an effect that it can be used even in a terminal that can handle only voice signals.

According to a ninth aspect of the present invention, the question content is converted into a voice file, the question mail receiving section receives the question mail attached to the electronic mail, and the question content from the received voice file is converted into text data by voice recognition. A voice recognition unit that converts the text data, a search unit that analyzes the text data and searches the database from the analyzed question contents, a response sentence creation unit that creates a response sentence from the question contents and search results, and the response It is characterized by including a transmitting unit for transmitting a sentence as a response mail, and has an effect that the communication cost only affects the time for transmitting and receiving the mail.

According to a tenth aspect of the present invention, in the mail response server device according to the ninth aspect, the transmitting section is provided with a voice synthesizing section for converting the response sentence into a voice file, and the voice file is attached and transmitted. It is characterized by
It has an effect that it can be used even in a terminal that can handle only audio signals.

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

(Embodiment 1) FIG. 1 is a diagram showing a configuration of an inquiry system according to Embodiment 1 of the present invention. In FIG. 1, the information terminal 11 is wirelessly connected to a relay device 12, the relay device 12 is connected to a provider 13 via a telephone line, and the provider 13 is connected to a mail response server 14 via an internet line. There is. The database 15 is directly the mail response server 14
Connected with. As the information terminal 11, a mobile phone,
A PDA or the like can be used.

The relay machine 12 sends the inquiry mail to the information terminal 11
Relay to the provider 13. The provider 13 transfers the inquiry mail from the relay machine 12 to the mail response server 14.

Next, a configuration diagram of the information terminal 11 is shown in FIG. 2 and will be described below.

The information terminal 11 inputs a spoken question from the microphone 21 and converts it into a voice file by the voice input unit 22. The question mail creating unit 23 attaches the converted voice file to the electronic mail. The mail sending unit 24 sends it as a question mail. Also, the response mail receiving unit 25
Then, the response mail to the question mail is received, and the attached voice file or text is reproduced or displayed on the output unit 26.

Next, a configuration diagram of the mail response server 14 is shown in FIG. 3 and will be described below.

The inquiry mail from the information terminal 11 is received by the mail reception unit 31, and the voice file attached to the inquiry mail is taken out. The voice recognition unit 32 converts the voice signal from the extracted voice file into text data by voice recognition. The analysis unit 33 analyzes the content of the question mail from the text data from the voice recognition unit 32. Search section 3
Reference numeral 4 searches the database 15 according to the content of the question mail, and the response sentence creating unit 35 creates a response sentence based on the content of the question email and the search result. The voice synthesizing unit 36 converts the created response sentence into a voice file by voice synthesis. The sending unit 37 creates a response mail with the created voice file attached, and sends it to the provider 13. The database 15 of the present embodiment is assumed to hold data such as telephone directory information in which correspondences between individual / group names, addresses, and telephone numbers are described.

The reply mail sent from the mail reply server 14 is received by the information terminal 11 via the provider 13 and the relay device 12.

Next, the operation of the inquiry system will be described in detail with reference to FIGS.

FIG. 4 is a flow chart showing an operation procedure when creating a question mail using the information terminal 11.

The user uses the voice input function of the information terminal 11 to input a question sentence (for example, "utterance of Mr. Taro Matsushita's telephone number") from the microphone 21, and the voice input unit 22 outputs the voice file. Create (step 21). The file format of the audio file is MP3 (MPEG
-1 Audio Layer3), WMA (Windows (registered trademark) Media Audio) or ADPC
M (Adaptive Differential Pulse Code Modulation)
Is a well-known standard format.

Next, the user is asked by the question mail creating section 23.
Start the mail creation software at, attach the voice file of the question text to the e-mail with nothing written in the subject and text,
A question mail in which the mail address of the mail response server 14 is written is created (step 22). next,
The mail sending unit 24 sends a question mail (step 23).

The inquiry mail is transmitted by the above processing procedure. Next, a flow chart showing an example of an operation procedure of a response mail creating process in which the mail response server 14 processes a question mail and creates a response mail is shown in FIG. 5 and will be described below.

The mail receiving unit 31 receives the question mail from the user (step 31) and takes out the voice file attached to the question mail (step 32).

Next, the voice recognition unit 32 obtains text data as a voice recognition result by voice recognition processing of the voice signal from the voice file (step 33). The speech recognition process converts a speech signal into a characteristic parameter, searches for a word sequence that is acoustically and linguistically likely to be input with reference to a prepared acoustic model, language model, and pronunciation dictionary of words. It can be realized by doing. Feature parameters are cepstrum coefficients, acoustic model is HMM, language model is N-gr
The am model and the beam search method are known as search methods. Information such as the sampling frequency can be obtained from the header of the voice file to set the control parameters for voice recognition. As a recognition result, for example, a text result "Taro Matsushita's telephone number is" is obtained.

Next, the analysis unit 33 analyzes from the recognition result of the voice recognition unit 32 that "Taro Matsushita" is the name and the inquiry is "telephone number". The analysis method can be realized by a known method such as a slot filling method. The slot filling method prepares necessary slots (items such as "name" and "object of inquiry") and keywords that apply to each slot in advance, and extracts the part that matches the keyword from the text data and stores it in the slot. This is a method to be applied, which enables analysis of inquiry contents of text data.

Then, the search unit 34 searches the database 15 for the telephone number of "Taro Matsushita" based on the analysis result. For example, the result is set to 1234-5678. Then, the response sentence creating unit 35 determines that the response sentence “Taro Matsushita's telephone number is 12” based on the analysis result and the search result.
It is 34-5678. The text "" is generated (step 34). The response sentence is generated in advance for each inquiry target (for example, the response sentence pattern for inquiring a telephone number is "[name] 's telephone number is [phone number]. ]] ”) Is prepared and the contents in [] are replaced with concrete contents.

Next, the voice synthesizing unit 36 performs voice synthesis processing on the text of the response sentence to create a voice file (step 35). The voice synthesis processing can be realized by converting a text sentence into a sequence of phonetic symbols and connecting waveform data for phonetic symbols prepared in advance according to the sequence of phonetic symbols.

Next, the transmission unit 37 attaches the voice file of the response sentence or the text of the response sentence to the electronic mail in which the subject and the text are not written, and writes the address of the sender of the question mail in the response. Create an email (step 3
6) A reply mail is transmitted (step 37).

The response mail is transmitted by the above processing procedure.

The reception of the reply mail is received by the automatic reception function of the information terminal 11. Response mail receiver 25
Receives a reply mail to the question mail. The output unit 26 reproduces or displays the attached voice file or text on the speaker 28 or the display 27.

Alternatively, the user can manually receive the mail at an appropriate timing. The user can know the telephone number by playing the audio file attached to the response mail.

By the above procedure, the inquiry process is completed by exchanging the question mail and the reply mail.

In the above explanation, the method of attaching the voice file to both the question text and the response text has been explained, but the case where the question text is attached to the voice file and the response text is transmitted in the text form in the mail body is also applicable. It can be similarly implemented.

Further, in the above description, the mail response server performs batch processing from the voice recognition to the generation of the response sentence, but the voice recognition result is temporarily returned to the user in a text format. The user may confirm the voice recognition result and resend it, and the mail response server may generate the response sentence based on the question sentence of the text sentence.

Also, in the above description, the method of directly attaching the voice file has been described, but the voice file may be compressed by the information terminal and the voice file may be decompressed by the mail response server.

Further, it is judged from the reliability of the voice recognition result whether or not the recognition result is reliable, and if the recognition result is not reliable, a guide sentence explaining the utterance method should be sent to the user together with a notification that the response cannot be made. You may

[0050]

As described above, according to the present invention, the voice file and the electronic mail are used instead of the voice recognition via the telephone, the recognition accuracy of the voice recognition is high, and the communication cost is high. The advantageous effect that a cheap inquiry system can be obtained is obtained.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing an inquiry system according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of an information terminal according to the first embodiment of the present invention.

FIG. 3 is a configuration diagram of a mail response server according to the first embodiment of the present invention.

FIG. 4 is a flowchart showing an operation procedure when creating a question mail.

FIG. 5 is a flowchart showing an example of an operation procedure when processing a question mail and creating a response mail.

FIG. 6 is a schematic configuration diagram showing a conventional inquiry system.

FIG. 7 is a flowchart showing an example of an operation procedure when processing a question voice and returning a response voice.

[Explanation of symbols]

11 Information terminal 12 repeaters 13 providers 14 Email response server 15 Database 21 microphone 22 Voice input section 23 Question mail composer 24 E-mail sending section 25 Response mail receiver 26 Output section 27 display 28 speakers 31 Email Receiver 32 Speech recognition unit 33 Analysis Department 34 Search Department 35 Response sentence creation section 36 Speech synthesizer 37 Transmitter

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G10L 15/28 H04M 3/42 R H04M 3/42 Z 11/00 302 G10L 3/00 551A 11/00 302 561H 571A (72) Inventor Natsuki Saito 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. F term (reference) 5D015 AA05 KK02 LL12 5K015 AA07 GA07 5K024 BB07 GG03 GG05 5K101 KK02 LL01 NN08 NN11 NN11 NN11

Claims (10)

[Claims] 1. The information terminal converts the question into a voice file, attaches it to an e-mail to form a question mail, transmits the question mail to a mail response server, and the mail response server uses voice recognition to reply the response mail. An inquiry method characterized by creating and transmitting and receiving a reply mail from the mail reply server at the information terminal. 2. The inquiry method according to claim 1, wherein the voice file is created without limiting the voice band of the telephone. 3. The inquiry method according to claim 1, wherein the response mail is in a text format. 4. An inquiry system comprising a plurality of information terminals and a mail response server, and a voice file creating section for converting the content of a question into a voice file, and attaching the voice file to an electronic mail to form a question mail, and the question mail is mailed. An information terminal having a mail sending unit for sending to a response server and a reply mail receiving unit for receiving a reply mail from the mail reply server; a question mail receiving unit for receiving a question mail from the information terminal; A voice recognition unit that converts question contents from a voice file into text data by voice recognition, a search unit that analyzes the text data and searches a database from the analyzed question contents, and a response from the question contents and search results A response sentence creating unit for creating a sentence; a voice synthesis unit for converting the response sentence into a voice file; An inquiry system, comprising: a mail response server including a transmission unit that attaches and transmits a voice recording file as a response mail. 5. The inquiry system according to claim 4, wherein the voice file is created without limiting the voice band of the telephone. 6. The inquiry system according to claim 4 or 5, wherein the response mail is in a text format. 7. The question converted into a text by receiving a question mail attached to an e-mail with the question content as a voice file, converting the voice file of the received question mail into a text sentence by using voice recognition, and converting into a text. A method for responding to an inquiry, which comprises analyzing a content, searching a database in which response contents are accumulated in advance from the analysis result, creating a response sentence from the search result and the analysis result, and transmitting it as a response mail. 8. The response mail is characterized in that the response sentence is attached as a voice file by voice synthesis.
How to respond to the inquiries listed. 9. A question mail receiving unit for receiving a question mail attached to an e-mail, the question contents being a voice file, and a voice recognition unit for converting the question contents from the received voice file into text data by voice recognition. , A search unit that analyzes the text data and searches a database from the analyzed question contents, a response sentence creation unit that creates a response sentence from the question contents and search results, and sends the response sentence as a response mail A mail response server device comprising a transmission unit. 10. The mail response server device according to claim 9, wherein the transmission unit is provided with a voice synthesizing unit for converting the response sentence into a voice file, and the voice file is attached and transmitted.