JP3857188B2 - Text-to-speech system and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a system for declaiming a text for precisely declaiming the text even when using a device having a limited storage capacity. <P>SOLUTION: In the method for declaiming a text in an environment wherein a server for managing data transmission/reception between terminals and a plurality of terminals utilizing the server are connected through a network, the server uses a large-scale word dictionary to perform word analysis of text data to be transmitted to a prescribed terminal being a destination and extracts words which are included in the large-scale word dictionary but are not in a word dictionary provided to the terminal being the destination, and further extracts reading information being information about how to read the words. The server then makes the extracted words and information into a user dictionary for the terminal being the destination and delivers it to the terminal. The terminal refers to the delivered user dictionary for the terminal and the previously prepared word dictionary to convert text data included in transmitted data to a voice. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a text-to-speech system and method that enable accurate text-to-speech even for portable terminals and the like with limited storage capacity.
[0002]
[Prior art]
With the rapid progress of the computer environment in recent years, the development of applications using speech synthesis technology with a heavy processing load has been diversified. Recently, it has been applied to mobile terminals such as mobile phones and PDAs (Personal Digital Assistants) in addition to conventional application fields such as car navigation and telephone answering systems. It is used. This application is a great help for visually impaired and elderly people when using these devices, and also allows healthy people to know the content of the text without taking their eyes off the screen, especially hands-free. However, it is highly convenient in the situation where is required.
[0003]
FIG. 1 is a block diagram of a conventional general text-to-speech system. In FIG. 1, when text data is input, the word analysis unit 11 in the language processing unit 1 includes a system dictionary 12 prepared in common by the system and a user dictionary 13 set individually by the text creator. Split a text string into words using two word dictionaries. Then, the syntax analysis unit 14 analyzes the dependency information of each word, and generates information such as how to pose.
[0004]
Next, the prosody generation unit 21 in the sound processing unit 2 sets the time length and pitch frequency pattern of each sound in the text, and the waveform generation unit 22 stores the speech elements stored in the speech unit data storage unit 23. A speech waveform is generated by connecting the pieces of data according to the time length, pitch frequency pattern, etc. of each sound set by the prosody generation unit 21.
[0005]
As an application example, FIG. 2 shows a configuration example of an electronic mail system equipped with a text-to-speech system. In FIG. 2, a mail server 101 and a plurality of terminals A31 to X51 are connected on a communication network 201. Mail transmitted from another terminal to the terminal A31 is received by the mail transmitting / receiving unit 301. When the mail is read out by voice, the voice synthesizer 302 generates a voice waveform using the system dictionary 304 prepared in advance in the system and the user dictionary 305 set individually by the user.
[0006]
Here, the speech synthesis unit 302 includes the word analysis unit 11, the syntax analysis unit 14, the prosody generation unit 21, the waveform generation unit 22, and the speech segment data 23 in FIG. 1. The generated voice waveform is output by the voice output unit 303. When there is a reading error in the audio output, the user dictionary 305 is updated by the user using the user dictionary editing unit 306 for each.
[0007]
[Problems to be solved by the invention]
In general, the reading accuracy of text data is greatly affected by the number of vocabularies stored in the two word dictionaries described above. However, since portable terminals and the like are required to be lightweight and compact, it is indispensable to reduce the storage capacity accompanying the reduction in the size of the storage device. Therefore, when the text-to-speech system is installed in a portable terminal or the like, the number of vocabularies in the word dictionary is often greatly reduced from the number of vocabularies in the word dictionary in PC software or stationary devices. For this reason, there is a problem that the reading accuracy of text data tends to be lower than usual.
[0008]
For example, there are thought to be hundreds of thousands of people's first and last names. Therefore, when only thousands of frequently used surnames are stored due to the capacity limit of the storage device of the system dictionary in a portable terminal or the like, the surnames not stored in the system dictionary can be read out by mistake. Increases nature.
[0009]
Specifically, the surname “Akutsu” is correctly read out if the reading is “Actu” in the system dictionary, but if the reading “Actu” is not stored in the system dictionary, A combination of readings of kanji will be mistakenly read as “Ahisatsu”.
[0010]
On the other hand, when the user or acquaintance of the terminal A31 actually has a person named “Akutsu”, it is considered that the mail transmitted to the terminal A31 frequently includes the character string “Akutsu”. . However, since the surname “Akutsu” is generally not a frequent surname, the surname is registered in the user dictionary instead of the system dictionary.
[0011]
However, it is necessary to register information such as notation, reading, accent, part of speech, etc. for word registration in the user dictionary, and the operation is complicated especially on a mobile terminal. There is also a problem that it is difficult.
[0012]
In order to solve this problem, Japanese Patent Application Laid-Open No. 9-135264 prepares a text conversion server, performs all text conversion (language processing) of e-mail on the server side, and reads / prosodic information as a result thereof. Is output to the user terminal.
[0013]
However, in such a system, since language processing is not prepared on the terminal side, it is necessary to access the mail server every time the user reads out text, and the convenience as a portable terminal may be impaired.
[0014]
Alternatively, in Japanese Patent Laid-Open No. 2000-20417, a method of using a user dictionary attached in addition to a system dictionary when a mail sender attaches user dictionary information to a mail and a mail server performs speech synthesis. It is disclosed.
[0015]
However, in this method, it is necessary for the mail sender to edit the user dictionary, and the user dictionary is transmitted / received every time the mail is exchanged on the communication line, which increases the load on the line. It was left.
[0016]
In order to solve the above-described problems, the present invention provides a text-to-speech system and method that can accurately read a text even when a device having a limited storage capacity such as a portable terminal is used. With the goal.
[0017]
[Means for Solving the Problems]
In order to achieve the above object, a text-to-speech system according to the present invention is a text-to-speech system in an environment in which a server that manages transmission and reception of data between terminals and a plurality of terminals that use the server are connected via a network. A server uses a large word dictionary to perform word analysis on text data transmitted to a predetermined terminal as a destination, and a destination included in the large word dictionary Create a user dictionary that extracts words that are not included in the word dictionary prepared for the terminal and the reading information that is information on how to read the word and creates it as a user dictionary for the destination terminal / Editing unit and a user dictionary distribution unit that distributes the user dictionary for the terminal that is the destination to the terminal, and is distributed at the terminal A user dictionary for terminals that are created on the server, with reference to the word dictionary that is prepared in advance, and converting the speech text data included in the transmitted data.
[0018]
With this configuration, it is possible to build the minimum necessary system dictionary with the same text reading accuracy as when using a large-capacity system dictionary on mobile terminals with limited storage capacity, and when the storage capacity is limited. Even if it exists, it becomes possible to realize text reading with high accuracy.
[0019]
The text-to-speech system according to the present invention preferably periodically transmits a user dictionary for the terminal from the server to the terminal. Alternatively, it is preferable that the terminal requests the server to distribute the user dictionary for the terminal, and the server distributes the user dictionary for the terminal in response to the request. This is because it is sufficient to update the user dictionary at an arbitrary timing according to the application.
[0020]
In the text-to-speech system according to the present invention, it is preferable that the server distributes the text data and the user dictionary for the terminal when data is distributed to the terminal. This is because, in particular, in an electronic mail system or the like, an updated user dictionary can be obtained at the same time as mail is transmitted.
[0021]
Further, the present invention is characterized by software that executes the functions of the text-to-speech system as described above as a processing step of a computer. Specifically, a server that manages transmission and reception of data between terminals, and a server A text-to-speech method in an environment in which a plurality of terminals using a network are connected via a network, in which a server uses a large-scale word dictionary and uses words for text data transmitted to a predetermined terminal as a destination Reading that is information about the analysis step, words that are included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal, and how to read the word. Extracting information and creating a user dictionary for the destination terminal and distributing the user dictionary for the destination terminal to the terminal The text data included in the transmitted data is converted into speech by referring to the terminal user dictionary created by the distributed server and the word dictionary prepared in advance at the terminal. A text-to-speech method and a computer-executable program that embodies such a process.
[0022]
With this configuration, by loading and executing the program on a computer, a system dictionary with the minimum necessary to make the text-to-speech accuracy equivalent to that when using a large-capacity system dictionary is built on portable terminals with limited storage capacity. Therefore, it is possible to realize a text-to-speech system that can accurately perform text-to-speech even when the storage capacity is limited.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, a text-to-speech system according to an embodiment of the present invention will be described with reference to the drawings. FIG. 3 is a configuration diagram of the text-to-speech system according to the embodiment of the present invention.
[0024]
As shown in FIG. 3, a server 110 including a mail server 101 includes a word analysis unit 102, a user dictionary creation / editing unit 103, a user dictionary distribution unit 104, a system dictionary (large) 105, a system dictionary (small) 106, a terminal User dictionaries 107 to 108 for A to terminal X are prepared. Here, the system dictionary (large) 105 is a word dictionary having a large capacity and an abundant vocabulary, and the system dictionary (small) 106 is a slightly small capacity word dictionary having a vocabulary of a subset of the system dictionary (large) 105. means. Accordingly, all vocabularies registered in the system dictionary (small) 106 are included in the system dictionary (large) 105. The system dictionary (small) 106 has the same contents as the system dictionary (small) 304 prepared in common for the terminals A to X.
[0025]
The terminal A31, which is one of a plurality of terminals connected to the communication network 201, includes a mail transmission / reception unit 301, a voice synthesis unit 302, a system dictionary (small) 304, a user dictionary 305, a voice output unit 303, as in the conventional example. In addition, a user dictionary editing unit 306 is prepared.
[0026]
When the mail server 101 receives a mail transmitted to the terminal A31 as a destination, the word analysis unit 102 divides the mail into words while referring to the system dictionary (large) 105. Then, the user dictionary creation / editing unit 103 stores, in the terminal A user dictionary 107, words that are not included in the system dictionary (small) 106 among the words included in the mail. The same processing is performed for mails that arrive at the mail server with terminal B to terminal X as destinations, and words that are not included in system dictionary (small) 106 are stored in user dictionary 107 to 108 for each terminal. .
[0027]
The contents of the user dictionary for terminal A created / edited in this way are distributed to terminal A together with the mail by the user dictionary distribution unit 104. In the terminal A31, the user dictionary editing unit 306 is used to update the user dictionary 305 of the terminal, so that at the time of mail distribution, the mail can be processed with the same accuracy as that of the large-capacity system dictionary (large) 105. The contents (text data) can be read out.
[0028]
Also, it is conceivable that the user dictionary is distributed at various timings. First, as shown in FIG. 4, it is conceivable to perform control so that distribution is performed periodically. In this case, it is necessary to provide the regular distribution control unit 109.
[0029]
The regular distribution control unit 109 has distribution records of the terminal user dictionaries 107 to 108 for each terminal, and periodically stores the terminal user dictionaries 107 to 108 when a predetermined time has elapsed since the last distribution. It controls to transmit to each terminal. For example, the user dictionary 107 for the terminal A is distributed to the terminal A by the user dictionary distribution unit 104 at regular intervals.
[0030]
In addition to the time interval, for example, when the update amount of each terminal user dictionary 107 to 108 reaches a certain value, the terminal user dictionary 107 to 108 is periodically transmitted to each terminal. It may be what you do.
[0031]
Alternatively, as shown in FIG. 5, it is possible to distribute the content when accessing the server 110 as necessary. In this case, the user dictionary requesting unit 307 is provided in the terminal A31, and the user dictionary requesting unit 307 accesses the server 110 as necessary, and distributes the user dictionary 107 to 108 for the terminal used by the claimant. Claim.
[0032]
On the other hand, the server 110 has a billing distribution control unit 111 and distributes the terminal A user dictionary 107 to the terminal A31 via the user dictionary distribution unit 104 in response to a request from the terminal A31. In the terminal A31, the user dictionary editing unit 306 is used to update the user dictionary 305 in the terminal.
[0033]
The contents stored in each of the terminal user dictionaries 107 to 108 are not limited to words that are not included in the system dictionary (small) 106. For example, a replica of the user dictionary at each terminal is stored as a server. It is also conceivable to store it on 110 and store it as a new user dictionary merged with the dictionary. In this case, the terminal user dictionary itself is distributed.
[0034]
In this case, as shown in FIG. 6, the server 110 has a user dictionary attachment unit 112 and attaches the terminal A user dictionary 107 to a mail addressed to the terminal A and transmits it to the terminal A31. In the terminal A31, the attached user dictionary 107 for the terminal A is extracted from the mail received by the user dictionary editing unit 306 and replaced with the existing user dictionary 305.
[0035]
Next, the processing flow of the program that implements the text-to-speech system according to the embodiment of the present invention will be described using mail as an example. FIG. 7 shows a flowchart of processing of a program in the server 110 that realizes the text-to-speech system according to the embodiment of the present invention.
[0036]
As shown in FIG. 7, first, a mail addressed to an arbitrary terminal is received (step S701), and the mail content is divided into words while referring to the system dictionary (large) 105 (step S702).
[0037]
Next, the system dictionary (small) 106 is inquired to check whether or not a word extracted from the mail content is included (step S703). If the word extracted from the mail content is not included in the system dictionary (small) 106, the word and its reading are stored in the user dictionary 107 for the terminal that is the destination of the mail (step S704). .
[0038]
Then, the contents of the updated user dictionary for the terminal that is the destination of the mail are distributed to the terminal together with the mail (step S705). In the terminal that has received the contents of the terminal user dictionary, the user dictionary 305 in the terminal is updated based on the contents, thereby reading out the read mail to the system dictionary (large) 105 existing in the server 110. It is possible to perform with the same accuracy as when using.
[0039]
As described above, according to the present embodiment, it is possible to construct a minimum required system dictionary that makes text reading accuracy equivalent to that when using a large-capacity system dictionary on a portable terminal having a limited storage capacity, Even when the storage capacity is limited, it is possible to realize text reading with high accuracy.
[0040]
Particularly in a reading system for e-mail, user dictionary setting itself to be performed by the user in order to improve reading accuracy can be simplified, and high-precision text reading can be easily realized.
[0041]
As shown in FIG. 8, the program for realizing the text-to-speech system according to the embodiment of the present invention is not only a portable recording medium 82 such as a CD-ROM 82-1 and a flexible disk 82-2, but also a communication line. It may be stored in any of the other storage device 81 provided in front of the recording medium or a recording medium 84 such as a hard disk or a RAM of the computer 83. When the program is executed, the program is loaded and stored in the main memory. Executed.
[0042]
Further, as for the personal dictionary generated by the text-to-speech system according to the embodiment of the present invention, only a portable recording medium 82 such as a CD-ROM 82-1 or a flexible disk 82-2 is used, as shown in FIG. Instead, it may be stored in any of the other storage device 81 provided at the end of the communication line and the recording medium 84 such as a hard disk or RAM of the computer 83. For example, the text-to-speech system according to the present invention Is read by the computer 83 when using.
[0043]
(Supplementary Note 1) A text-to-speech system in an environment in which a server that manages transmission and reception of data between terminals and a plurality of terminals that use the server are connected via a network,
The server is
A word analysis unit that performs word analysis on text data transmitted to a predetermined terminal using a large word dictionary;
Extracting words that are included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and reading information that is information on how to read the word; A user dictionary creation / editing unit for creating a user dictionary for the terminal as a destination;
A user dictionary distribution unit that distributes a user dictionary for the terminal as a destination to the terminal,
In the terminal,
The text which is converted into speech by converting text data included in the transmitted data with reference to the user dictionary for the terminal created by the distributed server and a word dictionary prepared in advance Reading system.
[0044]
(Supplementary note 2) The text-to-speech system according to supplementary note 1, in which a user dictionary for the terminal is periodically transmitted from the server to the terminal.
[0045]
(Supplementary note 3) The text-to-speech system according to supplementary note 1, wherein the terminal requests the server to distribute the user dictionary for the terminal, and the server distributes the user dictionary for the terminal in response to the request.
[0046]
(Supplementary note 4) The text-to-speech system according to any one of supplementary notes 1 to 3, wherein the server distributes the text data and the user dictionary for the terminal during data distribution to the terminal.
[0047]
(Supplementary note 5) A text-to-speech method in an environment in which a server that manages transmission and reception of data between terminals and a plurality of terminals that use the server are connected via a network,
The server is
Performing word analysis on text data transmitted to a predetermined terminal using a large word dictionary;
Extracting words that are included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and reading information that is information on how to read the word; Creating a user dictionary for the terminal as a destination;
Delivering the user dictionary for the terminal as a destination to the terminal,
In the terminal,
The text which is converted into speech by converting text data included in the transmitted data with reference to the user dictionary for the terminal created by the distributed server and a word dictionary prepared in advance Reading method.
[0048]
(Supplementary Note 6) A computer-executable program in the server for realizing a text-to-speech method in an environment in which a server for managing transmission / reception of data between terminals and a plurality of terminals using the server are connected via a network Because
Performing word analysis on text data sent to a predetermined terminal using a large word dictionary;
Extracting words that are included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and reading information that is information on how to read the word; Creating a user dictionary for the terminal as a destination;
Delivering a user dictionary for the terminal as a destination to the terminal.
[0049]
【The invention's effect】
As described above, according to the text-to-speech system according to the present invention, it is possible to construct a minimum necessary system dictionary that makes the text-to-speech accuracy equivalent to that when a large-capacity system dictionary is used on a portable terminal having a limited storage capacity. Thus, even when the storage capacity is limited, it is possible to realize text reading with high accuracy.
[Brief description of the drawings]
FIG. 1 is a block diagram of a conventional text-to-speech system. FIG. 2 is a configuration diagram of an e-mail system having a conventional text-to-speech system. FIG. 3 is a block diagram of a text-to-speech system according to an embodiment of the invention. FIG. 4 is a block diagram of a text-to-speech system according to an embodiment of the present invention. FIG. 5 is a block diagram of a text-to-speech system according to another embodiment of the present invention. FIG. 7 is a configuration diagram of a reading system. FIG. 7 is a flowchart of processing of a program in a server that realizes the text reading system according to the embodiment of the present invention.
DESCRIPTION OF SYMBOLS 1 Language processing part 2 Acoustic processing part 11,102 Word analysis part 12 System dictionary 13,305 User dictionary 14 Syntax analysis part 21 Prosody generation part 22 Waveform generation part 23 Speech unit data storage part 31 Terminal A
41 Terminal B
51 Terminal X
101 mail server 103 user dictionary creation / editing unit 104 user dictionary distribution unit 105 system dictionary (large)
106, 304 System dictionary (small)
107 User Dictionary for Terminal A 108 User Dictionary for Terminal X 109 Regular Distribution Control Unit 110 Server 111 Request Distribution Control Unit 112 User Dictionary Attachment Unit 201 Network 301 Mail Transmission / Reception Unit 302 Speech Synthesizer 303 Speech Output Unit 306 User Dictionary Editing Unit 307 User Dictionary requesting unit 81 Line destination storage device 82 Portable recording medium 82-1 such as CD-ROM or flexible disk CD-ROM
82-2 Flexible disk 83 Computer 84 Recording medium such as RAM / hard disk on computer

Claims (5)

A text-to-speech system in an environment in which a server that manages transmission and reception of data between terminals and a plurality of terminals that use the server are connected via a network,
The server is
A large word dictionary,
A small-sized word dictionary with the same contents as the word dictionary common to each terminal,
A user dictionary for each terminal,
A word analysis unit that performs word analysis on text data transmitted to a predetermined terminal using the large-scale word dictionary;
Referring to the small-capacity word dictionary , how to read a word that is included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and the word A user dictionary creation / editing unit that stores reading information, which is information related to the above, in the user dictionary for the terminal as a destination;
A user dictionary distribution unit that distributes a user dictionary for the terminal as a destination to the terminal,
In the terminal,
A text-to-speech system that converts text data included in transmitted data into speech with reference to a user dictionary for the terminal created by the distributed server and a word dictionary common to the terminals .   The text-to-speech system according to claim 1, wherein the terminal requests the server to distribute the user dictionary for the terminal, and the server distributes the user dictionary for the terminal in response to the request.   The text-to-speech system according to claim 1 or 2, wherein the server distributes the text data and the user dictionary for the terminal when data is distributed to the terminal. A text-to-speech method in an environment in which a server that manages transmission and reception of data between terminals and a plurality of terminals that use the server are connected via a network,
The server comprising a large-scale word dictionary, a small-capacity word dictionary having the same content as a word dictionary common to each terminal, and a user dictionary for each terminal ,
Using said large word dictionary, and performing word analysis for the text data to be transmitted to a predetermined terminal as a destination,
Referring to the small-capacity word dictionary , how to read a word that is included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and the word Storing the reading information, which is information related to this , in the user dictionary for the terminal as the destination;
Delivering the user dictionary for the terminal as a destination to the terminal,
In the terminal,
A text-to-speech method characterized in that text data included in transmitted data is converted to speech by referring to the terminal user dictionary created by the distributed server and the word dictionary common to the terminals . A large-scale word dictionary, a small-capacity word dictionary having the same contents as a word dictionary common to each terminal, a user dictionary for each terminal, a server for managing transmission and reception of data between terminals, and the server A computer-executable program in the server that embodies a text-to-speech method in an environment where a plurality of terminals to be used are connected via a network,
Using said large word dictionary, and performing word analysis for the text data to be transmitted to a predetermined terminal as a destination,
Referring to the small-capacity word dictionary , how to read a word that is included in the large-scale word dictionary but not included in the word dictionary prepared in the destination terminal and the word Storing the reading information, which is information relating to this , in the user dictionary for the terminal as the destination;
Program, characterized in that to perform the steps on a computer that distributes user dictionary for the terminal as the destination to the terminal.

Images