JP2003208191A - Speech synthesis system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech synthesis system where the designation of a reading manner of and customizing can easily be performed by a writer and speech synthesis as designated by the writer is realized when being read in creating a document on the premise of reading the document through speech synthesis. <P>SOLUTION: The speech synthesis system comprises: a document editing device 100 provided with document editors 120 and 130 for performing inputting of rhythm information showing the reading manner of a word/phrase accompanying document input and data filing of document data and the rhythm information; and a document output device 200 having a speech synthesis means 240 which performs the speech synthesis of the document data on the basis of generated data files 122, 132 and 133 and which performs synthesis of speech according to the rhythm information when the rhythm information related with the document data exists and document editors 220 and 230 capable of displaying and outputting only a document expressing the document data without displaying the rhythm information. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique useful when applied to a system for creating a document on the assumption that voice synthesis is performed. For example, when exchanging a voice synthesized document among a plurality of users. The technology is particularly useful to use.

[0002]

2. Description of the Related Art For example, there is software that reads out and outputs an electronic mail by voice synthesis. When using such software to exchange e-mails on the premise that voice synthesis is performed, the sender inputs a document from a keyboard or the like to create and send an e-mail message, while the recipient receives it. You can listen to the e-mail text by synthesizing it with a voice synthesis program. Speech synthesis programs generally analyze the Japanese words such as what words a sentence is composed of and where the boundary between the main part and the predicate is, and then dictionary data that stores the voice information of various words. While associating the linguistic information obtained as a result of the above Japanese analysis with the acoustic features of each word,
Speech synthesis of a series of sentences is performed.

[0003]

However, in the e-mail system for exchanging e-mails on the premise of speech synthesis as described above, when a new word not registered in the dictionary data is used, the word There is a problem that pronunciation is not performed correctly. Further, even if the sender wants to customize the intonation of the phrase and have the other party read it, it is very difficult to satisfy such a desire.

An object of the present invention is to allow a writer to easily specify and customize a reading method when a document is created on the premise that it is read out by voice synthesis, and at the same time, a voice according to a writer's specification is read out. It is to provide a speech synthesis system that enables synthesis. The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

[0005]

The typical ones of the inventions disclosed in the present application will be outlined below. That is, in the document editing device, prosody information of words and phrases can be input along with document input, and while these document data and prosody information are made into a data file, the data output device performs data synthesis when performing voice synthesis of the document data. For words and phrases having prosodic information in the file, speech synthesis is performed according to the prosodic information.
According to such a means, it becomes easy to create a document so that the document can be read aloud according to the specified reading, and the document can be read aloud as specified.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining the outline of the speech synthesis system according to the embodiment of the present invention. The voice synthesis system of this embodiment is an editing apparatus 100 for inputting a document.
And an output device 20 for displaying a document and outputting a voice
And the editing device 100.
The output device 200 and the output device 200 are realized on separate personal computers by, for example, software configurations. Of course, both functions of the editing device 100 and the output device 200 may be provided on one computer.

The speech synthesis system shown in FIG. 1 has three types of editing methods and output methods. It should be noted that each of these methods is independent, and it is not necessary to provide all three types. If only any one type is provided, the function as the speech synthesis system according to the present invention is fulfilled. Next, each method will be described.

In the first method, prosody information is directly inserted into a document in a text data format and input, and a document is displayed and a voice synthesis output is performed from a data file containing the prosody information and the document data. is there. In the editing apparatus 100 in the first method, the editor 11 having a function as an input interface and a data file generating means is provided.
0 is used, and the output device 200 uses an editor 210 having a document display function and the like. In the first method, document input can be performed using a general-purpose text editor 110 capable of inputting text data, underline, and the like. Then, when a new word that is not registered in the dictionary for voice synthesis is used during inputting a document using the editor 110 or when it is desired to customize the pronunciation of the word, the prosody information is converted into text data as described later. It corresponds by directly inputting in the document in the format.

When the document input is completed, the text editor 110 combines the document data and the prosody information together into a text data format (extended text data format including underline information, or HTML (Hyper Text Mark-up).
(including language) etc.) is generated. On the output device 200 side, as a module related to the first method, a general-purpose text editor 210 that can handle a data file in the above-mentioned text data format, and a Japanese language processing module 240 as a voice synthesizing means for performing Japanese voice synthesizing. A rule synthesis processing module 260 and the like are provided. These text editor 210, Japanese processing module 240, rule composition processing module 26
0 is a software configuration realized by a computer. When the data file 112 is transferred to the output device 200, first, the data file 112 is transferred.
Is input to the text editor 210, and the document is displayed and output. However, the contents displayed and output here include prosody information as well as document data.

The data file 112 is also sent to the Japanese language processing module 240 as it is, the prosody information inserted in the document is separated from the document data by the module 240, and the Japanese data is analyzed and analyzed. Done. Then, the reading prosodic symbols of each word (data including prosodic information such as kana, accents, and word phrases) according to the grammar and clause are extracted from the dictionary, and the entire document data is converted into reading prosodic symbols. Go. Then, the reading prosody symbol data 250 is thereby generated. At the time of this conversion processing into reading prosodic symbols, the prosody information is prioritized and converted into the reading prosodic symbols for the part having prosody information inserted in the document. After that, in the output device 200, the reading prosody symbol data 250 is sent to the rule synthesis processing module 260, and regular speech synthesis according to the reading prosody symbol is performed. Then, the synthesized voice is output from the speaker 270 or the like. In this way, the above-mentioned document editing and document output methods allow the user to easily specify or customize the reading of words and phrases during document input, and at the side of the reader of the data file, the voice synthesis output according to the reading specified by the writer. You can do it.

2 to 5 show some examples of inputting prosody information in text format in the text editor 110 of the editing apparatus 100. The example of FIG. 2 illustrates a case where a word accent is designated. As shown in the same figure, for example, a new word "Ohhh" or the like is not specified where the accent is present even if it is recognized as one word in the usual speech synthesis process. As described above, it is converted into a reading prosodic symbol with no accent, such as "Ocher." Here, “.” Is a punctuation mark.
Therefore, when it is desired to add an accent to the new word "Ohha", an accent symbol (for example, "'") is inserted after the character to be accented, as shown in FIG. By doing so, in the voice synthesis process, the accented part is converted into a reading prosody symbol in which the accented part is designated by the user, "Oh, ha.".

The example of FIG. 3 explains a case where a word phrase segment having one accent is designated. Since a new word "i-mode" or the like is recognized as two words "i" and "mode" in a normal speech synthesis process, as shown in FIG. . ”And is divided into two word phrases and converted into a reading prosodic symbol with two accents. Here, "/" is a word phrase separator. Therefore, when it is desired to recognize a word or phrase recognized as a plurality of words as one word, the word or phrase to be recognized as one word is underlined as a decoration symbol as shown in FIG. 3B. By doing so, in the voice synthesis process, "a'i mode."
It is converted into a reading prosody symbol designated by the user as one word.

The example shown in FIG. 4 illustrates a case where the reading is modified in addition to the example shown in FIG. The word "IT revolution" or the like is recognized as two words "IT" and "revolution" in the normal speech synthesis process, and further, "IT" is recognized as an incorrect pronunciation such as "it". As shown in (A), it is converted into a reading prosody symbol called "It / Kakume". Therefore, to correct the reading of a word recognized as having an incorrect pronunciation, the reading kana is inserted in parentheses after the word as shown in FIG. 4 (B). Thereby,
The speech synthesis processing of the word will be performed using the phonetic alphabet designated by the user. Further, in the example of FIG. 4 (B), one word is also recognized by being underlined, and as a result, it is converted into a reading prosodic symbol called "itika'kume". The first accent of "Revolution" is
It can be added by inputting an accent symbol, but here, the accent of "revolution" registered in the dictionary as a word with a prefix is attached.

The example of FIG. 5 illustrates the case where various prosody parameters are specified for a phrase. When specifying prosodic parameters such as pitch of a sound, loudness of a sound, and duration of a sound (output time length) for a word or a phrase, as shown in FIGS. Later, symbols and characters that are preset for specifying prosody parameters are inserted in parentheses so that they are not parenthesized. The characters set for the prosody parameter are:
For example, "fast" and "slow" that represent the length of the sound, "high" and "low" that represent the pitch of the sound, "big" and "small" that represent the loudness of the sound, and "emphasis" that makes the sound loud and long. is there. Then, when the document is input, the prosody parameter is specified by adding the above symbol after the word. When such a designation is made, as shown in FIGS.
It is converted into a reading prosody symbol with a prosody parameter by a voice synthesis process. In the figure, “S”, “P”, “F”, and “V” indicate the types of prosody parameters, and the numbers indicate the parameter values.

Further, when setting a prosodic parameter for a phrase in which a plurality of words are gathered, as shown in FIG. 5 (e), the phrase is delimited by a hook and a prosodic parameter symbol is inserted after that. In this case, the voice synthesis process converts the prosodic parameters into a plurality of words into reading prosodic symbols.
In the figure, “>” is a symbol indicating the end of the specified range of the prosody parameter in the case of a plurality of words. As described above, prosody information can be input during text input in the text editor 110.

Next, the second method will be described. The second method is such that prosody information is input during document input by command selection different from the document input method, and the document data and the prosody information are handled as a dedicated data file in a different form. is there. In the second method, the editing apparatus 100 side uses a dedicated editor 120 having a function as an input interface and a data file generating means, and the output apparatus 200 side has a dedicated editor 2 having a function as an output interface.
20 is used.

FIG. 6 shows an image diagram where the prosody information is input in the dedicated editor 120. The dedicated editor 120 is almost the same as a general-purpose document editing module capable of adding various extended information such as formats and fonts in addition to document input. The dedicated editor 120 of this embodiment further includes prosodic information of words and phrases as extended information that can be added to document data. To input the prosody information, as shown in FIG. 6, after selecting a phrase to which the prosody information is added, the dedicated editor 12
This is done by selecting the prosody information insertion command 126 from the menu 125 provided in 0. 2 to 5 in the balloon display 127 appearing by this command selection.
Enter the reading prosodic symbol as shown in.

It should be noted that instead of directly inputting the reading prosody symbols into the balloon display 127, the prosody information may be input using a symbol system that is defined so as to be easily understood by the user. Further, the method of command selection, the display for inputting prosody information (callout display 127), and the like can be changed to various methods. When the input of the document is completed in the dedicated editor 120, the dedicated editor 120 associates the document data (“Kanji-kana mixed sentence”) and the prosody information with different forms and the prosody information is associated with which phrase. A data file is created in the state, and a dedicated data file 122 is generated.

On the output device 200 side, when the data file 122 is received, the data file 122 is received.
It is possible to display and output only the document without the prosody information by using the dedicated editor 220 corresponding to the format. It should be noted that a function of displaying the prosody information by a special method may be added by, for example, displaying the prosody information added to the phrase by displaying it by designating the phrase. On the output device 200 side, the data file 122 is the Japanese language processing module 24.
0, the Japanese analysis of the document is performed, the reading prosodic symbols of each word related to the language information are extracted, and the entire document data is converted into reading prosodic symbols. Then, the reading prosody symbol data 250 is generated. At the time of conversion of the reading prosody symbols, the conversion processing is performed for words and phrases associated with the prosody information included in the data file 122 with priority given to the prosody information. Then, this reading prosodic symbol 250 is sent to the rule synthesis processing module 260,
Regular voice synthesis is performed according to the reading prosodic symbols, and the synthesized voice is output from the speaker 270 or the like.

As described above, also in this second method,
The user can easily specify or customize the reading method of the phrase during input of the document, and the reader side of the data file can perform voice synthesis output according to the reading method specified by the writer. Further, when the document is displayed on the reader side, it is possible to prevent the prosodic information indicating the reading from being displayed, which is natural and preferable.

In the third method, the data file format is the second.
Unlike the system, the other method of inputting prosody information is similar to that of the second system. In the third method, the editing apparatus 100 side uses a dedicated editor 130 having a function as an input interface and a data file generating means, and the output apparatus 200 side has a dedicated editor 23 having a function as an output interface.
0 is used. In the third method, the document data input method and the prosody information input method using the dedicated editor 230 are the same as those shown in FIG. On the other hand, the data file generated by the dedicated editor 230 in the third method is a text data file 132 in which document data, format information, etc. are stored, and a prosody in which prosody information is stored in association with words of the document data. A data file 133 and two data files.

When the output device 200 receives these two data files 132 and 133, the dedicated editor 230 processes the text data file 132 to display and output the document, while the Japanese processing module 240 processes the text. Both the data file 132 and the prosody data file 133 are input, and the voice synthesis processing is performed as in the second method. Then, similarly, the synthesized voice is output from the speaker 270 or the like. As described above, according to the speech synthesis system of the above-described embodiment, when writing a document on the premise that the document is read out by speech synthesis, the reading method intended by the writer can be easily specified during document input, and the reading side Also in, the voice synthesis can be performed in accordance with this designation.

In the example of FIG. 1, the writer side is provided with the editing device 100 and the reader side is provided with the output device 200. However, both the writer side and the reader side have the editing device 100 and the output device, respectively. It is general that the system is configured so that it is provided with the device 200 so that the documents for voice synthesis can be exchanged bidirectionally. With such a configuration, it is also possible to proceed with the document creation and the designation of the reading while confirming with the ear the voice created by the user for the document created by the user.

Although the invention made by the present inventor has been specifically described based on the embodiments, the invention is not limited to the above embodiments, and various modifications can be made without departing from the scope of the invention. Needless to say. For example, as a method of speech synthesis, a method of converting Japanese into a reading prosody symbol using dictionary data after Japanese analysis of a document, and then regularly synthesizing the reading prosody symbol was shown. Any method may be used as long as it is a method of speech synthesis that reflects prosody information. Further, the method of inputting the prosody information concretely shown in the embodiments can be variously changed as long as the user can easily input the text input.

In the above description, the invention made by the present inventor is mainly used in a situation where documents are exchanged between a plurality of users on the assumption that voice synthesis is output which is a field of use which is the background of the invention. Although the speech synthesis system has been described, the present invention is not limited thereto, and can be widely used, for example, a speech synthesis system used to create a speech synthesis signal or synthetic speech data.

[0026]

The effects obtained by the representative one of the inventions disclosed in the present application will be briefly described as follows. That is, according to the present invention, when writing a document on the premise that the document is read out by voice synthesis, it is possible to easily specify the intended reading on the writer side during document input, and the reader side There is an effect that the voice can be synthesized by the reading according to the designation.

[Brief description of drawings]

FIG. 1 is a schematic diagram illustrating a speech synthesis system according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a first example of inputting prosody information in a text format.

FIG. 3 is a diagram illustrating a second example of inputting prosody information in a text format.

FIG. 4 is a diagram illustrating a third example of inputting prosody information in a text format.

FIG. 5 is a diagram illustrating a fourth example of inputting prosody information in a text format.

FIG. 6 is a display example showing input of prosody information with a dedicated editor of the voice synthesis system.

[Explanation of symbols]

100 editing device 110 text editor 112 data files 120, 130 dedicated editor 122, 132, 133 data files 200 output device 220, 230 dedicated editor 240 Japanese processing module 260 Rule Synthesis Processing Module

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Tadashi Horio             5-22-1 Kamimizuhonmachi, Kodaira-shi, Tokyo Stock             Ceremony Company Hitachi Cho-LS System             Within (72) Inventor Takeo Mori             5-22-1 Kamimizuhonmachi, Kodaira-shi, Tokyo Stock             Ceremony Company Hitachi Cho-LS System             Within (72) Inventor Mitsuhiko Seo             5-22-1 Kamimizuhonmachi, Kodaira-shi, Tokyo Stock             Ceremony Company Hitachi Cho-LS System             Within F-term (reference) 5D045 AB02

Claims (5)

[Claims] 1. An input interface for inputting prosodic information indicating how to read a phrase when a document is input, and data for making a data file of the inputted document data and making a data file of the prosody information associated with the document data. A document editing apparatus having a file generation means, and voice synthesis of document data is performed based on the data file of the document data and prosody information, and if there is prosody information associated with the document data, the prosody information is followed. A document output device having a voice synthesizing unit for performing voice synthesis, and an output interface capable of displaying and outputting only the document represented by the document data without displaying the prosody information based on the data file of the document data and the prosody information. A speech synthesis system comprising: 2. An input interface for inputting prosody information indicating how to read a phrase in the same format as a document input when a document is input, and input document data and the prosody information integrally and in a text data format. A document editing apparatus having a data file generating means for converting to a data file, and based on the data file of the document data and the prosody information, the document data and the prosody information are separated to perform voice synthesis of the document data and the prosody information is added. And a document output device having a voice synthesizing means for performing voice synthesizing in accordance with the prosody information when there is a phrase, and a voice synthesizing system. 3. The prosody information is input by adding a decoration symbol to a phrase in the input interface,
3. The speech synthesis system according to claim 2, further comprising accent delimiter information that regards the words or phrases to which the decoration symbol is continuously given as a word having one accent. 4. The prosody information includes accent information indicating that a character is input by adding an accent mark to the character in the input interface and indicates that the character with the accent mark is to be accented. The speech synthesis system according to claim 2 or 3. 5. The prosody information includes prosody parameter information that sets a pronunciation prosody of a phrase as a prosody according to a prosody parameter, and the prosody parameter information adds the prosody parameter before or after the phrase in the input interface. 5. The voice synthesis system according to claim 2, wherein the voice synthesis system inputs the voice.