JPH09258763A - Voice synthesizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice synthesizing device that can read out correctly even in the case of including special words needing special way of reading. SOLUTION: When a kana-kanji mixed sentence to be read out is inputted, a read-out control means 1 supplies it to a tag information retrieving means 2, and a sentence implementor retrieves tag information added at the time of preparing a sentence. In the case of detecting tag information that assigns the way of reading, the special word information of this tag is registered in a special word dictionary 3. In the case of detecting tag information that assigns another way of reading in the same notation, an object word is stored in a read-out buffer 4, substituting for the phoneme-rhythm information of this tag, and the object word is stored in a display buffer 8, substituting for the noted information of this tag. A next processing means 6 converts read-out data, stored in the read-out buffer 4, into a pronounciation mark row, referring to the special word dictionary 3 and a general word dictionary 5 and supplies it to a voice synthesizing means 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing apparatus for reading a voice as a voice by performing language processing on a sentence containing kanji and kana and synthesizing the result.

[0002]

2. Description of the Related Art In recent years, a speech synthesizer has been put into practical use, in which a sentence containing a mixture of Japanese kanji and kana is subjected to language processing, and the result is subjected to speech synthesis processing to read it out as speech. In this speech synthesizer, language processing is performed by referring to a word dictionary in which reading information, accent information, and the like for words in a sentence mixed with Kanji and kana are registered, and the result is used to read aloud.

However, the words in the kanji and kana mixed sentences may include not only general and general-purpose words but also special words such as uncommon proper nouns and technical terms. To give correct reading and accent not only to general-purpose words but also to uncommon words,
Although it is sufficient to register all these words in one word dictionary, it is practically impossible to configure one word dictionary so as to cover all of these pieces of information. Therefore, the read sentence is not always what the sentence creator intended.

Conventionally, this type of speech synthesizer is described in, for example, Japanese Patent Application Laid-Open No. 4-331998. The speech synthesizer described in this publication embeds in a sentence special word information that defines information such as reading and accent for a word that is not registered in a word dictionary, and uses this information to write a sentence. Accurate reading is performed according to the reading intended by the creator.

The invention described in this publication will be described with reference to the drawings. FIG. 15 is a block diagram of the speech synthesizer described in the publication. As shown in FIG. 15, this speech synthesizer performs processing on a sentence containing Kana-Kanji mixed characters to be read aloud. Therefore, embedded special word information is extracted from the input Kana-Kanji mixed sentence and the special word information is extracted. A special word extraction unit 102 which stores the special word information in the sentence into the notation of this special word while saving it in the word information buffer 101; and analyzes a sentence mixed with kana-kanji resulting from the special word information being replaced with the notation of the special word. do it,
A language processing unit 103 for converting this into phoneme / prosodic information;
A special word conversion unit 104 for replacing the phoneme and prosody information obtained by the language processing unit 103 with the phoneme and prosody information of the special word stored in the special word information buffer 101;
It is composed of a voice synthesis unit 105 which performs voice synthesis based on the phoneme and prosody information from the special word conversion unit 104 and reads the voice.

FIG. 16 shows an example of a kana-kanji mixed sentence which is input to the system shown in FIG. In FIG. 16, special words (MIC and LI
For NE), this notation information (MIC, LINE)
And phonological / prosodic information (ma'iku, la'in) are separated by a colon (:), and special word information consisting of these information is sandwiched by a semicolon (;), and a sentence containing kana-kanji to be processed is mixed. Embedded in advance.

Next, the operation will be described. Now, assuming that the sentence creator has created a sentence mixed with Kana-Kanji characters, for example, as shown in FIG. 16, and inputting this into the system of FIG. 15, the special word extraction unit 102 embeds it in the sentence mixed with Kana-Kanji characters. Extracted special word information.
That is, when a semicolon (;) is detected in a sentence containing Kana-Kanji, the space between this semicolon (;) and the colon (:) after it is extracted as the notation information of the special word, and the colon (:) The part up to the semicolon (;) is extracted as phoneme and prosody information of the special word. As a result, for example, with respect to the special word information “; MIC: ma'iku;”, “MIC” is extracted as the notation information, and “ma'iku” is extracted as the phoneme and prosody information.

After extracting the special word information in this way, the special word extracting unit 102 stores the result in the special word information buffer 101. In addition, the character string from semicolon to semicolon is replaced with the notation of this special word, and the part corresponding to the special word is clearly marked (underlined) in the sentence mixed with Kana and Kanji.

As a result, the language processing unit 103 is sent from the special word extracting unit 102 a sentence mixed with Kana-Kanji in which the notation of the special word is marked. The language processing unit 103 analyzes this and converts each word in this sentence into phonemic and prosodic information. At this time, the marked portion is processed so as to store the position information. For processing each word, a known general word dictionary not shown in FIG. 15 is used.

Thereafter, in the special word conversion unit 104, regarding the phoneme and prosody information corresponding to the marking position in the phoneme and prosody information obtained by the language processing unit 103, this is corresponded to in the special word information buffer 101. Replace with phoneme and prosody information of special words.

As a result, as a result of the processing by the language processing 103, the special words can be replaced with those stored in the special word information buffer 101 as intended by the sentence creator and sent to the speech synthesizer 105. The speech synthesis unit 105 synthesizes speech using a known rule speech synthesis technique and reads it aloud.

[0012] Also, when the same special word repeatedly exists in a sentence mixed with Kana and Kanji, special word information is added to a sentence mixed with Kana and Kanji in a footnote style, and an example of a speech synthesizer utilizing this is also provided. It is described in the above-mentioned JP-A-4-331998. This speech synthesizer aims at eliminating the need to specify special word information many times even when the same special word is present in a sentence many times.

This prior art will be described with reference to FIG. As shown in FIG. 17, this speech synthesizer performs a process on a sentence mixed with Kana-Kanji characters, so a special word extraction unit 111 for extracting special word information added to a sentence mixed with Kana-Kanji characters; A special word dictionary creating unit 113 that creates a special word dictionary 112 based on the extracted special word information; a general word dictionary 114 in which phoneme and prosody information of general words is stored in advance; special words that have been added A language processing unit 115 for inputting only a Kana-Kanji mixed sentence in a form in which information has been deleted, and performing a language process on the Kana-Kanji mixed sentence by referring to both the special word dictionary 112 and the general word dictionary 114; A voice synthesis unit 116 that performs voice synthesis based on the phoneme and prosody information from the voice 115 and reads the voice aloud.

FIG. 18 shows an example of a kana-kanji mixed sentence which is input to the system shown in FIG. 17 to perform a voice synthesis process. In FIG. 18, regarding the special words (MIC and LINE), this notation information (MIC, LIN)
E) and phonological / prosodic information (ma'iku, la'in) are separated by a colon (:), and special word information consisting of these information is sandwiched by semicolons (;) to be processed kana kanji. Add it to the mixed text in advance.

Next, the operation will be described. Now, when the sentence creator inputs a kana-kanji mixed sentence with special word information as shown in FIG. 18 into the system of FIG. 17, the special word extraction unit 111 adds it to the kana-kanji mixed sentence. Extract the special word information.
That is, in the example of FIG. 18, when a semicolon (;) at the beginning of a line is detected, the space between this semicolon (;) and the colon (:) after it is extracted as the notation information of the special word,
The phoneme of the special word from this colon (:) to the end of the line,
Extract as prosodic information.

When the special word extracting unit 111 extracts the special word information in this way, the result is given to the special word dictionary creating unit 113, and the special word dictionary creating unit 113 specializes based on the extracted special word information. The word dictionary 112 is created.

When the special word extracting unit 111 extracts the special word information, the special word extracting unit 111 deletes the special word information and gives only the sentence containing Kana-Kanji to the language processing unit 115.

As a result, the language processing unit 115 is sent from the special word extracting unit 111 only the sentence containing the kana and kanji characters in which the special word information is deleted.

The language processing unit 115 analyzes this and converts each word in this sentence into phoneme and prosody information. At this time, when performing language processing on a certain word, the language processing unit 115 refers to the special word dictionary 112 and the general word dictionary 114, but general word information is generally stored in the special word dictionary 112. Since it is not registered and the information of the special word is not generally registered in the general word dictionary 114, the general word dictionary 114 is referred to for general words, and the special word “MIC” in the sentence is referred to. ,
For "LINE", refer to the special word dictionary 112. As a result, the language processing unit 115 causes the special word “MI
For “C” and “LINE”, these are replaced with phoneme and prosody information registered in the special word dictionary 112,
It can be sent to the voice synthesizer 116. Speech synthesizer 11
In 6, a voice is synthesized by using a known rule voice synthesis technique and read out aloud.

[0020]

The first problem is that the sentence creator must add the special word information of the special word every time there is a target word.

In the first example described in Japanese Patent Application Laid-Open No. 4-331998, the sentence creator previously embeds special word information in a sentence containing a mixture of kana and kanji, and utilizes the embedded special word information. The text is read aloud as intended by the author. For this reason, it becomes necessary for the sentence creator to embed special word information for each target word that he / she wants to read.

The second problem is that plural readings cannot be designated for words having the same notation.

In the second example disclosed in Japanese Patent Laid-Open No. 4-331998, special word information is added to a footnote style in a sentence mixed with Kana and Kanji, and the added special word information is registered in the special word dictionary. When reading aloud, the target word is replaced with the phoneme and prosody information of this special word dictionary. For this reason, it becomes impossible to read aloud with phonemes, phonemes other than prosody, and prosody information registered in the special word dictionary. Further, when it is desired to read aloud other than the registered phoneme and prosody information, it is necessary to re-register the special word information.

A third problem is that when sentence data containing kana-kanji characters is input in page units or the like, footnote-like special word information may not be obtained, and it may not be possible to read aloud as intended by the sentence creator. .

In the second example described in Japanese Patent Laid-Open No. 4-331998, special word information is added to a sentence mixed with Kana and Kanji in a footnote style. When the sentence data to be input is freely selected, the special creator may add special word information, but the special word information cannot be acquired, so that the intended reading may not be performed.

Therefore, the present invention solves the above-mentioned problems of the prior art, and it is not necessary to add special word information every time there is a target word. The purpose of the present invention is to provide a speech synthesizer capable of designating a reading method and reading the speech in a reading method intended by the sentence creator.

[0027]

The first speech synthesizer of the present invention comprises means for retrieving tag information added to a sentence mixed with kana and kanji to be read (2 in FIG. 1), and a special word obtained from the tag information. It has a special word dictionary (3 in FIG. 1) for registering information, and a reading control means (1 in FIG. 1) for converting an input kana-kanji mixed sentence into display data and voice data by using the special word information. .

The second speech synthesizing device of the present invention is a means for retrieving tag information added to a sentence containing kana-kanji mixed characters to be read out (2 in FIG. 5) and converting a target word into tag information and inputting it. It has a reading control means (1 in FIG. 5) for converting a kana-kanji mixed sentence to be displayed data and voice data by using special word information.

The third voice synthesizing device of the present invention is a control means (1 in FIG. 7) for distinguishing and inputting header information and a sentence mixed with kana and kanji to be read into the reading control means (1 in FIG. 7).
1), means for retrieving tag information (2 in FIG. 7), and registering the special word information in the header information in the special word dictionary (3 in FIG. 7), and using the special word information for a sentence mixed with kana and kanji to be read. And a reading control means (1 in FIG. 7) for converting the display data and the voice data.

The fourth speech synthesizer of the present invention is a control means (1 in FIG. 9) for distinguishing and inputting header information and a sentence mixed with kana and kanji to be read into the reading control means (1 in FIG. 9).
1), means for searching tag information (2 in FIG. 9), and converting the target word into special word information of the tag information of the header information, and using the special word information for a sentence mixed with kana and kanji to be read. It has reading control means (1 in FIG. 9) for converting display data and voice data.

The reading control means registers the special word information obtained by searching the tag information in the special word dictionary, and at the time of converting it into display data, replaces the tag information with the notation information and converts it into voice data. When converting, refer to the special word information registered in the special word dictionary, convert to the correct reading, or replace the reading of the word to which the tag information is added with the phoneme of the tag information, prosodic information, I try to read it out correctly.

Further, the special word information obtained by searching the tag information is not registered in the special word dictionary, and the correct reading is added to the word whose reading is not specified by the same character string as this notation information. .

[0033]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of the present invention. As shown in FIG. 1, the speech synthesis device according to the first embodiment of the present invention reads aloud sentence data and display data from a sentence mixed with Kana and Kanji to which tag information of reading information specifying the reading intended by the sentence creator is added. And a tag information retrieval unit 2 for retrieving tag information from input sentence data containing Kana-Kanji; a special word dictionary 3 for registering notation of tag information for specifying the reading, phoneme, and prosody information. And; a reading buffer 4 for storing text data to be read; a general word dictionary 5 in which phoneme and prosody information of commonly used words is recorded in advance; and reading data stored in the reading buffer 4 for special words Text processing means 6 for converting into a phonetic symbol string using the dictionary 3 and the general word dictionary 5;
A voice synthesis processing means 7 for converting phonetic symbol string data into voice waveform data by using a voice synthesis technique and reading the voice;
A display buffer 8 for storing display data of a sentence mixed with kana-kanji input; display processing means 9 for converting the display data stored in the display buffer 8 into a display format; and displaying the data converted into the display format It includes a display and a display device 10 such as a liquid crystal display device.

Texts mixed with kana and kanji input to the reading control means 1 will be described with reference to FIGS. 3 and 4.
The kana-kanji mixed sentence has a structure in which the sentence creator adds in advance tag information for designating a reading style and tag information for designating another reading style with the same notation. The tag information is composed of a tag enclosed by "<" and ">" indicating the beginning of the tag information and a tag enclosed by "</" and ">" indicating the end of the tag information.

FIG. 3 shows a format of tag information that specifies the reading. The tag indicating the beginning of the tag information includes a character string (WCHG) indicating the type of tag, phoneme of the target word, and prosodic information (for example, Mike) is written. Between the tag showing the beginning and the tag showing the end, the notation (for example, MIC) of the target word is written. FIG. 4 shows a format of tag information that specifies different readings in the same notation, and a tag indicating the beginning of tag information includes a character string (TCHG) indicating tag information that specifies different readings in the same notation, The phoneme and prosody information (eg, tomorrow, tomorrow) of the target word are written. The notation (for example, tomorrow) of the target word is written between the tag indicating the beginning and the tag indicating the end.

The operation of the present invention will be described with reference to FIG. FIG. 2 is a flowchart showing the processing of the first embodiment of the present invention.

First, a sentence mixed with kana-kanji added with the replacement information is input to the reading control means 1. The reading control means 1 supplies the input mixed kana-kanji characters to the tag information searching means 2 and searches for the presence or absence of tag information (steps A1 and A2). When the tag information can be found in the sentence mixed with Kana and Kanji, the type of the found tag information is determined (steps A3 and A4). When the found tag information is the tag information that specifies the reading, the notation, the phoneme, and the prosody information are acquired from the tag information, and the acquired notation, the phoneme, and the prosody information are registered in the special word dictionary 3 (step A5). , A6).

When the found tag information is the same notation that specifies a different reading, the notation of the tag information, the phoneme,
Prosodic information is acquired, and the notation information is replaced with the acquired phoneme and prosodic information and stored in the reading buffer 4 (steps A7 and A8). The found tag information is stored in the display buffer 8 after being replaced with the acquired notation information (step A9).

When the tag information cannot be found in the sentence mixed with Kana-Kanji, the retrieval character string is stored in the reading buffer 4 and the display buffer 8 and it is judged whether or not the retrieval of the inputted sentence mixed with Kana-Kanji has been completed, and the process ends. If not, the process returns to step A1 to perform the search again, and if completed, the reading process ends (steps A15, A16, A).
17).

As described above, in the first embodiment described above, when the tag information designating the reading added by the sentence creator is found in a sentence mixed with Kana / Kanji, the replacement information is registered in the special word dictionary. , If the registered notation information is in a sentence mixed with Kana and Kanji, the registered phoneme,
It reads it by replacing it with prosody information. This eliminates the need for the sentence creator to add special word information for each target word. Further, when tag information designating different readings with the same notation is found in a sentence mixed with Kana-Kanji, it is replaced by the phoneme and prosody information designated by the tag information and read aloud. As a result, a word having the same notation or a word whose reading is specified by the tag information specifying the reading can be read aloud by a plurality of readings intended by the sentence creator.

A second embodiment of the present invention will be described with reference to the drawings. In the second embodiment, the tag information that is added to a sentence mixed with kana and kanji is not registered in the dictionary, and the reading is intended by the sentence creator. FIG. 5 is a block diagram showing the second embodiment of the present invention. As shown in FIG. 5, in the second embodiment of the present invention, the special word dictionary 3 is deleted from the configuration of the speech synthesizer in the first embodiment shown in FIG.

FIG. 6 is a flow chart showing the processing of the second embodiment of the present invention. The operation of the second embodiment of the present invention is the same as that of the first embodiment shown in FIG.
The process after finding the tag designating the reading is different (step A6 in FIG. 1). In this embodiment, when a tag designating the reading is found, special word information of the tag information designating the reading is acquired, and the acquired notation information is searched from the kana-kanji mixed character string. If found, the notation information is replaced with the tag information designating different reading with the same notation (step B1). Step A1 in FIG.
Since the operation of the reading means 1 in the second embodiment shown by -A5 and A7-A12 is the same as the operation in the first embodiment, the description thereof will be omitted.

As described above, in the second embodiment, when the tag information designating the reading in the input mixed Kana-Kanji character is found, the special word information is acquired and the same as this notation information. A character string is searched from a sentence mixed with Kana and Kanji, and this character string is replaced with tag information which is read differently with the same notation. This eliminates the need to register the notation of tag information that specifies the reading, phoneme, and prosody information, and it is possible to delete the special word dictionary,
No need to manage a dictionary.

A third embodiment of the present invention will be described with reference to the drawings. In the third embodiment, even when the user inputs a sentence mixed with kana and kanji that is freely selected, the sentence creator reads the sentence according to the intended reading. FIG. 7 is a block diagram showing the third embodiment of the present invention. Referring to FIG. 7, in the third embodiment of the present invention, the control means 11 and the input device 12 are added to the configuration of the speech synthesizer in the first embodiment shown in FIG. The control unit 11 first reads the header information describing the tag information that specifies the reading method added to the area different from the kana-kanji mixed sentence that can be selected by the user.
The user, and then the user enters an input device 1 such as a keyboard.
The sentence mixed with kana and kanji selected using 2 is input to the reading control means 1.

The operation of the present invention will be described with reference to FIG.
FIG. 8 is a flowchart showing the processing of the third embodiment of the present invention. First, the control means 11 inputs to the reading control means 1 header information stored in a different area from the kana-kanji mixed text to be read. The reading control means 1 determines whether the input data is header information or text data to be read (step C1). When the input data is header information, the input data is supplied to the tag information searching means 2 to search for the presence or absence of tag information (steps C1 and C2). If the tag information can be found in the header information data, the type of the found tag information is determined (steps C3 and C4). When the found tag information is the tag information that specifies the reading, the notation, the phoneme, and the prosody information are acquired from the tag information, and the acquired notation, the phoneme, and the prosody information are registered in the special word dictionary 3 (step C5). , C
6). This process is repeated until the input header information is completed (step C7).

When the processing of the header information input to the reading control means 1 is completed, the control means 11 inputs to the reading control means 1 a kana-kanji mixed sentence to be read.
The operation of the reading means 1 in the third embodiment shown by steps A1-A3 and A9-A12 in FIG. 8 is the same as the operation in the first embodiment, and therefore its explanation is omitted.

In the third embodiment, the tag information designating the reading method is stored in an area different from the kana-kanji mixed sentence to be read and registered in advance in the special word dictionary. As a result, even in the case of a sentence mixed with kana-kanji characters freely selected by the user, it is possible to read aloud as intended by the sentence creator by using the special word dictionary registered in advance.

A fourth embodiment of the present invention will be described with reference to the drawings. A block diagram showing a fourth embodiment of the present invention is shown in FIG. As shown in FIG. 9, in the fourth embodiment of the present invention, the special word dictionary 3 is deleted from the configuration of the speech synthesizer in the third embodiment shown in FIG.

FIG. 10 is a flow chart showing the processing of the fourth embodiment of the present invention. The operation of the fourth embodiment of the present invention is different in the processing after discovering the tag designating the reading method (step C6 in FIG. 8). The processing shown in steps C1-C5 and C7 of FIG. 10 is the same as the processing of the third embodiment of the present invention, and therefore description thereof is omitted. The tag information that specifies the reading method described in the header information that is input from the control unit 11 to the reading control unit 1 is found,
After acquiring the special word information of this tag information (step C
5) Replace the same character string as the acquired notation information in the sentence with mixed kana and kanji to be read out with tag information for different reading with the same notation (step D1). This process is repeated until the input header information is completed (step C
7). Steps A1-A9, A15 and A1 of FIG.
Since 6 is the same as the second embodiment of the present invention, the description thereof is omitted.

In the fourth embodiment, the tag information for specifying the reading method, which is stored in advance in a different area from the kana-kanji mixed sentence to be read out, is acquired and the same in the kana-kanji mixed sentence to be read out. Replace the words in the notation with tag information that reads differently in the same notation. This eliminates the need to register special word information in the dictionary,
Eliminates the need for dictionary management.

[0051]

EXAMPLE An example of the first embodiment of the present invention will be described. FIG. 11 (a) shows an example of a kana-kanji mixed sentence input to the reading control unit 1. The reading control means 1 performs tag information search processing, and registers the notation (tomorrow) of the tag information (WCHG) designating the reading, the phoneme and the prosody information (tomorrow) in the special word dictionary 3 (step A).
6). For the prosody information, for example, a special character may be added to the first character such as "A'shita" to indicate an accent.

FIG. 11 (b) shows the display data supplied from the reading control means 1 and stored in the display buffer 8. The data stored in the display buffer 8 is the same as that embedded in a kana-kanji mixed sentence. It indicates that the tag information (TCHG) that specifies another reading in the notation is replaced with the notation information (tomorrow) of this tag.

This display data is supplied to the display control means 9, determines the display position and display size of the character string, generates the character string at the time of display, and supplies it to the display device 10 for display.

FIG. 11 (c) shows voice data output from the voice synthesizing means 7, and a tag for designating another reading by the same notation embedded in the inputted mixed Kana / Kanji sentence of FIG. 11 (a). The information (TCHG) is replaced by phoneme and prosody information (tomorrow) in the rewriting processing means 1, and further, the text processing means 9 refers to the special word dictionary 3 to refer to the special word dictionary 3 and the same notation as the tag information (WCHG) that specifies the reading. It shows that the word with information (tomorrow) is replaced with the phoneme and prosody information (tomorrow) of the special word information.

An example of the second embodiment of the present invention will be described. FIG. 12A shows an example of a sentence containing kana-kanji characters input to the reading control unit 1.

FIG. 12B shows data obtained by replacing the same character string as the notation information (tomorrow) of the tag information (WCHG) designating the reading method with the tag information (TCHG) designating another reading method with the same notation. , The phonological and prosody information (tomorrow) of this data is tag information (WCHG) that specifies the reading method.
Using the phoneme and prosody information acquired from the above, the phoneme and prosody information (tomorrow) of the tag information that specifies a different reading with the same notation added in advance is not replaced.

FIG. 12C shows the display data supplied from the reading control means 1 and stored in the display buffer 8, and the tag information (TCHG) of the input kana-kanji mixed sentence data shown in FIG. 12A. ) Is replaced by the notation information (tomorrow) of this tag information.

FIG. 12D shows the voice data output from the voice synthesizing means 7, and shows that the tag information (TCHG) is replaced with phoneme and prosody information (tomorrow, tomorrow, etc.).

An example of the third embodiment of the present invention will be described. FIG. 13A shows an example of header information input to the reading control unit 1. In the header information, for example, tag information (WCHG) designating how to read is described. In addition to this, other tag information such as display-related tag information may be included. Reading control means 1
Performs tag information search processing of header information and registers the phoneme and prosody information of the acquired tag in the special word dictionary 3.

FIG. 13B shows an example of text data mixed with kana and kanji to be read aloud.

FIG. 13C shows display data supplied from the reading control means 1 and stored in the display buffer 8.
It is indicated that the tag information (TCHG) added to the kana-kanji mixed sentence data shown in FIG. 13B is replaced with the tag notation information (tomorrow, today).

FIG. 13D shows voice data output from the voice synthesizing means 7. Tag information (TCHG) designating different readings with the same notation is phoneme and prosody information (Myoni, Tomorrow, Konnichi). To the reading of words with the same notation as the tag information (WCHG) that specifies the reading.
Indicates that the tag has been replaced with the phoneme and prosody information (tomorrow, today).

An example of the fourth embodiment of the present invention will be described. FIG. 14A shows an example of header information input to the reading control unit 1, and FIG. 14B shows an example of text data mixed with Kana and Kanji to be read. The header information includes, for example, tag information (WC
HG) and the like are described. In addition to this, other tag information such as display-related tag information may be included. The reading control unit 1 performs a tag information search process of the header information, and acquires special word information of tag information (WCHG) that specifies the reading method. The same character string as the acquired notation information (tomorrow, today) is searched from a sentence mixed with kana-kanji characters shown in FIG. 14B, and replaced with tag information (TCHG) that specifies different reading with the same notation.

FIG. 14C shows data obtained by converting the target word in a sentence mixed with kana and kanji to be read into tag information (TCHG) designating the reading method using the header information. This data is supplied to the reading control means 1 and subjected to tag search processing to be converted into display data and reading data.

FIG. 14D shows display data stored in the display buffer 8, and tag information (T) for designating the reading embedded in the kana-kanji mixed sentence of FIG. 14B.
CHG) is converted into the notation information (tomorrow and today) of the tag.

FIG. 14 (e) shows voice data output from the voice synthesizing means 7, and tag information for designating another reading method with the same notation embedded in the kana-kanji mixed sentence of FIG. 14 (b) ( TCHG) is converted into the phoneme and prosody information of the tag (Myoni, Tomorrow, and Hi),
The phoneme and prosody information of the words (tomorrow and today) having the same notation as the tag information (WCHG) that specifies the reading of the header information in FIG. 14A is the phoneme and prosody information of the tag described in the header information ( Tomorrow and today).

[0067]

The first effect is that even when there are a plurality of special words that have the same reading in a kana-kanji mixed sentence, the sentence creator adds the special word information such as reading and accent to the sentence in advance when creating the sentence. By keeping
There is no need to add special word information every time it exists.

The reason is that the special word information of the tag information that specifies the reading of a plurality of words having the same reading is acquired,
This is to register in the special word dictionary and replace the part corresponding to the special word with the correct phoneme and prosody information registered in the special word.

The second effect is that even if there is a special word which is read differently in the same notation in a sentence mixed with kana and kanji, the sentence creator prepares the special word information such as the reading and accent in the sentence at the time of writing the sentence. By embedding, each special word can be read out in multiple ways.

The reason is that the special word information of the tag information that specifies the reading of a plurality of readings with the same notation is acquired, and the portion corresponding to the special word is replaced with the correct phoneme and prosody information of the special word information. Is.

The third effect is that a dictionary for registering special word information is not required.

The reason is that, when the sentence creator obtains the special word information added to the sentence in advance at the time of writing the sentence, the portion corresponding to the special word of the input sentence data is replaced with the correct phonological and prosodic information of the obtained special word information. This is because.

The fourth effect is that the text data freely selected by the user is read aloud with correct phoneme and prosody information.

The reason is that when the sentence creator acquires the special word information added to the sentence at the time of writing the sentence and registers it in the special word dictionary in advance, and the user reads the sentence mixed with kana and kanji freely selected by the user. This is because the registered special word dictionary is replaced with the correct phoneme and prosody information.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating the operation of the first exemplary embodiment of the present invention.

FIG. 3 is a diagram showing an example of tag information that specifies how to read special words.

FIG. 4 is a diagram showing an example of tag information designating different readings with the same notation.

FIG. 5 is a block diagram showing a second embodiment of the present invention.

FIG. 6 is a flowchart for explaining the operation of the second embodiment of the present invention.

FIG. 7 is a block diagram showing a third embodiment of the present invention.

FIG. 8 is a flowchart for explaining the operation of the third exemplary embodiment of the present invention.

FIG. 9 is a block diagram showing a fourth embodiment of the present invention.

FIG. 10 is a flow chart for explaining the operation of the fourth embodiment of the present invention.

FIG. 11 is a diagram for explaining an example of the first embodiment of the present invention.

FIG. 12 is a diagram for explaining an example of the second embodiment of the present invention.

FIG. 13 is a diagram for explaining an example of the third exemplary embodiment of the present invention.

FIG. 14 is a diagram for explaining an example of the fourth embodiment of the present invention.

FIG. 15 is a block diagram showing an example of a conventional speech synthesizer.

FIG. 16 is a diagram showing an example of a kana-kanji mixed sentence in which special words are embedded.

FIG. 17 is a block diagram showing an example of a conventional speech synthesizer.

FIG. 18 is a diagram showing an example of a kana-kanji mixed sentence to which special words are added.

[Explanation of symbols]

 1 reading control device 2 tag search means 3 special word dictionary 4 reading buffer 5 general word dictionary 6 text processing means 7 voice synthesis processing means 8 display buffer 9 display control means 10 display device 11 control means 12 input means 101 special word buffer 102 special Word extraction unit 103 Language processing unit 104 Special word conversion unit 105 Speech synthesis unit 111 Special word extraction unit 112 Special word dictionary 113 Special word dictionary creation unit 114 General word dictionary 115 Language processing unit 116 Speech synthesis unit

Claims (7)

[Claims] 1. A tag search means for searching tag information describing phoneme, prosody information, etc. of a special word embedded in a sentence to be processed, and a special word of the special word based on an analysis result of the detected tag information. A speech synthesis apparatus comprising: a reading control means for generating phonological and prosody information. 2. The voice synthesizing apparatus according to claim 1, further comprising a reading control means for replacing the special word to which the tag information is added with the notation information of the tag and storing it in the display buffer. 3. The speech synthesis apparatus according to claim 1, further comprising a reading control means for replacing the special word to which the tag information is added with the phoneme and prosody information of the tag and storing it in the reading buffer. 4. Obtaining special word information of the detected tag,
4. The speech synthesizer according to claim 3, further comprising reading control means for registering the acquired notation, phoneme, and prosody information of the special word information in the special word dictionary. 5. A reading control means for detecting a character string that matches the detected notation information of the tag from a sentence to be processed and replacing the notation information of the matched character string with phoneme and prosody information and storing it in a reading buffer. The speech synthesizer according to claim 3, wherein 6. The speech synthesis according to claim 4, further comprising text processing means for converting the reading sentence data stored in the reading buffer into a phonetic symbol string by referring to the special word dictionary and the general dictionary. apparatus. 7. Control means for separately supplying data of only tag information describing phonemes, prosodic information, etc. of a special word and a sentence to be processed in which special word information is embedded to a reading control means, and tag information. 7. The speech synthesizer according to claim 5, further comprising: a reading control unit that acquires tag information from the data of only the data and registers the acquired notation, phoneme, and prosody information of the special word payment in the special word.