JP3280729B2 - Phonetic symbol creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a phonetic symbol creation device used in a text-to-speech synthesis system or the like, for converting a sentence mixed with kanji kana into a phonetic symbol string.

[0002]

2. Description of the Related Art In recent years, a text-to-speech synthesis system has been developed in which a Japanese sentence containing kanji and kana is subjected to linguistic analysis, converted into a phonetic symbol string containing reading and accent information, and then output as a synthesized speech by a speech synthesizer. The applications of this type of text-to-speech synthesis system include text-to-speech and text-to-speech editing, and the so-called online use of converting characters displayed on the screen of a personal computer into speech and outputting them. There is a so-called off-line application in which a phonetic symbol string is prepared in advance, such as a voice manual or a telephone service using synthetic voice, and the same sentence is repeatedly output using the generated voice symbol string.

[0003] In the case of online use, it is sufficient if the contents of a sentence can be conveyed by voice as a whole, so there is not much problem even if the way of reading does not suit the user's wish somewhat.
When the same sentence is repeatedly heard, such as in an off-line application, it is necessary to eliminate reading errors and to make the reading such as the pause position match the user's desire. However, input documents have various ambiguities such as homomorphic words (words with the same notation but different readings) and differences in how to read numbers (stick reading, digit reading, etc.). May not always be as desired by the user.

In order to correct a portion that does not satisfy the user's desire, a conventional method has been proposed in Japanese Patent Application Laid-Open No. Hei 2-24495.
As disclosed in the issue, a method of inserting a phonetic symbol string in the input Kanji kana mixed sentence in advance, or correcting the final created phonetic symbol string with a normal text editor Methods have been proposed.

[0005]

However, in the method of inserting a phonetic symbol string in a kanji kana-mixed sentence to be input in advance, a phonetic symbol string automatically generated by the system from a kanji kana-mixed sentence is inserted in advance. There is a problem in that it is difficult to obtain consistency with the phonetic symbol string that has been used, and it is difficult to reflect user's requirements very finely. Also,
In the method of modifying the final generated phonetic symbol string with an ordinary text editor, the user needs to be proficient in the definition of the phonetic symbol string in order to output the desired sound, and the burden on the user is reduced. There was a problem of being big.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a phonetic symbol creation device capable of creating a phonetic symbol string that meets a user's request in a detailed manner without imposing a burden on the user. And

[0007]

In order to achieve the above object, according to the present invention, a language analysis means, a language analysis result storage means, a phonetic symbol generation means, A user interface control means for controlling the voice synthesizing means is provided, and the user interface means can actively support a phonetic symbol creation operation. As a result, it is possible to determine a user's request and reflect it in detail, and to create a phonetic symbol string that meets the user's request without imposing a burden on the user. Also, language analysis
Specify the word information of a specific word in advance before performing
With the provision of word information designating means, language
The ambiguity at the time of analysis can be reduced, and post-editing work can be reduced.

Further, the invention according to claim 2 is provided with division position designation means for designating various division positions in advance before performing language analysis. This can reduce ambiguity during language analysis and reduce post-editing work.

The invention according to claim 3 is provided with a language analysis method designating means for designating which language analysis method is to be used for a portion where a plurality of language analysis methods are considered before performing the language analysis. ing. This can reduce ambiguity during language analysis and reduce post-editing work.

[0010]

Further, the invention according to claim 4 is provided with a speech synthesizing means for outputting the contents of the language analysis result storage buffer from any position designated by the user as synthesized speech in various utterance segments. As a result, the user can easily find the corrected portion in the text.

Further, the invention according to claim 5 is provided with re-language analysis activation means for re-linguing a part of the input sentence and updating a part of the contents of the linguistic analysis result storage buffer. Thus, the burden on the user in post-editing the phonetic symbols can be reduced.

According to a sixth aspect of the present invention, there is provided a search means for searching the word information in the linguistic analysis result storage buffer by using a part of the word information such as a word notation, a part of speech, and a reading as a key. I have. As a result, the user can easily find a portion where an error similar to the once corrected portion has occurred.

Further, the invention according to claim 7 is provided with a replacing means for replacing word information at a designated position in a language analysis result storage buffer with designated contents. This allows
It is possible to easily correct a portion where an error similar to the once corrected portion occurs.

[0015]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a phonetic symbol generation device according to the present invention. With reference to FIG. 1, this phonetic symbol creation device performs a linguistic analysis of a Japanese sentence containing kanji and kana, and converts the sentence into a sequence of word information including reading, part of speech, accent type, word delimiter information, and the like. And a linguistic analysis result storage buffer 2 for storing the analysis result of the linguistic analysis unit 1 and a part or all of the contents of the linguistic analysis result storage buffer 2 as input to perform accent combining processing and the like to convert them to phonetic symbol strings. A phonetic symbol generation unit 3, a speech synthesis unit 4 that generates a synthesized voice based on a phonetic symbol string generated by the phonetic symbol generation unit 3, and, in response to an instruction from a user,
It has a language analysis unit 1, a language analysis result storage buffer 2, a phonetic symbol generation unit 3, and a user interface control unit 5 for controlling the speech synthesis unit 4. Here, it is assumed that the language analysis unit 1 includes a word dictionary and other grammatical rules.

FIG. 2 is a diagram showing a configuration example of the user interface control unit 5. In the configuration example of FIG. 2, the user interface control unit 5 basically includes the edit screen 10
And a pre-editing unit 11 and a post-editing unit 12. Based on the user's instruction from the edit screen 10, a linguistic analysis unit 1, a linguistic analysis result storage buffer 2, a phonetic symbol generation unit 3, a voice The combining unit 4 is controlled. FIG. 3 shows a configuration example of the edit screen 10.
In the example of FIG. 3, the edit screen has a menu section MN for allowing the user to make various designations, and a text display section CG for allowing the user to confirm.

The pre-editing section 11 specifies the delimiter position to the language analysis section 1 before the language analysis section 1 analyzes the language of the Japanese sentence containing kanji and kana.
The language analysis method and the word information of a specific word can be specified. That is, the pre-editing unit 11 allows the user to set various delimiter positions such as the position of a pause and the position of a break of an accent phrase before the language analysis is performed, and gives the information to the language analysis unit 1. The user designates which language analysis method is to be used for a part such as a numeral or a symbol, for which a plurality of language analysis methods can be considered according to the contents of the sentence, and specifies this information to the language analysis unit 1. The given language analysis method designating unit 22 and the specific word information designating unit 23 that allows the user to specify word information of a specific word such as a word or proper noun for which a homonym is predicted, and gives this information to the language analysis unit 1 Have.

When the above-mentioned designation information is given by the pre-editing unit 11 of the user interface control unit 5, the language analyzing unit 1 executes the language designated by the pre-editing unit 11 in accordance with the designation. An analysis is performed, and a part not particularly specified by the preediting unit 11 is subjected to a linguistic analysis as usual, and a word composed of a word notation pointer, a notation length, a reading, an accent type, a part of speech, and word delimiter information. A sequence of information is created and stored in the linguistic analysis result storage buffer 2.

The post-editing unit 12 of the user interface control unit 5 outputs a synthesized speech in various utterance categories from an arbitrary position designated by the user as a synthesized speech from the contents stored in the language analysis result storage buffer 2. When the user specifies a partial range to be corrected among the contents stored in the linguistic analysis result storage buffer 2, the range of the input sentence corresponding to this partial range is transmitted to the linguistic analysis unit 1 again. Language analysis, language analysis result storage buffer 2
Re-language analysis activation unit 26 for updating the content of the word, and a search unit 27 for searching word information in the language analysis result storage buffer 2 using a part of word information such as notation, part of speech, and reading as keys.
And a replacement unit 28 that replaces word information at a specified position in the language analysis result storage buffer 2 with specified content.

Next, the processing operation of the phonetic symbol generation device having such a configuration will be described. The phonetic symbol creation device of this embodiment is basically based on the premise that phonetic symbol creation work is performed in the order of pre-editing, language analysis, and post-editing. Prior to causing the language analysis unit 1 to analyze the language of a Japanese sentence mixed with kanji and kana, the user first uses the edit screen as shown in FIG. The analysis unit 1 can specify a break position, a language analysis method, and word information of a specific word. Specifically, to specify the break position, move the cursor to a desired position of the text displayed on the character display section CG of the edit screen, and open the “pause” menu or the “accent phrase” menu of the menu section MN. In this case, the delimiter position specification unit 21 recognizes the position of the cursor as a pause or a break of an accent phrase, and gives the recognition result to the language analysis unit 1 as delimiter position information.

The language analysis method can be designated by selecting the "language analysis designation" menu in the menu section MN of the edit screen. Assuming that there are two types of analysis methods, "stick reading" or "digit reading" for numbers, and "not read" or "speak as symbols" for symbols, the "language analysis designation" menu is selected. And a language analysis method designation unit 22
Automatically searches for a number or symbol part after the current cursor position in a sentence and asks the user how to analyze that part. At this time, the user may select an analysis method that suits his or her wish. For example, digit reading is selected by default if the user does not specify the number analysis method, and "not read" is selected unless the user specifies the symbol analysis method. The language analysis method designated and selected in this way is provided from the language analysis designation unit 22 to the language analysis unit 1.

The specification of a specific word is performed by the user specifying a range of words and first selecting the "word specification" menu in the menu section MN of the edit screen. In this specification, the word information is limited in advance for words for which the post-editing work is considered to be easier if the word information is specified in advance, such as words that are expected to have the same form and proper nouns. Is made if it is good. When the user specifies a word range and selects the “word specification” menu, the specific word information specifying unit 23 searches a dictionary (not shown) for a word candidate corresponding to the notation in the range, and edits the screen. The word candidate display window is displayed on the character display section CG of. FIG. 4 is a diagram showing an example of a word candidate display window when "January" is designated as the word notation.
In this example, when the word notation “January” is specified by the user, the specific word information specifying unit 23 determines that the word candidate corresponding to this notation is the first candidate having the pronunciation “Ichigatsu” and the reading The second candidate of “one-touch” is displayed in the window. If there is a desired one, the user selects and confirms the candidate. If there is no desired one, the user inputs a word reading, an accent, and a part of speech in a user input field of the word candidate display window, and inputs a specific word. You can specify. The specific word specified and selected in this way is provided from the specific word information specification unit 23 to the language analysis unit 1.
By performing pre-editing as described above, ambiguity during language analysis is reduced, and the load of post-editing work can be reduced.

After the pre-editing operation is performed, the user can activate the language analysis section 1 by selecting the "language analysis" menu. The linguistic analysis unit 1 performs a linguistic analysis on the part specified in the pre-editing according to the specification, and performs a linguistic analysis on the part not specified in the pre-editing.
Perform normal linguistic analysis to find word pointers,
A string of word information including the length of the notation, reading, accent type, part of speech, and word delimiter information is created and stored in the linguistic analysis result storage buffer 2. Here, the notation pointer is a pointer indicating which notation part in the sentence this word information corresponds to, and the word delimiter information means that this word becomes a break of a phrase or an accent. It refers to information such as a break in a phrase or a pause position.

After the input sentence is subjected to the linguistic analysis as described above, and the result is stored in the linguistic analysis result storage buffer 2, the user can perform post-editing based on the result. In the post-editing, the user can activate the synthesized speech output unit 25, convert the contents of the language analysis result storage buffer 2 into synthesized speech, and listen to the ear to search for a portion that needs correction. At this time, the reading can be started from an arbitrary position designated by the user, and can be stopped at an arbitrary position. Specifically, the user can specify the reading start position by moving the cursor to a desired position of the text displayed on the character display section CG. The user specifies the reading start position,
When the "voice output" menu is selected, the synthesized voice output unit 2
5 starts reading aloud from a specified reading start position in a specified utterance section. There are four types of utterance classification for reading aloud: a punctuation according to phonetic symbols, a punctuation according to accent punctuation, a punctuation according to a phrase, and a punctuation according to a word. This can be selected in the "Change utterance category" menu. Also,
With the “forward” menu and the “backward” menu, the reading start position can be moved back and forth in units of speech segments. The user first finds the approximate position of the portion that needs to be corrected while listening to the voice in the utterance segment according to the phonetic symbols, then finely divides the utterance segment in the vicinity and re-listens to determine the portion to be corrected. Can be easily found.

When the partial range to be corrected is determined and the user selects the "language analysis" menu, the re-linguistic analysis activation unit 26 activates the language analysis unit 1 again, and determines the partial range to be corrected. And the next candidate can be selected. At this time, the re-language analysis activation unit 26 causes the re-analysis range to be analyzed as a phrase at first, and displays the analysis result candidates in the word candidate display window in the same manner as described above, but the desired result is still obtained. If not, the range can be narrowed further and the range of the word can be specified for analysis. If a word with a specified range of notation does not exist in the dictionary, the user is requested to input the pronunciation, accent type, and part of speech of the word. At this time, the user
What is necessary is just to input these information in a predetermined field. At this time, it is also possible to additionally register this word in the dictionary. When the linguistic analysis result is determined in the linguistic analysis unit 1, the linguistic analysis result is sent to the linguistic analysis buffer 2, and the contents of the linguistic analysis result storage buffer 2 are updated.

As described above, the correcting operation can be repeatedly performed until there is no portion to be corrected.
The same word or phrase is often used in multiple places in the same sentence. In the phonetic symbol generation device of the present embodiment, the search unit 27 further searches the word information in the linguistic analysis result storage buffer 2 using a part of the word information such as the word notation, part of speech, and reading as a key. In addition, the replacement unit 28 can be used to replace the word information or the string of word information in the searched portion with the specified word information or the string of word information. Thus, when the same word or phrase is used in a plurality of places in the same sentence, the work of correcting this part can be performed more easily. That is, when the “search” menu is selected, the search unit 27 first displays a search window on the character display unit CG. FIG. 5 shows an example of the search window.
This search window contains notation, reading,
A field for inputting a part of speech is provided. When the user inputs any combination of these keys and presses return, the search unit 27 matches the search key located behind the current cursor position. Move the cursor to the beginning of a word. At this time, when the user selects the “replace” menu, the replacement unit 28 inquires, for example, how to perform replacement, and in response to this inquiry, Is specified, the word information or the column of the word information at the current cursor position is replaced with the specified word information or the column of the word information. However, replacement is performed only when the notation of the replaced portion and the notation of the range as an example after the replacement completely match. By performing such a search and replacement process, it is possible to easily correct a portion where an error similar to the once corrected portion occurs.

As described above, when the synthesized speech that meets the user's desire can be finally obtained from the synthesized speech output unit 25, the user selects the "save" menu to execute the language analysis. The contents of the result storage buffer 2 can be stored in a file. By selecting the "pronunciation symbol output" menu, the utterance symbol generation unit 3 is activated, the contents of the linguistic analysis result storage buffer 2 are converted into a pronunciation symbol string, and the result can be output to a file.

[0028]

As described above, according to the first aspect of the present invention, the language analysis means, the language analysis result storage means, the phonetic symbol generation means, and the speech synthesis means are controlled in accordance with an instruction from the user. User interface control means is provided, and the user interface means can actively support the phonetic symbol creation work, so that the user's request can be reflected in detail and a phonetic symbol string that meets the user's wishes can be provided to the user. It can be created without much burden. In addition, before performing linguistic analysis, the specific word
A specific word information specifying means for specifying word information is provided;
This reduces ambiguity during language analysis and reduces post-editing work.
Can be reduced.

According to the second aspect of the present invention, since the delimiter position specifying means for specifying various delimiter positions is provided in advance before performing the language analysis, the ambiguity at the time of the language analysis can be reduced. Editing work can be reduced.

According to the third aspect of the present invention, a language analysis method designating means for designating which language analysis method is to be used for a part in which a plurality of language analysis methods are considered before performing language analysis. Since it is provided, ambiguity during language analysis can be reduced, and post-editing work can be reduced.

[0031]

According to the fourth aspect of the present invention, there is provided a speech synthesizing means for outputting the contents of the linguistic analysis result storage buffer from any position designated by the user as synthesized speech in various utterance segments. Therefore, it is possible for the user to easily find the corrected portion in the text.

According to the fifth aspect of the present invention, there is provided a re-language analysis activation means for re-linguing a part of the input sentence and updating a part of the contents of the linguistic analysis result storage buffer. Therefore, the burden on the user in post-editing the phonetic symbols can be reduced.

According to the sixth aspect of the present invention, there is provided a search means for searching the word information in the linguistic analysis result storage buffer using a part of the word information such as the word notation, part of speech, and reading as a key. Therefore, the user can easily find a portion where an error similar to the once corrected portion has occurred.

According to the seventh aspect of the present invention, the replacement means for replacing the word information at the specified position in the linguistic analysis result storage buffer with the specified content is provided, so that the once corrected portion The same error can be easily corrected.

[Brief description of the drawings]

FIG. 1 is a block diagram of an embodiment of a phonetic symbol generation device according to the present invention.

FIG. 2 is a diagram illustrating a configuration example of a user interface control unit.

FIG. 3 is a diagram showing a configuration example of an edit screen.

FIG. 4 is a diagram illustrating an example of a word candidate display window.

FIG. 5 is a diagram illustrating an example of a search window.

[Description of Signs] 1 Language analysis unit 2 Language analysis result storage buffer 3 Phonetic symbol generation unit 4 Voice synthesis unit 5 User interface control unit 10 Edit screen 11 Pre-editing unit 12 Post-editing unit 21 Delimiter position specifying unit 22 Language analysis method specification Unit 23 specific word information designating unit 25 synthesized speech output unit 26 relanguage analysis activation unit 27 search unit 28 replacement unit

Claims (7)

(57) [Claims] 1. A linguistic analysis means for linguistically analyzing a Japanese sentence mixed with kanji and kana, and converting it into a sequence of word information including reading, part of speech, accent type, word delimiter information, and the like, and an analysis result of the linguistic analysis means Linguistic analysis result storage means for storing a part of or all of the contents of the linguistic analysis result storage means, and performing accent connection processing or the like to convert it into a phonetic symbol string; A voice synthesizing means for generating a synthesized voice based on the phonetic symbol sequence generated by the means, and controlling a language analyzing means, a language analysis result storing means, a phonetic symbol generating means, and a voice synthesizing means in accordance with an instruction from a user. User interface control means , wherein the user interface
The language control means performs language analysis in the language analysis means.
Before the user, the word information of a specific word is given to the user in advance.
Specific word information to be given to the language analysis means
A phonetic symbol creation device characterized by having means . 2. The phonetic symbol creation device according to claim 1, wherein the user interface control means allows the user to designate various break positions in advance before the language analysis is performed by the language analysis means. A phonetic symbol creation device characterized by having a delimiter position designating means provided to the means. 3. The phonetic symbol creation device according to claim 1, wherein the user interface control means determines a part where a plurality of language analysis methods are considered in advance before the language analysis is performed by the language analysis means. A phonetic symbol creation device comprising a language analysis method designating means for allowing a user to designate a language analysis method and giving the language analysis means. 4. A language analysis of a Japanese sentence containing kanji and kana
Reading, part of speech, accent type, word delimiter information, etc.
Language analysis means for converting into a sequence of word information containing
Language analysis result storage means for storing analysis results of word analysis means
And part or all of the contents of the language analysis result storage means.
To perform accent connection processing etc.
A phonetic symbol generation means for converting the phonetic symbol into a symbol string;
Synthesized speech based on the phonetic symbol sequence generated by the synthesizing means
Voice synthesizing means for generating
Language analysis means, language analysis result storage means, phonetic symbol generation
Forming means, a user interface for controlling the speech synthesizing means
The user interface control means , wherein the user interface control means has a speech synthesis means for outputting the contents of the linguistic analysis result storage means from any position designated by the user as synthesized speech in various utterance segments. A phonetic symbol creation device characterized by the following. 5. The phonetic symbol creation device according to claim 4 , wherein the user interface control means causes the language analysis means to re-language a part of the input sentence.
A phonetic symbol creation device comprising re-language analysis activation means for updating a part of the contents of the language analysis result storage means. 6. A language analysis of a Japanese sentence containing kanji and kana
Reading, part of speech, accent type, word delimiter information, etc.
Language analysis means for converting into a sequence of word information containing
Language analysis result storage means for storing analysis results of word analysis means
And part or all of the contents of the language analysis result storage means.
To perform accent connection processing etc.
A phonetic symbol generation means for converting the phonetic symbol into a symbol string;
Synthesized speech based on the phonetic symbol sequence generated by the synthesizing means
Voice synthesizing means for generating
Language analysis means, language analysis result storage means, phonetic symbol generation
User interface for controlling the
And a scan control unit, the user interface control means, the word information in the language analysis result storage unit, notation of words, parts of speech, read retrieval means part to search the key word information, such as A phonetic symbol creation device characterized by having: 7. The pronunciation symbol generating apparatus according to claim 6 , wherein said user interface control means has a replacement means for replacing word information at a specified position in said linguistic analysis result storage means with specified contents. A phonetic symbol creation device characterized by the following.