JP3102636B2 - Notation string conversion device and notation string conversion method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To more precisely recognize voice. SOLUTION: A notation character string outputted from a voice data conversion means 5 is stored in a non-decided character string storage means 12 in a non-decided state and a read character string is stored in a read character string storage means 11. A language rule storage means 17 stores a language rule. An automatic conversion means 15 judges the relation of the specified notation character string in a plurality of notation character strings stored in the non-decided character string storage means 12 based on the language rule of the language rule storage means 17. When the relation of the notation character strings is judged to match the language rule, the notation character string in the non-decided state is changed to match the language rule. A display means 19 displays the character string of the non-decided character string storage means. Thus, precise recognition can be executed by changing recognized voice data based on the language rule.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition technology, and more particularly, to an improvement in the accuracy of conversion of a written character string.

[0002]

2. Description of the Related Art To input a character string to a computer, a voice input system as shown in FIG. In the speech conversion dictionary, a feature amount of speech data and a notation character string corresponding to the feature amount are registered in advance as a table. When a voice is input from the microphone 61, the voice conversion unit 63 extracts a predetermined feature amount from the input voice and calculates a matching with a feature amount of each written character string registered in the voice conversion dictionary 65. Then, the one with the highest matching degree is output as a notation character string.

However, the above-mentioned voice input system has the following problems. Since the matching degree is determined based on the feature amount of the voice data, if voice data other than the registered text string is given, the voice data cannot be converted into a text string desired by the operator.

[0004] For this reason, it is conceivable to register many notation character strings. However, as the number of written character strings to be registered increases, the matching operation takes more time, and even if the pronunciation method is slightly shifted, it is recognized as another conversion candidate. For this reason, the conversion accuracy may be reduced.

[0005] In particular, the speech conversion dictionary of the speech input system stores the characteristic amount of a notation character string that is statistically frequently used, so that a sentence in an unusual specialized field may be erroneously converted. High in nature.

The present invention solves the above-mentioned problem, and provides a typographical character string conversion apparatus and method capable of improving conversion efficiency and converting to an unregistered typographical character string without making the speech conversion dictionary so large.

[0007]

Means for Solving the Problems and Effects of the Invention 1) The notation character string conversion device according to the present invention is characterized by the characteristics of voice data.
The notation character string corresponding to the quantity and the reading
By using a speech conversion dictionary that stores only character strings
The input voice data is recorded in the voice conversion dictionary.
To the notation character string that best matches the
The notation string given from the voice data conversion means for output together with the read character string, a writing character string converter for converting to match the language rules, 1) language correspondence between specific writing character string According to the rules
A corresponding reading sentence that is stored and corresponds to the notation character string
Language rule storage means for storing character strings together ; 2) undetermined character string storage means for storing the notation character string provided from the voice data conversion means in an undetermined state; 3) provided from the voice data conversion means. Reading character string storage means for storing the read character string, and whether or not the relationship between specific notation character strings among the plurality of notation character strings stored in the undetermined character string storage means matches the correspondence relationship Judge
And if not, read the character string
The specific notation of the reading character string stored in the storage means;
The reading character string corresponding to the character string and the corresponding reading character string
Match the specific notation string with the correspondence based on
Automatic conversion means for converting into a notation character string .

[0008] As described above, using the speech conversion dictionary,
Character strings and reading character strings with a high degree of
If given by the conversion means, this notation character string
And memorize the reading character string,
The relationship of a specific notation string among the plurality of notation strings
Judgment based on language rules stored in advance
Judge that the relationship between strings does not conform to the language rules
If not, the written character string in the undetermined state is
Change to match the rule. Therefore, many notations
Even if there are no speech conversion rules for character strings,
A notation character string that matches the rule can be obtained. Also before
Considering the written string, the relationship between the specific written string
And change the notation character string. This
As described above, by considering the reading string,
You can make judgments and make changes.

2) A notation character string conversion method using a computer according to the present invention is a notation character string conversion method for converting a notation character string provided from voice data conversion means so as to conform to language rules. The audio data conversion means includes a notation character string corresponding to a feature amount of the audio data and the table.
A speech conversion dictionary that stores the reading character string corresponding to the written character string
The input voice data can be
Notation characters that best match the features recorded in the voice conversion dictionary
By converting to a column and outputting it with the corresponding reading character string
The computer executes the following processing, and determines the correspondence between specific notation character strings based on language rules.
And the corresponding reading corresponding to the notation character string.
A character string is also stored, and the notation character string provided from the voice data conversion means is stored in an undetermined state, and a read character string provided from the voice data conversion means is stored in the undetermined state. The relationship of a specific notation character string among the plurality of notation character strings stored in the step matches the correspondence.
Judge whether or not the character is read, and if not , read the character given by the voice data conversion means.
The specific notation sentence based on a column and the corresponding reading character string
The character string is converted into a written character string that matches the correspondence .
Therefore, it is possible to obtain a written character string that conforms to the language rules even if there are no speech conversion rules for many written character strings. In addition, the determination regarding the relationship of the specific written character string and the change of the written character string are performed in consideration of the read character string. As described above, the determination and the change can be performed more accurately by considering the read character string.

[0010] 4,5) in notation string conversion device or method according to the present invention, the language rules storing means, the writing character string having a reading similar <br/> stored for a read character string The automatic conversion means outputs the undetermined character string.
A specific table among a plurality of notation character strings stored in the storage means
If it is determined that the relationship between the character strings does not match the correspondence , the character string to be automatically converted is converted to the character string having the similar pronunciation. Therefore, even if a slightly deviated reading is given, an appropriate written character string can be obtained.

The automatic conversion means corresponds to the processing of step ST11 of the CPU 23 in the embodiment.
Further, in the embodiment, it is determined whether or not the character string matches the Kana-Kanji conversion rule without referring to the reading character string in the reading buffer 27b. However, the reading character string may be referred to.

In the embodiment, the manual conversion means performs step ST17, step ST21, step ST25, step ST27, step ST29, step ST3 performed by the CPU 23 according to the type of key operation by the operator.
1. This corresponds to the process of step ST33.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Description of Functional Block Diagram One embodiment of the present invention will be described with reference to the drawings. FIG.
2 shows a functional block diagram of the notation character string conversion device 10.
The notation character string conversion device 10 converts the notation character string given from the sound data conversion means 5 that converts the voice data input by the operator into the corresponding notation character string and reading character string and outputs the converted character string so as to match the language rule. The apparatus includes a language rule storage unit 17, an undetermined character string storage unit 12, an automatic conversion unit 15, a read character string storage unit 11, a manual change unit 13, and an output unit 14.

First, a description will be given of the voice data conversion means 5 for providing a written character string to the written character string conversion device 10. When the operator inputs a voice to the voice input unit 3, the voice conversion unit 6 extracts a predetermined feature amount from the input voice data. A plurality of conversion candidates are stored in the voice conversion dictionary 7 in advance. Each conversion candidate is composed of a feature amount, its notation character string, and its reading character string. The voice conversion means 6 determines the degree of coincidence between the extracted feature quantity and the feature quantity with each of the conversion candidates, and outputs a notation character string and a reading character string of the best matching candidate.

The written character string output from the voice data conversion means 5 is stored in the undetermined character string storage means 12 in an undetermined state, and the read character string is stored in the read character string storage means 11. The language rule storage means 17 stores a language rule. In the present embodiment, a conversion dictionary for kana-kanji conversion is adopted as a language rule. More specifically, co-occurrence information composed of a notation character string having a co-occurrence relation and a reading character string thereof is employed.

The automatic conversion means 15 determines the relation of a specific notation character string among a plurality of notation character strings stored in the unconfirmed character string storage means 12 based on the language rules of the language rule storage means 17, If it is determined that the relationship between the notation character strings does not match the language rule, the notation state of the notation character string is changed to match the language rule.

In this embodiment, the automatic conversion means 15
Performs the above determination and change in consideration of the read character string stored in the read character string storage unit 11.

When a conversion command is given by the operator, the manual change unit 13 changes the character string written in the undetermined character string storage unit 12 based on the language rules stored in the language rule storage unit 17. For example, when an instruction to convert an undetermined character string to another candidate is given as a conversion instruction, the undetermined character string is converted to another candidate. Also, the manual changing means 13
When a command is given by the operator, the notation character string in the undetermined character string storage means 12 is changed with reference to the reading character string in the reading character string storage means 11. For example, when a clause break position change command is given as an operation command, new clause break processing is performed.

The output means 14 is an undetermined character string storage means 1
2 or the character string stored in the read character string storage means 11.

2. 2. Hardware Configuration (2.1) Outline FIG. 2 shows a kana-kanji conversion device 40 incorporating a voice conversion device and a display character string conversion device. The kana-kanji conversion device 40 includes an input device 41, a control device 43, a display device 45, and a storage device 47. The input device 41 is capable of voice input or key input of a character string, a conversion command, a selection command, and the like. The storage device 47 stores a dictionary for voice conversion and a dictionary for kana-kanji conversion.

The operator inputs voice data from the input device 41. The control device 43 converts the input speech data into a notation character string using the speech conversion dictionary, and converts the converted notation character string using the kana-kanji conversion dictionary to match the kana-kanji conversion rule. It is checked whether the conversion has been performed, and if the conversion that conforms to the Kana-Kanji conversion rule has not been performed, it is converted to a written character string that conforms to the Kana-Kanji conversion rule. The display device 45 displays the converted notation character string.

(2.2) Details FIG. 3 shows an example of a hardware configuration in which the kana-kanji conversion device 40 shown in FIG. 2 is realized using a CPU.

The kana-kanji conversion device 40 includes a CPU 23, a memory 27, a hard disk 26, a CRT 30, a FDD2
5, keyboard 28, mouse 31, interface 3
3, a microphone 34 and a bus line 29.

Audio data is input to the microphone 34.
The input audio data is digitally converted by the interface 33 and temporarily stored in a work buffer (not shown) of the memory 27.

The hard disk 26 stores a voice conversion program 26a, a voice conversion dictionary 26b, a kana-kanji conversion program 26c, a kana-kanji conversion dictionary 26d, and an application program 26p. FIG. 4 shows the data structure of the voice conversion dictionary 26b. As described above, the written character string and its read character string are stored for each voice data. The speech conversion dictionary 26b also stores inter-word connection data indicating how words are connected (not shown). FIG. 5 shows an example of the co-occurrence dictionary stored in the kana-kanji conversion dictionary 26d. Notation character strings having a co-occurrence relationship are stored in association with each other. The voice conversion program 26a and the kana-kanji conversion program 26c will be described later. The application program 26p
The notation character string is received from the kana-kanji conversion program 26c and output to the CRT 30.

The CPU 23 shown in FIG. 3 controls each unit via the bus line 29 according to the two programs stored in the hard disk 26. These programs are read from the flexible disk 25a in which the programs are stored via the FDD 25 and installed on the hard disk 26. In addition to the flexible disk, a hard disk may be installed from a computer-readable storage medium in which a program such as a CD-ROM or an IC card is substantially integrated. Furthermore, you may make it download using a communication line.

In the present embodiment, the program is installed on the hard disk 26 from the flexible disk, so that the computer indirectly executes the program stored on the flexible disk. However, without being limited to this, the program stored in the flexible disk is stored in the FDD25.
Alternatively, it may be executed directly from. Note that, as a program executable by a computer, not only a program that can be directly executed by simply installing it as it is, but also a program that needs to be temporarily converted into another form (for example,
Decompression of data that has been compressed, etc.), and also includes those that can be executed in combination with other module parts.

The memory 27 has an audio data buffer 27
a, a feature buffer 27f, a post-conversion buffer 27e, a reading buffer 27b, a semi-determined buffer 27c, and a defined buffer 27d.

The audio data buffer 27a, the feature buffer 27f and the post-conversion buffer 27e are buffers for an audio conversion program. Audio data buffer 27a
Stores digital audio data obtained by digitally converting input audio data. The voice feature extracted by the voice conversion program 26a is stored in the feature buffer 27f. The post-conversion buffer 27e stores the written character string specified by the voice conversion program 26a and its read character string.

Read buffer 27b, semi-determined buffer 27
c, the determination buffer 27d is a buffer for a kana-kanji conversion program. The reading character strings stored in the converted buffer 27e are sequentially stored in the reading buffer 27b. The semi-determined buffer 27c includes a post-conversion buffer 27e.
Are sequentially stored. The notation character string determined by the determination command given from the keyboard 28 is stored in the determination buffer 27d. The memory 27 also stores various calculation results and the like.

On the CRT 30, a notation character string in an undetermined state or a notation character string in a confirmed state is displayed.

3. Next, a program stored in the hard disk 26 will be described with reference to FIGS. In this embodiment, the application program 26p is:
The voice conversion program 26a and the kana-kanji conversion program 26b are called in a state where a character string can be input, and a kana-kanji character string is input. The voice conversion program 26a specifies the notation character string (FIG. 6, step ST0 to step ST0).
3) The kana-kanji conversion program 26b
Is passed to the application program 26p in an undetermined state. The application program 26p displays the notation character string in an undetermined state.

The kana-kanji conversion program 26b examines the notation character string, and if it does not match the predetermined kana-kanji conversion rule, converts the notation character string based on the kana-kanji conversion rule (steps ST4 to ST33 in FIG. 7). ).
Notation character strings converted based on the Kana-Kanji conversion rules are:
It is passed to the application program 26p. In the present embodiment, “ATOK11” (trademark) manufactured by Just System Co., Ltd., which is one of IME (input method editor), is used as the kana-kanji conversion program 26b, and “Ichitaro 8” (made by the company) is used as the application program 26p. Trademark).

First, the application program 26p
Will be described. Application program 26p
Is a Japanese document creation program, and when activated, a character string input enabled screen is displayed on the CRT 30 (not shown).

In this state, the voice conversion program 26a is called. The following processing is executed by the voice conversion program 26a. The operator operates the microphone 34 shown in FIG.
Input audio data from. The CPU 23 includes a microphone 34
It is determined whether or not there is a voice input (step ST0 in FIG. 6).
The digital audio data digitally converted by the interface 33 is stored in a (step ST1). For example, when "Sansei-Arikari-no-Sui-Eki" is uttered, the digitally converted data is stored in the audio data buffer 27a.

The CPU 23 has an audio data buffer 27a
A feature amount is extracted from the digital audio data stored in (step ST2). Next, the voice data and the inter-word concatenation data registered in the voice conversion dictionary 26b shown in FIG. 3 are collated to obtain a character string having the highest matching degree, and stored in the converted buffer 27e (step ST3). For example, in this case, a conversion candidate is specified for each word, such as "sansei / to / arikasei / no / suoiyoueki /", and a conversion candidate is specified for each word. Agree/
And / alkaline // aqueous solution ”and the reading character string“ sansei / to / akarisei / no / suoiyoueki ”are stored.

In step ST0 in FIG. 6, if there is no voice input, the voice conversion program ends.

Next, the kana-kanji conversion program 26b is called. The following processing is executed by the kana-kanji conversion program 26b. The CPU 23 determines whether or not a new character string is stored in the post-conversion buffer 27e (step ST4 in FIG. 7). If a new character string is stored, the conversion unit (phrase) of the kana-kanji conversion program is used. ) Is corrected, and the written character string is stored in the undetermined buffer 27c and the read character string is stored in the read buffer 27b (step ST5 in FIG. 7). In the above case,
For example, the notation character string “agreement / alkaline / aqueous solution” is stored in the semi-fixed buffer 27c, and the reading “sansanto / arukasei / suoiyoeki” is stored in the reading buffer 27b.

Next, the CPU 23 executes the semi-determined buffer 2
The notation character string "agreement and alkaline aqueous solution" stored in 7c is passed to the application program 26p with the condition that it is in a semi-determined state (step ST in FIG. 7).
7). When the application program 26p is executed, it is displayed on the CRT 30, for example, in a semi-determined state.

Next, the CPU 23 executes the semi-determined buffer 2
It is determined whether the character string stored in 7c matches the Kana-Kanji conversion rule (step ST9 in FIG. 7). In this example, the determination is made based on the co-occurrence dictionary shown in FIG. In this case, it is found that the co-occurrence relationship with the notation character string “alkaline” is not the notation character string “agree” but the notation character string “acidity”.

Therefore, the CPU 23 determines that the character string stored in the semi-fixed buffer 27c does not conform to the Kana-Kanji conversion rule, and converts the written character string stored in the semi-fixed buffer 27c based on the Kana-Kanji conversion rule. (Step ST11 in FIG. 7). Specifically, “agree” is replaced by “acid”, and “agree and alkaline aqueous solution” is stored in the semi-fixed buffer 27c.

Next, the CPU 23 executes the semi-determined buffer 2
The notation character string “acidic and alkaline aqueous solution” stored in 7c is transferred to the application program 26p in a semi-determined state (step ST13 in FIG. 7). Thereby, CR
It is displayed in a semi-determined state at T30.

If it is determined in step ST9 that the values match, the processes in steps ST11 and ST13 are not performed.

Next, the CPU 23 determines whether or not there is a key operation (step ST14). If there is no key operation, the kana-kanji conversion program ends, and the voice conversion program shown in FIG. 6 is executed.

On the other hand, if there is a key operation in step ST14, the flow advances to step ST15 to determine the type of key.

If the type of the key is the decision key, the process proceeds to step ST17 in FIG. 8, and the data of the semi-decision buffer 27c is stored in the decision buffer 27d. Then, it is passed to the application program 26p with the condition of the fixed state. The application program 26p causes the CRT 30 to display, for example, the notation character string “acidic and alkaline aqueous solution” stored in the determination buffer 27d.

If the key type is a break position change key, the break position of the reading buffer 27b is changed (FIG. 8).
In step ST21, the read character string whose delimiter position has been changed is passed to the application program 26p (step ST23). Then, returning to step ST14, it is determined whether or not a key operation has been performed.

If the key type is the cancel key, the data in the semi-determined buffer 27c and the read buffer 27b are cleared (step ST25 in FIG. 8). Then, the process ends.

If the type of the key is a katakana key, the notable phrase in the semi-fixed buffer 27c is subjected to katakana conversion (FIG. 8).
Step ST27). Then, returning to step ST14, it is determined whether a key operation has been performed.

If the key type is a hiragana key, the target phrase in the semi-fixed buffer 27c is converted into hiragana (FIG. 8).
Step ST29). Then, returning to step ST14, it is determined whether a key operation has been performed.

If the key type is the noticeable phrase change key,
The target phrase in the semi-determined buffer 27c is changed and converted (FIG. 8).
Step ST31). Then, returning to step ST14, it is determined whether a key operation has been performed.

If the type of the key is a conversion key, the notation of the target phrase in the semi-fixed buffer 27c is converted to the next candidate (step ST33 in FIG. 8). Then, returning to step ST14, it is determined whether a key operation has been performed.

If the type of key is a half-width conversion key, the target phrase in the half-fixed buffer 27c is half-width converted (step ST35 in FIG. 8). Then, returning to step ST14, it is determined whether a key operation has been performed.

If no character string is additionally stored in the post-conversion buffer 27e in step ST4 in FIG. 7, the process proceeds to step ST14 to determine whether a key operation has been performed.

The processing by the kana-kanji conversion program and the processing by the voice conversion program are sequentially repeated.

As described above, in the present embodiment, the notation character string converted by the voice conversion program is passed in an undetermined state to the Kana-Kanji conversion program, which is an IME, and inspected. The conversion is performed again based on the conversion rules of the conversion program. As a result, even if the speech conversion program has not been trained, appropriate Kana-Kanji conversion can be performed by utilizing the Kana-Kanji conversion rules already accumulated by the operator.

4. Other Embodiments In the present embodiment, the case where co-occurrence information is used as the kana-kanji conversion rule has been described, but other rules, such as learning information, may be used. Also,
Not only kana-kanji conversion to Japanese but any language rules may be used. For example, the same can be applied to voice input in other languages such as English and Chinese.

When there is a homonymous object word, by using the co-occurrence information as in the present embodiment, it is possible to obtain an appropriate written character string from data input by voice without an operation by an operator. On the other hand, reading voice data (technical terms and the like) not registered in the voice conversion dictionary of the voice conversion program cannot be correctly converted.
This is because the voice conversion program converts the character string into a written character string having a reading very similar to the reading. For example, if there is no term for a patent, it is converted to a similar Tokyo, etc., and its reading is also converted to “Tokyo”. In general, a kana-kanji conversion program searches for a conversion candidate by referring to a written character string or a reading given from a voice conversion program. Therefore,
Even in such a case, a written character string having a similar pronunciation may be stored in the kana-kanji conversion program as a kana-kanji conversion rule so that conversion can be performed accurately. In this way, by storing the similar word reading data and predicting the correct reading, an appropriate written character string can be obtained even if a slightly shifted reading is given from the speech conversion program.

Also, words that are in Japanese English,
For example, as for “system” and the like, the spelling “system” may be stored as a notation character string and converted to this.

In the present embodiment, as the voice conversion program, a program manufactured by IBM Japan, Ltd.
Although iaVoice (trademark) was adopted, the present invention is not limited to this, and other voice recognition programs can be similarly applied. In this way, when a commercially available voice conversion program is used as a voice input engine, even if the conversion rate is not so good due to the conversion target field, the conversion rules of the kana-kanji conversion stored by the operator are used. Thus, a desired written character string can be obtained.

Further, in the present embodiment, a case has been described in which the conversion candidates separated for each word are specified from the input voice data. However, the present invention is not limited to this, and the candidates may be specified for each phrase.

Further, the case where the voice conversion program is realized by software has been described, but it may be realized by hardware. That is, the audio data conversion means is a concept including the audio data conversion device.

In this embodiment, the application program 26p is displayed on the CRT 30 in a semi-determined state. That is, the kana-kanji conversion program outputs the display control command so that the notation character string stored in the undetermined buffer is displayed in an undetermined state. However, as for the display in the undetermined state, not only is the application program 26p executed in this way, but the Kana-Kanji conversion program receives a reply command indicating that no display was made from the application program 26p, and displays it. You may. In addition, such display may be realized by sharing with the operating system (OS).

In the present embodiment, the functions shown in FIG. 1 are realized by using the CPU 23 and software. However, some or all of them may be realized by hardware such as a logic circuit.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a notation character string conversion device 10 according to the present invention.

FIG. 2 is a diagram illustrating an example of a hardware configuration of a notation character string conversion device 10 illustrated in FIG.

FIG. 3 is a block diagram of a notation character string conversion device 10 shown in FIG.
FIG. 3 is a diagram illustrating an example of a hardware configuration realized using the H.23.

FIG. 4 shows a data structure of a speech conversion dictionary.

FIG. 5 shows a data structure of a kana-kanji conversion dictionary.

FIG. 6 is a flowchart of a voice conversion process.

FIG. 7 is a flowchart of a kana-kanji conversion process.

FIG. 8 is a flowchart of a kana-kanji conversion process.

FIG. 9 is a functional block diagram of a conventional voice conversion system.

[Explanation of symbols]

 3 ... Voice input means 5 ... Voice data conversion means 6 ... Voice conversion means 7 ... Conversion conversion dictionary 10 ... Notation character string conversion device 11 ... ..Reading character string storage means 12... Undetermined character string storage means 13... Manual conversion means 14... Output means 15... Automatic conversion means 17. 19 ... display means 23 ... CPU 27 ... memory

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-9-160750 (JP, A) JP-A-64-70865 (JP, A) JP-A-5-79657 (JP, U) (58) Field (Int.Cl. ⁷ , DB name) G06F 17/22

Claims (6)

(57) [Claims] 1. A notation character string corresponding to a feature amount of audio data.
And a reading character string corresponding to the reading character string corresponding to the notation character string.
By using the voice conversion dictionary, the input voice data
Best matches the feature value recorded in the voice conversion dictionary
A notation character string conversion device for converting a notation character string given from a voice data conversion means for converting into a notation character string and outputting it together with a corresponding reading character string so as to conform to a language rule, comprising a specific notation character string. The correspondence between the two is described based on language rules.
Corresponding reading characters corresponding to the notation character string
A language rule storage means for storing the strings together; an unconfirmed character string storage means for storing the notation character string provided from the audio data conversion means in an unconfirmed state; and a read character string provided from the audio data conversion means. A reading character string storing means for storing, a relationship between specific notation character strings among a plurality of notation character strings stored in the unconfirmed character string storage means matches the correspondence relationship.
Judge whether or not they match, and do not match
In the case, the reading character stored in the reading character string storage means
A reading character string corresponding to the specific notation character string in the column;
The specific notation character string based on the corresponding reading character string
Automatic conversion means for converting the notation character string into a notation character string that matches the correspondence relationship . With 2. A computer, a writing character string conversion method for converting to match the writing character string supplied from the sound data conversion unit into language rules, the audio data converting means, the audio data Notation character string corresponding to the feature amount and the notation sentence
A speech conversion dictionary that stores the reading character string corresponding to the character string
By using the input voice data,
Character string that most closely matches the feature value recorded in the translation dictionary
It is a means to convert and output with the corresponding reading character string.
The computer executes the following processing, and records the correspondence between specific notation character strings based on language rules.
Corresponding reading characters corresponding to the notation character string
A string is also stored, and the notation character string given from the voice data conversion means is stored in an undetermined state, and the voice data conversion means
Storing the read character string given al, or relationship of a particular writing character string among the plurality of writing character string stored in the undetermined state meets the relationship
Judge whether it is not the same as above.
The character string read from the voice data conversion means and the pair
The specific notation character string based on the unread character string
A notation character string conversion method using a computer , which performs conversion to a notation character string that matches the correspondence . 3. A storage medium storing a program for causing a computer provided with an input device, a control device, an output device, and a storage device to function as a notation character string conversion device, wherein the program executes the following means: Function based on language rules.
Corresponding reading characters corresponding to the notation character string
Language rule storage means for storing strings together, a notation character string corresponding to a feature amount of voice data, and the notation sentence
A speech conversion dictionary that stores the reading character string corresponding to the character string
By using the input voice data,
Character string that most closely matches the feature value recorded in the translation dictionary
An undetermined character string storage unit that stores a notation character string provided from the audio data conversion unit that converts and outputs the character string together with the corresponding reading character string in an undetermined state; and stores a read character string that is supplied from the audio data conversion unit. read character string storage means, if the correspondence relationship between certain writing character string among the plurality of writing character string stored in said unfixed character string storage means
Judge whether or not they match, and do not match
In the case, the reading character stored in the reading character string storage means
A reading character string corresponding to the specific notation character string in the column;
Writing character string of the specific based on said corresponding read character string
Automatic conversion means for converting into a notation character string that matches the correspondence relationship, a storage medium storing a program. 4. The notation character string conversion device according to claim 1, wherein the language rule storage means includes a notation character string having a similar reading to a certain reading character string.
The automatic conversion means stores a plurality of written characters stored in the unconfirmed character string storage means.
If the relationship between certain writing character string of the string is determined not to conform to the correspondence, to convert the representation string automatically converted into writing character string having a reading of the similarity, the Character string conversion device. 5. The notation character string conversion method according to claim 2, wherein the computer further includes a notation character string having a similar reading to a certain reading character string.
Stores the, especially among the plurality of writing character string that the stored in the undetermined state
If the relationship between the constant notation string is determined not to conform to the correspondence, to convert the representation string automatically converted into writing character string having a reading of the similarity, and wherein Notation string conversion method. 6. A storage medium storing a program according to claim 3, wherein said language rule storage means stores a notation character string having a similar reading to a certain reading character string, and said automatic conversion means , a plurality of writing character stored in said unfixed character string storage means
If it is determined that the relationship between specific notation character strings among the columns does not match the correspondence , the undetermined character string
Converting a plurality of notation character strings stored in the storage means into notation character strings having similar readings.