JPH0836395A - Generating method for voice data and document reading device - Google Patents

Abstract

PURPOSE:To read a document in a natural manner which always fits to the context and the scene of the document. CONSTITUTION:An analysis result buffer 7 stores the Japanese analysis results of a specified document among a document data file 1 done by a Japanese analysis section 4. A voice data generating section 8 generates voice data from the Japanses analysis results. During the data generation, a special reading is assigned to the alphanumeric column located before and after the special character detected by a special character detection section 17 in accordance with the rule corresponding to a special character processing rule table 21. Moreover, when the alphanumeric column of a special pattern is detected by a special pattern detecting section 19, a special reading is assigned to the detected column in accordance with the corresponding rule in a special pattern processing rule table 22, voice data are generated and stored in a voice data file 10. A voice synthesizer 11 converts the voice data into voice signals which are outputted by a voice output section 13. Thus, alphanumeric columns such as telephone numbers and postal code numbers in the document being read are adequately read because the synthesizer 11 converts the voice data to voice signals which are outputted by the section 13.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention reads a document by analyzing the reading target document in Japanese, generating voice data, and outputting the voice obtained by synthesizing the voice data to the outside. The present invention relates to a device, and more particularly to a method of generating audio data.

[0002]

2. Description of the Related Art In a conventional speech reading apparatus, in an analysis result obtained by analyzing the reading target document data in Japanese, for example, for a character string of numbers, a bar reading is performed in a preset mode, The voice data is generated by reading either the digit or the digit. As a concrete example, as the analysis result, "03 (123) 456" is obtained.
If “7” is obtained, the voice data generation unit normally responds to “Zerosa ^ // ^^ / hyakuni ^ 30 // ^^ / onse for this numeric character string. It generates a reading "^ / Gohyakukurojana ^ na", which is the customary way of reading when it is a telephone number, "Zerosa ^ no / Ichinii ^ san / Yongo-Rokuna". I didn't generate a reading saying "^".

Also, even if a character string of a number including a symbol for special reading such as "10:13", which has some variations in reading, is obtained as the analysis result, the voice data generating section Depending on the words that appear before and after, it is necessary to distinguish whether "juta ^ i / ju ^ u-san" is "ju ^ uji / ^ jusa ^ mpun", which is either one of the preset settings. I could only read it. Therefore, when reading the document data with the conventional document reading device, if the above example appears, it does not necessarily make a natural reading suitable for the context or scene, which gives the listener a feeling of discomfort. At the same time, there was a problem that listeners sometimes got the meaning wrong. Incidentally, in this example, those including special symbols such as parentheses and ":" in the numeric character string are collectively read as the numeric character string.

[0004]

In the conventional voice reading device as described above, when the Japanese analysis result of the read target document data is a numeric character string indicating a telephone number or a postal code, the time and ratio are set. If it is a numeric character string containing a special symbol indicating, only the preset reading is performed, so there are cases where natural reading appropriate for the context or scene is not done. There is a problem that the listener feels uncomfortable and that the listener sometimes misunderstands the meaning.

Therefore, the present invention eliminates the above-mentioned drawbacks, and detects a numerical character string which is conventionally read in a special way by a special word appearing in the Japanese analysis result of the document data to be read out. In addition to being able to generate the voice data of the special reading, if a numeric character string of a special pattern that conventionally has a special reading appears in the Japanese analysis result of the document data to be read, An object of the present invention is to provide a voice data generation method capable of detecting and generating voice data of the special reading, and further, voice-synthesizing voice from voice data generated by the voice data generation method. By creating a signal and outputting it to the outside, it is possible to provide a document reading device that can always perform natural reading suitable for the context or scene. It is the target.

[0006]

According to a first aspect of the present invention, voice data is generated as voice data according to a voice data generation rule from an analysis result obtained by performing Japanese analysis on document data in a form that can be handled by a computer. In the generating method, when a predetermined special word is detected from the Japanese analysis result, the numeric character string existing before and after the special word is followed by the rule corresponding to the detected special word. It has a method of generating voice data from the numerical character string by applying a reading.

A second aspect of the present invention is a voice data generating method for generating voice data according to a voice data generating rule from an analysis result obtained by performing Japanese analysis on document data in a form that can be handled on a computer. A method of generating voice data from a numeric character string by detecting a numeric character string of a predetermined special pattern from a Japanese analysis result and applying a reading according to a rule corresponding to the pattern to the numeric character string. Have.

A third aspect of the present invention is a voice data generating method for generating voice data according to a voice data generating rule from an analysis result obtained by performing Japanese analysis on document data in a form that can be handled on a computer. When a predetermined special word is detected from the Japanese analysis result, by applying a reading according to the rule corresponding to the detected special word to the numeric character strings existing before and after the special word, Generate voice data from the numeric character string, or
When a numeric character string of a predetermined special pattern is detected from the Japanese analysis result, by applying a reading according to the rule corresponding to the pattern to the numeric character string,
There is a method of generating voice data from the numeric character string.

According to a fourth aspect of the present invention, when the detection of the special word and the detection of the numeric character string of the special pattern occur at the same time for the same numeric character string, the one having a higher priority set in advance is detected. The method has a method of executing voice data generation processing to be activated and generating voice data from a corresponding numeric character string.

A fifth aspect of the present invention has a method of prioritizing voice data processing activated by detection of the special word.

According to a sixth aspect of the present invention, voice data is generated according to a voice data generation rule from an analysis result obtained by analyzing the document data to be read out in Japanese, and the voice data is converted into an electric voice signal by a voice synthesizer. In a document reading device which converts the obtained voice signal into a voice by a voice output device and outputs the voice to the outside, a special word detecting means for detecting a predetermined special word from the Japanese analysis result, and a special word detecting device. Extraction means for extracting numeric character strings existing before and after the special word detected by the word detection means, and the numeric character string extracted by this extraction means corresponding to the special word detected by the detection means And a voice data generating means for generating voice data from the numerical character string by applying reading according to a predetermined rule.

According to a seventh aspect of the present invention, the special word detecting means has table data listing a plurality of special words,
The special word is searched by matching the word appearing in the Japanese analysis result with the table data.

According to an eighth aspect of the present invention, the voice data generating means has table data in which a list of rules determined corresponding to each of the plurality of special words is listed, and the numeric character string extracted by the extracting means. In addition, by applying reading according to the rule in the table data corresponding to the special word detected by the detecting means, the numerical character string is generated into voice data.

According to a ninth aspect of the present invention, voice data is generated according to a voice data generation rule from an analysis result obtained by analyzing the document data to be read aloud in Japanese, and the voice data is converted into an electric voice signal by a voice synthesizer. In the document reading device which converts the obtained voice signal into a voice by the voice output device and outputs the voice to the outside, a special pattern detecting means for detecting a numeric character string of a predetermined special pattern from the Japanese analysis result. And a voice that generates voice data from the numeric character string by applying reading to the numeric character string of the special pattern detected by the special pattern detection means in accordance with a predetermined rule corresponding to the special pattern. And a data generating means.

According to a tenth aspect of the present invention, the special pattern detection means has table data in which a plurality of special pattern numeric character strings are listed, and the numeric character strings appearing in the Japanese analysis result are collated with the table data. By doing so, the special word is searched for.

According to an eleventh aspect of the present invention, the voice data generating means has table data in which a list of rules determined corresponding to each of the plurality of special pattern numeric character strings is provided.
The numeric character string of the special pattern detected by the detection means is read according to the rule in the table data corresponding to the special pattern, and the voice data is generated from the numeric character string.

According to a twelfth aspect of the present invention, voice data is generated according to a voice data generation rule from the analysis result obtained by analyzing the document data to be read out in Japanese, and this voice data is converted into an electric voice signal by a voice synthesizer. In the document reading device that converts the obtained voice signal into a voice by a voice output device and outputs the voice to the outside, a special word detection unit that detects a predetermined special word from the Japanese analysis result,
Extraction means for extracting numeric character strings existing before and after the special word detected by the special word detection means,
Voice data is generated from the numeric character string by applying reading to the numeric character string extracted by the extracting means according to a predetermined rule corresponding to the special word detected by the special word detecting means. The first voice data generating means, the special pattern detecting means for detecting a numeric character string of a predetermined special pattern from the Japanese analysis result, and the numeric character string detected by the special pattern detecting means. A second voice data generating means for generating voice data from the numeric character string by applying reading according to a predetermined rule corresponding to the special pattern of the character string.

According to a thirteenth aspect of the present invention, a voice data generation process activated by detection of a special word by the special word detection means and a voice data generation activated by detection of a numeric character string of a special pattern by the special pattern detection means. Setting means for setting which of the processing is prioritized, determination means for determining whether or not the numeric character string extracted by the extraction means and the numeric character string detected by the special pattern detection means are matched, When the determination means determines that the two numeric character strings match, the control means executes a voice data generation process according to the priority information set in the setting means.

[0019]

In the method of generating voice data according to the first aspect of the present invention, when the voice data is being generated from the Japanese analysis result, if there is a predetermined special word in the Japanese analysis result, it is detected and The numerical character strings existing before and after the detected special word are read according to the rule corresponding to the detected special word.

In the method of generating voice data according to the second aspect of the present invention, when the voice data is being generated from the Japanese analysis result, if a numeric character string of a predetermined special pattern is present in the Japanese analysis result, this is detected. Therefore, the reading according to the rule corresponding to the detected numeric character string of the special pattern is applied.

In the method for generating voice data according to the third aspect of the present invention, when the voice data is being generated from the Japanese analysis result, there is a predetermined special word in the Japanese analysis result,
When this is detected, the numeric character strings existing before and after the detected special word are read according to the rule corresponding to the detected special word. As a result, reading voice data that matches the scene or context limited by the special word is created from the numeric character string, or if there is a numeric character string of a special pattern that is predetermined in the Japanese analysis result. Since this is detected, the reading according to the rule corresponding to the detected numeric character string of the special pattern is applied.

In the method of generating voice data according to the present invention, when the detection of the special word and the detection of the numeric character string of the special pattern occur simultaneously for the same numeric character string, a preset priority order is set. The voice data generation process that is activated from the detection of the higher one is performed to generate voice data from the corresponding numeric character string.

In the method for generating voice data according to the fifth aspect of the present invention, the voice data processing activated by the detection of the special word is prioritized.

In the document reading apparatus of the invention of claim 6,
The special word detection means detects a predetermined special word from the Japanese analysis result. The extracting means extracts a numeric character string existing before and after the special word detected by the special word detecting means. The voice data generating means applies voice reading to the numeric character string extracted by the extracting means according to a predetermined rule corresponding to the special word detected by the detecting means, thereby converting the voice data from the numeric character string. To generate.

In the document reading apparatus of the invention of claim 7,
The special word detecting means has table data listing a plurality of special words, and searches for the special word by matching the words appearing in the Japanese analysis result with the table data.

In the document reading apparatus according to the invention of claim 8,
The voice data generating means has table data in which a list of rules determined corresponding to each of the plurality of special words is listed, and the special character detected by the detecting means is added to the numeric character string extracted by the extracting means. By applying the reading according to the rule in the table data corresponding to the word,
The voice data is generated from the numeric character string.

In the document reading apparatus of the invention of claim 9,
The special pattern detecting means detects a numeric character string of a predetermined special pattern from the Japanese analysis result. The voice data generation means applies voice data to the numeric character string of the special pattern detected by the special pattern detection means in accordance with a predetermined rule corresponding to the special pattern to obtain voice data from the numeric character string. To generate.

In the document reading apparatus of the tenth aspect of the invention, the special pattern detecting means has table data listing a plurality of special pattern numeric character strings, and the numeric character strings appearing in the Japanese analysis result and the numeric character string. The special pattern numeric character string is searched by collating with the table data.

In the document reading apparatus according to the invention of claim 11, the voice data generating means has table data in which a list of rules decided corresponding to each of the plurality of special pattern numeric character strings is included, The numeric character string of the special pattern detected by the detecting means is read according to the rule in the table data corresponding to the special pattern to generate voice data from the numeric character string.

In the document reading apparatus of the twelfth aspect of the invention, the special word detecting means detects a predetermined special word from the Japanese analysis result. The extracting means extracts a numeric character string existing before and after the special word detected by the special word detecting means. The first voice data generating means applies the reading according to a predetermined rule corresponding to the special word detected by the special word detecting means to the numeral character string extracted by the extracting means, thereby reading the numeral. Generate voice data from a character string. The special pattern detecting means detects a numeric character string of a predetermined special pattern from the Japanese analysis result. The second voice data generation means applies the reading according to a rule determined in advance corresponding to the special pattern of the numeric character string to the numeric character string detected by the special pattern detection means, thereby Generates voice data from a numeric character string.

In the document reading apparatus according to the thirteenth aspect of the present invention, the setting means is a voice data generating process activated by the detection of the special word by the special word detecting means, and a numeric character string of the special pattern by the special pattern detecting means. It is set which of the voice data generation processes activated by the detection of is prioritized. The determining means determines whether or not the numeric character string extracted by the extracting means matches the numeric character string detected by the special pattern detecting means. When the determination means determines that the two numerical character strings match each other, the control means executes a voice data generation process according to the priority information set in the setting means.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS 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 voice data generating method of the present invention and a document reading apparatus of the present invention using this method. Reference numeral 1 is a document data file storing document data in a form that can be handled on a computer, 2 is an input device for inputting various setting data at the time of reading, and 3 is a control unit for performing comprehensive control when reading the document data. Reference numeral 4 denotes a Japanese analysis unit that morphologically, syntactically and semantically analyzes the read document data with reference to the word dictionary 6, 5 is a setting buffer in which various setting data at the time of reading is stored, and 6 is document data. A word dictionary in which headings, parts of speech, readings, accents, meanings, and other information for analysis are collected as a list, 7 is an analysis result buffer for storing the analysis results by the Japanese analysis unit 4, and 8 is Japanese. A voice data generation unit that generates voice data corresponding to the analysis result of the analysis unit 4, 9
Is a voice data generation rule file storing a voice data generation rule referred to when the voice data generation unit 8 generates voice data, and 10 is a voice for storing the voice data generated by the voice data generation unit 8. data file,
Reference numeral 11 denotes a voice synthesizing device for regularly synthesizing voice signals based on the voice data read from the voice data file 10.
12 is a special character table that holds a list of special characters such as "zip code" and "ratio" that read the numerical character strings preceding and following these characters in a special reading manner, and 13 outputs a voice signal. A voice output unit such as a speaker, 14 is a display device such as a CRT or LCD that displays display data on the screen, 15 is a display data creation unit that creates display data to be displayed on the display device 14 from voice data, and 16 is display data Is a display data file for storing, 17 is a special character detection unit for detecting a special character held in the special character table 12 from the Japanese analysis result, and 18 is a special numeric character such as a numeric character string indicating a telephone number. A special pattern table that holds column patterns as a list, and 19 detects special patterns that are held in the special pattern table 18 from the Japanese analysis result. The special pattern detecting unit 20 reads the rule information in the special character processing rule 21 or the special pattern processing rule table 22 when processing the voice data for the special numeric character string and gives the special numeric character to the voice data generating unit 8. A processing unit, 21 is a special character processing rule table that holds a list of voice data generation rules for the special characters, and 22 is a special that holds a list of voice data generation rules for the special numeric character string patterns described above. It is a pattern processing rule table.

Next, the operation of this embodiment will be described with reference to the flow chart shown in FIG. The document data stored in the document data file 1 is a plurality of documents stored in a form that can be processed by a computer, and these document data are created by a document creation device such as a Japanese word processor or by a computer such as OCR. It has been read in.

When the control unit 3 receives an instruction from the input device 2 such as a keyboard, the document reading process is started. At this time, the control unit 3 is given an instruction from the input device 2 by an operator to specify a document to be read aloud and a condition for reading the document. The operator can set this work by the control unit 3 while following the guide displayed on the display device 14 such as a CRT or LCD.

In response to the operation of the operator as described above,
The control unit 3 reads the document specified by the input device 2 from the document data file 1 in step 201 of FIG.
A process of selecting a document to be read is performed, and the selected document is passed to the Japanese analysis unit 4. Next, the control unit 3 executes step 202
The reading condition input from the input device 2 is written in the setting buffer 5. Here, a part of the document data specified by the input from the operator's input device 2 is as shown in FIG. 3, for example. Further, the operator can set the reading condition from the input device 2 while the control unit 3 follows the guide displayed on the display device 14. The settings include basic reading speed, sound quality, Height, strength, set time for reading end, presence / absence of special reading of emphasized characters, changes during special reading of emphasized characters, presence / absence of reading, rest length,
Whether or not the reading of special characters has been changed, etc., the above condition data is stored in the setting buffer 5 as shown in FIG.

The Japanese analysis unit 4 which receives an instruction from the control unit 3.
Is a word dictionary 6 containing morphologically, syntactically and semantically analyzed the passed document in step 203, and morphological information, reading information, accent information, co-occurrence information between words and the like. The document is divided into words, divided into words, and the analyzed result is written in the analysis result buffer 7, and then the control unit 3 is informed that the document analysis is completed. Send a signal to indicate.

Here, the word dictionary 6 used by the Japanese analysis section 4 has a data structure example as shown in FIG.
The analysis result buffer 7 in which the analysis result by the Japanese analysis unit 4 is stored has a data structure example as shown in FIG. When the Japanese analysis unit 4 analyzes the document and guides the reading, and when there are a plurality of readings for one word, the co-occurrence information described in the word dictionary is used. The appropriate reading shall be determined based on.

Upon receiving the signal indicating that the analysis of the document is completed, the control unit 3 activates the voice data generation unit 8. The voice data generation unit 8 refers to the data in the voice data generation rule file 9 to generate voice data corresponding to the Japanese analysis result in the analysis result buffer 7, and stores this in the voice data file. At this time, the voice data generation unit 8 refers to the special character table 12 to find the special characters (words) registered in the special character table 12 shown in FIG. 10 in the Japanese analysis result in the analysis result buffer 7. ) Is present in step 204, and if it is present, the rule corresponding to the searched special character in the special character table 12 is read out, and then step 2
If not, the process proceeds to step 205.

When the voice data generation unit 8 proceeds to step 207, it extracts a numeric character string before and after the detected special character from the Japanese analysis result, and extracts the numeric character string and the special character table. The rule read from 12 is given to the special numeral character processing unit 15 to activate it. As a result, the special numeral character processing unit 15 reads the rule corresponding to the given rule from the data in the special character processing rule table 21 as shown in FIG. It is given to the generation unit 8. The voice data generation unit 8 generates voice data by adding readings and accents in accordance with the rules corresponding to the given number character string in step 209.

On the other hand, the voice data generation unit 8 performs step 20.
When the process proceeds to step 5, the special pattern table 18 as shown in FIG. 13 is referred to, and whether or not there is a special pattern registered in the special pattern table 18 in the Japanese analysis result in the analysis result buffer 7. Is determined in step 205, and if it exists, the special pattern table 1
After reading the rule corresponding to the searched special pattern in 8, the process proceeds to step 206, and when it does not exist, the process proceeds to step 209. When the voice data generation unit 8 proceeds to step 206, the special pattern table 18
The rule read out from is given to the special numeric character processing unit 20 to activate it. As a result, in step 208, the special numeric character processing unit 20 reads the rule corresponding to the given rule from the data in the special pattern processing rule table 22 having the structure shown in FIG. It is given to the voice data generation unit 8. The voice data generation unit 8 generates voice data by adding readings and accents in accordance with the rules corresponding to the detected special numeral character string in step 209.

Next, the specific processing contents of steps 204 to 209 will be described. For example, when "postal code" is not registered in the special character table 12 as a special character string for the sentence "His postal code is 123-45" as shown in the upper part of FIG. Part of "1
"23-45" is "hyaku ni ^ / san / mai eggasu /
Although the reading "Yojuju / go" is added, if the "zip code" is registered in the special character table 12 as shown in FIG. It is detected as a special character, and the special numeric character processing unit 20 is activated, and the special numeric character processing unit 20 refers to the special character processing rule table 21 to generate a reading. In this example, the special character processing rule table 21 has a rule that the adjacent numbers are read as stick readings and the reading of "no" is given to "-". According to this rule, 123- As shown in the lower part of FIG. 9, the reading of 45 is “Ichini ^ i / san's / yogo ^-”.

The character string "0" as shown in the upper part of FIG.
3 (123) 4567 ”in the document, the special pattern table 18 referred to by the special pattern detection unit 19
When the pattern that corresponds to is not registered, "Zerosa ^ / / ^ kko / Hyaku ni ^ san / ka ^ kko /
"Yonsei / Goyakurokujana" is added. However, if the special pattern table 18 having the structure as shown in FIG. 13 has a registration of “numeral (numeral) numeral”, the special pattern detection unit 19 collates the patterns, and the input character string becomes the special pattern table. If it is in the data in 18, the special numeric character processing unit 20 is activated, and the special numeric character processing unit 20 refers to the special character processing rule table 21 as shown in FIG. 14 to generate the reading. In this example, the special character processing rule table 21 describes a rule that the adjacent numbers are read as stick readings, and "(") is applied to "(" and ")". , The reading for the input string is "Zeros
Good-good! ” Therefore, the character string "03
The reading given to (123) 4567 "is as shown in the lower part of FIG.

Similarly, "10: 0" in the reading target document.
For the character string "0", the special pattern table 1
If the pattern "Numeric: Numerical" is registered in 8, the special pattern detector 19 detects the pattern "Numeral: Numerical" by comparing the character string with the pattern, and the special numeral character processor 20 To start up. As a result, the special numeral character processing unit 20 refers to the data in the special pattern processing rule table 22 to generate the reading. In this example, ":" stands for ":"
There is a rule that the word "Jun" and the word "Fun" are added after the number in the back. Incidentally, in this example, the reading "ju ^ uji / zero ^ un" or "ju ^ uji" can be generated according to the rules described in the special character processing rule table 21. Voice data creation unit 8
Will generate voice data in response to the above reading in step 209.

In the reading target document, for example, "12.
If there is a numerical character string including a decimal point such as “3456”, the special pattern detection unit 19 converts the numerical character string into the one shown in FIG.
The pattern "number.number" is detected by comparing it with the pattern registered in the special pattern table 18 as shown in FIG.
Activate for 0. As a result, the special numeric character processing unit 20 refers to the data in the special pattern processing rule table 22 as shown in FIG. 14 to generate the reading. In this example, "." Is read for ".", The digits above the decimal point are digit-read, the digits below the decimal point are stick-read, and if there are two or more digits below the decimal point, A rule is written that stick reading is performed so that a set of accent phrases is generated for every two characters. Therefore, after the decimal point in the above example,
It will be read as "Sanyo ^ / Goro ^ ku". After all,
In this example, the reading "Ju Uden / Sanyo ^ / Goro ^ ku" is generated according to the rule described in the special character processing rule table 21. The voice data creation unit 8 will create voice data for this reading in step 209.

However, in the numerical pattern as described above, since the numbers after the decimal point are not always a multiple of 2, there is a case where a remainder occurs when the numerical patterns are grouped in units of 2 as described above. For example, in the numerical character string "12.34567", if the decimal places are combined in units of two characters, 7 will be left over. In such a case, it is only necessary to generate voice data as an accent phrase in which the last three characters are grouped together, and to generate a reading of “sanyo ^ / guro ^ kunana ^ na”. Is written as a rule in the special pattern processing rule table 22 as shown in FIG. As a measure when the number after the decimal point does not become a multiple of 2, apart from the above, the last one character is used as an independent accent phrase to generate a reading, and "Sanyo ^ / Goro ^ ku / However, the former example is adopted in this example shown in FIG.

If there is a character string that matches both the special character detection unit 17 and the special pattern detection unit 19 as a numeric character string sent to the special numeric character processing unit 20, the rules of the special character detection unit 17 are set. Prioritize. For example, regarding the sentence “the ratio is 10:13”, if the numeric character string “10:13” is registered in the data in the special pattern table 18 as “number: number”, the special pattern detection unit 19 When the input character string matches the special pattern data, the special numeric character processing unit 20 is activated, and the special numeric character processing unit 20 refers to the data of the special pattern processing rule table 22. Then, the reading is generated. In this example, the rule is to put the reading "ji" for ":" and add the reading "fun" after the number after it, which is called "ju ^ uji / jusa ^ mpun". Be read. The voice data creation unit 8 will create voice data for the above reading in step 209.

However, the "ratio" is displayed in the special character table 12.
If is registered, the special character detection unit 17 detects the “ratio” as a special character from the special character table 12, and the special numeric character processing unit 20 is activated. The reading is generated with reference to the character processing rule table 21. In this example, the special character processing rule table 21 describes a rule that the adjacent numbers are read as stick readings and "tai" is read for ":".
The reading for "0:13" will be "Jutta ^ i / Ju ^ san", and the whole reading will be "the secret
You can get the reading "I / Ju U-san / Aru".
In such a case, as described above, the reading obtained by the detection by the special character detection unit 17 is prioritized, so that the voice data generation unit 8 is "the secret is / 13 ^ / / The readings "Ru" are adopted, and voice data for these readings is created in step 209.

When the special character detection and the special pattern detection described above occur simultaneously for the same numeric character string, the priority order can be set in advance from the input device 2 to the control unit 3. Therefore, the control unit 3 sets this priority in the voice data generation unit 8 so that the above processing is performed. Therefore, if the priorities are set in the control unit 3 in reverse, the special pattern detection unit 19
The voice data generation process by the detection of is prioritized and executed.

The voice data generator 8 has an analysis result buffer 7
For the portion of the Japanese analysis result data that is not a special character or a special pattern, the voice data is generated after referring to the voice data generation rule file 9 in step 208, and the special character or the special pattern is applied to these. Voice data is generated for the read and stored in the voice data buffer 10. When the creation of the audio data is completed, the audio data generation unit 8 sends a signal indicating that the creation of the audio data is completed to the control unit 3.

Here, a part of the voice data generation rule is shown in FIG.
The output voice data shown in FIG. FIG.
This is an example of a rule in which the accent form is O-type when the accent form is not O-type in the five-verb and the inflectional form is pre-formed. Further, FIG. 8 shows the format of the audio data file 10. However, in the reading character string data, katakana characters represent voice data, “^” represents the position of accent, and “.” Represents a break of the length set in the setting buffer 5.

Upon receiving the signal indicating that the voice data creation is completed, the control unit 3 transfers the voice data in the voice data file 10 to the voice synthesizer 11. The voice synthesizer 11 sets the speed set in the setting buffer 5 in step 210,
A voice is output from the voice output unit 13 such as a speaker by converting the voice data into an electric signal according to sound quality, height, and strength. Here, the speech synthesizer 11 includes a phoneme example,
It is a device for converting data, which is composed of a special control code, into an electric voice signal, and the data has the same format as the voice data in the voice data file 10. That is, when the voice synthesizer 11 receives a character string in the format of the voice data file 10, it can be said that the voice synthesizer 11 is capable of performing regular voice synthesis. Rule composition is performed on the incoming character string.

The document is read aloud by a series of steps 203 to 210 shown in FIG. Control unit 13
Determines in step 211 whether the reading target document has been read to the end, and if it has not been read,
After step 212, the process returns to step 203, and when the reading is finished, the process is finished.

According to the present embodiment, when the voice data is created from the Japanese analysis result, the numeric character string and the normal digit reading which are not digit-reading such as telephone numbers and postal codes are used. The numeric character string to be performed is identified, and for the numeric character string to be read aloud in a special way, the pattern of this numeric character string, or the telephone number or zip code before and after this numeric character string, etc. Since the special characters can give a reading suitable for the context or scene, it is possible to always read appropriately the numerical character string as described above. As a result, when special characters such as "(,)" and "-" are adjacent to numbers that meet the conditions,
"Zerosan's / Ichinii's / Japanese-Good
, Etc., it is possible to realize natural reading of documents suitable for reading phone numbers and postal codes, and to eliminate the possibility of giving a sense of discomfort to the listener when reading the document or making a mistake in the meaning of reading the document. The performance of the device can be improved.

[0054]

As described above, according to the invention of claim 1 or 6, it is possible to create reading voice data that matches the scene or context limited by the special word from the numeric character string. You can read a number string naturally.

According to the second or ninth aspect of the invention, it is possible to create the voice data of the reading that matches the scene or context limited by the numeric character string of the special pattern, and read the numeric character string naturally. It can be performed.

According to the third or twelfth aspect of the invention, reading voice data suitable for the scene or context limited by the special word is created from the numeric character string, or the numeric character string is used to read the numeric character. It is possible to create reading voice data that matches the scene or context limited by the pattern of the string, and perform natural reading of the numeric character string.

According to the fourth or thirteenth aspect of the present invention, the reading voice data suitable for the scene or the context, which is limited to either the special word or the pattern of the numeric character string, is created from the numeric character string. It is possible to read the numeric character string naturally.

According to the invention of claim 5, it is possible to create reading voice data that matches the scene or context limited by the special word from the numeric character string, and read the numeric character string naturally. It can be carried out.

According to the invention of claim 7, it is possible to surely search for the special word.

According to the invention of claim 8, the reading according to the rule corresponding to the special word can be surely obtained.

According to the tenth aspect of the invention, it is possible to surely search the pattern numeral character string.

According to the eleventh aspect of the present invention, it is possible to surely obtain the reading according to the rule corresponding to the numeric character string of the special pattern.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a document reading apparatus according to the present invention.

FIG. 2 is a flowchart showing a document reading process of the apparatus shown in FIG.

FIG. 3 is a diagram showing a part of the document data in the document data file shown in FIG.

FIG. 4 is a diagram showing an example of contents of a setting buffer shown in FIG.

5 is a diagram showing an example of the structure of the word dictionary shown in FIG.

6 is a diagram showing an example of contents of an analysis result buffer shown in FIG.

7 is a diagram showing an example of data in the audio data generation rule file shown in FIG.

FIG. 8 is a diagram showing an example of audio data in the audio data file shown in FIG.

FIG. 9 is a diagram showing an example of special character in a document and audio data generated by applying a special character processing rule.

10 is a diagram showing an example of special characters in the special character table shown in FIG.

11 is a diagram showing an example of special character processing rules in a special character processing rule buffer shown in FIG.

FIG. 12 is a view showing an example of a special pattern in a document and an example of audio data generated by applying a special pattern processing rule.

13 is a diagram showing an example of a special pattern in the special pattern table shown in FIG.

14 is a diagram showing an example of a special pattern processing rule in the special pattern processing rule buffer shown in FIG.

[Explanation of symbols]

1 ... Document data file 2 ... Input device 3 ... Control unit 4 ... Japanese analysis unit 5 ... Setting buffer 6 ... Word dictionary 7 ... Analysis result buffer 8 ... Voice data generation unit 9 ... Voice data generation rule file 10 ... Voice data file 11 ... Voice synthesizer 12 ... Special character table 13 ... Voice output unit 14 ... Display device 15 ... Display data generation unit 16 ... Display data file 17 ... Special character detection unit 18 ... Special pattern table 19 ... Special pattern detection unit 20 ... Special Numerical character processing unit 21 ... Special character processing rule table 22 ... Special pattern processing rule table

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI technical display location // G06F 17/28

Claims (13)

[Claims] 1. A voice data generation method for generating voice data according to a voice data generation rule from analysis results obtained by subjecting document data in a form that can be handled on a computer to Japanese analysis. When a predetermined special word is detected, by applying a reading according to the rule corresponding to the detected special word to the numeric character string existing before and after the special word, from the numeric character string. A voice data generation method characterized by generating voice data. 2. A voice data generation method for generating voice data according to a voice data generation rule from a analysis result obtained by subjecting document data in a form that can be handled by a computer to Japanese analysis. When a numeric character string of a predetermined special pattern is detected, the numeric character string is read according to a rule corresponding to the pattern to generate voice data from the numeric character string. Data generation method. 3. A voice data generation method for generating voice data according to a voice data generation rule from a analysis result obtained by subjecting document data in a form that can be handled by a computer to Japanese analysis. When a predetermined special word is detected, by applying a reading according to the rule corresponding to the detected special word to the numeric character string existing before and after the special word, from the numeric character string. When the voice data is generated or a numeric character string of a predetermined special pattern is detected from the Japanese analysis result, the numeric character string is read according to a rule corresponding to the pattern to obtain the numeric character. A voice data generation method characterized by generating voice data from a character string. 4. The voice data generating method according to claim 3, wherein the detection of the special word and the detection of the numeric character string of the special pattern occur simultaneously for the same numeric character string,
A voice data generation method characterized by performing voice data generation processing started from detection of a preset higher priority, and generating voice data from a corresponding numeric character string. 5. The voice data generation method according to claim 4, wherein the voice data processing activated by the detection of the special word is prioritized. 6. A voice data is generated according to a voice data generation rule from an analysis result obtained by performing Japanese analysis on document data to be read out, and this voice data is converted into an electric voice signal by a voice synthesizer to obtain the voice data. In a document reading device that outputs the obtained voice signal as a voice by a voice output device to the outside, a special word detecting unit that detects a predetermined special word from the Japanese analysis result and a special word detecting unit that detects the special word. Extraction means for extracting a numeric character string existing before and after the special word, and the numeric character string extracted by this extraction means is predetermined in correspondence with the special word detected by the detection means. A document reading apparatus, comprising: voice data generating means for generating voice data from the numeric character string by applying reading according to the rules. 7. The document reading apparatus according to claim 6,
A document characterized in that the special word detecting means has table data in which a plurality of special words are listed, and the special words are searched by matching words appearing in the Japanese analysis result with the table data. Reading device. 8. The document reading apparatus according to claim 6 or 7, wherein said voice data generation means has table data in which a list of rules determined corresponding to each of said plurality of special words is listed, and said extraction is performed. In the numerical character string extracted by means,
A document reading apparatus, which generates voice data from the numeric character string by applying a reading according to a rule in the table data corresponding to the special word detected by the detecting means. 9. A voice data is generated according to a voice data generation rule from an analysis result obtained by analyzing the document data to be read out in Japanese, and the voice data is converted into an electric voice signal by a voice synthesizer to obtain the voice data. In a document reading device that outputs the generated voice signal to the outside as a voice by a voice output device, special pattern detecting means for detecting a numeric character string of a predetermined special pattern from the Japanese analysis result, and the special pattern. A voice data generating means for generating voice data from the numeric character string by applying reading according to a predetermined rule corresponding to the special pattern to the numeric character string of the special pattern detected by the detecting means. A document reading device characterized by being provided. 10. The document reading apparatus according to claim 9, wherein the special pattern detection means has table data listing a plurality of special pattern numeric character strings, and a numeric character string appearing in the Japanese analysis result is stored. A document reading device characterized by searching the special word by collating with the table data. 11. The document reading apparatus according to claim 9 or 10, wherein said voice data generating means has table data in which a list of rules determined corresponding to each of the plurality of special pattern numeric character strings is provided. Then, by applying reading according to the rule in the table data corresponding to the special pattern to the numeric character string of the special pattern detected by the detecting means, voice data is generated from the numeric character string. Document reading device. 12. A voice data is generated according to a voice data generation rule from an analysis result obtained by analyzing Japanese of document data to be read aloud, and this voice data is converted into an electric voice signal by a voice synthesizer to obtain the voice data. In a document reading device that outputs the obtained voice signal as a voice by a voice output device to the outside, a special word detecting unit that detects a predetermined special word from the Japanese analysis result and a special word detecting unit that detects the special word. Extracting means for extracting a numeric character string existing before and after the special word, and the numeric character string extracted by the extracting means, corresponding to the special word detected by the special word detecting means, in advance. A first voice data generating means for generating voice data from the numerical character string by applying a reading according to a predetermined rule, and a Japanese language analysis result for predicting the voice data. A special pattern detecting means for detecting a numeric character string of a special pattern defined by the special pattern, and a predetermined rule corresponding to the special pattern of the numeric character string detected by the special pattern detecting means. And a second voice data generating means for generating voice data from the numerical character string by applying the reading according to the document reading device. 13. The document reading apparatus according to claim 12, wherein the voice data generating process activated by the detection of the special word by the special word detecting unit and the detection of the numeric character string of the special pattern by the special pattern detecting unit. By the setting means for setting which of the voice data generation processing started by the priority character string and the numeric character string extracted by the extracting means and the numeric character string detected by the special pattern detecting means are coincident with each other. And a control means for executing a voice data generation process according to the priority information set in the setting means when the judgment means judges that the two numeric character strings match. Document reading device characterized by.