JPH05314302A - Automatic reading-out reader - Google Patents

Abstract

PURPOSE:To read out KANJI (Chinese character), to which FURIGANA (Japanese phonetic syllabary written at the side of KANJI) is added because this KANJI has a difficult or special reading, in the way of reading according to FURIGANA characters without KANJI-KANA (Japanese syllabary) conversion. CONSTITUTION:This automatic reading-out reader optically reads a document sentence A and outputs it with a voice, and a layout processing part 3 of an optical character reader 1 is provided with a means 3a which recognizes FURIGANA characters, a means 3b which recognizes KANJI in the text corresponding to FURIGANA characters, and a means 3c which relates FURIGANA characters and KANJI corresponding to them to each other. A KANJI-KANA conversion part 7 of a language processing device 6 is provided with a means 7a which extracts KANJI with FURIGANA characters from a character code string and a means 7b, which handles the code of FURIGANA characters as a HIRAGANA (cursive form of Japanese syllabary) code as it is without KANA- KANJI conversion of KANJI.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reading device for automatically reading a book or printed manuscript.

[0002]

2. Description of the Related Art Conventionally, books such as OCR (optical character reader: optical character reader), language processing technology, voice synthesis technology, voice output means, etc. (Generally referred to as a sentence)
A reading device that reads aloud automatically has been proposed and is being prototyped.

To explain the outline, first, a document sentence is read by an image sensor in an optical character reading device, and the image data is collated with reference character data in an internal dictionary for character recognition. The character code determined by character recognition is converted into a hiragana code by kanji-kana conversion (conversion from kanji to kana) in a language processing device. Then, the voice synthesizer generates a synthetic voice from the open code and outputs the voice from the speaker.

[0004]

The problem here is that, in the conventional example, although the kana of the main text in the manuscript and the reading of the kanji are considered, the kanji with ruby It means that no one has considered the reading aloud.

For this reason, even if the Chinese characters that are difficult to read (for example, the Chinese characters of the second level or higher) are purposely rubbed, if the Chinese characters are not stored in the internal dictionary, they cannot be read aloud. was there.

[0006] In addition, when a kanji that is not particularly difficult to read but is specially read has ruby, for example, "poetry" is read as "song" or "woman" is read as "person." For this reason, even though [Uta] and [Human] are attached to the ruby characters, they cannot be read in a special way, and the contents of novels and poems, including their nuances, cannot be tasted faithfully according to the original text. There was an inconvenience.

The present invention was devised in view of the above circumstances, and includes kanji characters that are difficult to read and cannot be converted to kanji characters, or kanji characters that have a special reading and are attached with ruby characters. The purpose is to be able to read aloud as it is in the original text in the way the ruby characters are read.

[0008]

An automatic reading-type reading device according to the present invention is an optical character reading device for performing character recognition from image data obtained by optically reading a document text,
A language processing device for converting a character code determined by character recognition into a hiragana code, a voice synthesizing device for creating synthetic voice data from the hiragana code, and a voice output means for outputting a voice corresponding to the synthetic voice data. An automatic reading-type reading device provided, wherein the optical character reading device recognizes ruby characters, a means for recognizing the kanji in the text corresponding to the ruby characters, and the ruby characters and the corresponding ruby characters. And a means for associating kanji with each other, wherein the language processing device extracts means for extracting kanji with ruby characters from character codes to be processed in time series, and when extracting kanji with ruby characters. Instead of converting the kanji into kanji and kana, the corresponding ruby character code is directly processed as a hiragana code. The one in which the features.

[0009]

[Operation] When reading a manuscript sentence consisting of a sentence containing kanji with ruby characters, when the kanji with ruby characters are recognized, the ruby characters and the corresponding kanji in the text are recognized. , Both are associated. Then, when the character code is processed in time series and converted into a hiragana code, when the kanji with ruby characters is extracted, kanji-kana conversion is not performed for that kanji and the ruby character corresponding to that kanji is The code of is treated as it is as an open code. In the speech synthesis, synthesized voice data of the reading of the ruby characters, not the reading of the kanji characters, is generated at the location of the kanji with the ruby characters, and the reading of the ruby characters is voice output from the voice output means.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an automatic reading-type reading device according to the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing an electrical schematic configuration of an automatic reading-type reading device according to an embodiment.

In the figure, A is a manuscript sentence such as a paperback book, a book or a magazine, 1 is an optical character reading device (OCR),
2 is an image sensor, 3 is a layout processing unit, 4 is a character recognition unit, 5 is an internal dictionary, 6 is a language processing device, 7 is a Kanji / Kana conversion unit, 8 is a morphological analysis unit, 9 is a voice synthesis device, 10
Is a phonological / prosodic control unit, 11 is a voice synthesis processing unit, 12 is a voice output control unit, and 13 is a speaker as a voice output means.

The optical character reader 1 comprises an image sensor 2, a layout processing section 3, a character recognition section 4 and an internal dictionary 5. The language processing device 6 includes a kanji / kana conversion unit 7
And a morphological analysis unit 8. The voice synthesis device 9 includes a phoneme / prosody control unit 10, a voice synthesis processing unit 11, and a voice output control unit 12.

FIG. 2A shows an optical character reading device 1
3 is a block diagram showing a detailed configuration of the layout processing unit 3 in FIG. 3, 3a is a ruby character recognizing means, 3b is a ruby character corresponding kanji character recognizing means for recognizing a kanji character in the text corresponding to the ruby character, and 3c is a ruby character. It is a ruby-kanji associating means for associating with the corresponding kanji.

FIG. 2B is a block diagram showing a detailed configuration of the kanji-kana conversion unit 7 in the language processing device 6, where 7a is a kanji character extraction unit with ruby characters, and 7b is a kanji character with ruby characters extracted. It is a ruby character hiragana code handling means that sometimes handles the corresponding ruby character code as it is as a hiragana code instead of converting that kanji into kanji.

FIG. 3 is an example of a sentence including Kanji with ruby characters, which is extracted from the novel "I am a cat" by Soseki Natsume. In this manuscript sentence A, for example, ruby is attached to kanji such as "registration" (reading is "kento"), "disease"("doaku"), and medicine can ("kettle"). There is.

FIG. 4 shows a layout process of image data A _{0 obtained by} reading the original sentence A (for example, one page) shown in FIG. 3 by the image sensor 2 in the optical character reading apparatus 1 and the text layout of the image data A _0. The state of the vertical and horizontal projection data A _V , A _H when processed by the unit 3 is shown.

In FIG. 4, 21 is a page area, 22 is a text line area, 23 is a ruby area, and 22a is a ruby-corresponding word / phrase area.

Here, the general operation of the automatic reading-type reading device having the above-mentioned configuration will be described.

First, the image sensor 2 of the optical character reading device 1 optically reads one page of the original sentence A. Next, the layout processing unit 3 performs layout processing on the read image data A ₀ . As shown in FIG. 4, the layout process projects the image data A ₀ in the vertical direction and the horizontal direction, and based on the obtained vertical projection data A _V and horizontal projection data A _H , the text line area The positions and widths of the character 22 and the ruby area 23, the length of the text line area 22, the length of one character, and the like are detected.

Next, the character recognition unit 4 cuts out the image data of each character, and based on the pattern matching with the reference character data sequentially read from the internal dictionary 5, the cut-out image data of one character is extracted. Character recognition is performed, the character code is determined, and the character code is transferred to the language processing device 6. Such character recognition and transfer are sequentially executed in time series.

The kanji-kana conversion unit 7 of the language processing device 6 converts the character code transferred in time series into a hiragana code. Then, the morphological analysis unit 8 analyzes the sentence by dividing it into word units and phrase units. The unit of the hiragana code string thus processed is transferred to the speech synthesizer 9 in time series.

The phoneme / prosody control unit 10 of the speech synthesizer 9.
Generates a parameter for speech synthesis by performing phonological control or prosodic control that is accenting of words based on an accent dictionary or the like. Then, in the voice synthesis processing section 11, a DSP (digital signal processor) or the like converts the input hiragana code string into synthesized voice data based on the voice synthesis parameters. Next, the audio output control unit 12 performs output control such as D / A conversion and filtering. Such processing is sequentially performed in time series, and the synthesized voice is sequentially output from the speaker 13.

Next, the reading operation of the manuscript sentence A, which is a sentence including kanji with ruby characters, will be described.

FIG. 5 is a flow chart showing a routine of processing of kanji with ruby characters in the layout processing by the layout processing section 3.

In step S1, the text line area 22 and the ruby area 23 are detected. That is, in the vertical direction projection data A _V , the case where the line width is larger than a predetermined threshold value is the body line area 22, and the case where the line width is less than the threshold value is the ruby area 23.

In step S2, the ruby corresponding word / phrase area 22a in the body line area 22 corresponding to the ruby area 23 is determined. That is, the position of the ruby area 23 is determined based on the length from the beginning of the line and the number of character columns. Then, the position is applied to the text line area 22 to obtain the ruby-corresponding word area 22a.

In step S3, the ruby corresponding word / phrase area 22a
The identifier for the character code of the Kanji in
Byte data) is added to associate the ruby character with the corresponding kanji. An example thereof is shown in FIG.

FIG. 7A shows an identifier x shown by
Shows the body character code string 30 to which ₁ , x ₂ ...
FIG. 7B shows a ruby character code string 40 associated with each of the identifiers x ₁ , x _2, ... Identifiers x ₁ , x ₂ ...
Are numbered serially from the beginning of the page, which makes them distinguishable from each other.
The ruby characters and the corresponding kanji characters are associated with each other. That is, the identifier x for the "register" of the Chinese character
_The ruby character “kento” is associated with ₁ as a medium,
Ruby character "How evil" is associated with the identifier x ₂ with respect to "Doaku" of Chinese characters as an intermediary.

In step S4, it is determined whether or not one page is the end, and the above steps S1 to
S3 is repeatedly executed.

Next, the operation of performing language processing on the text character code string 30 with an identifier and the ruby character code string 40 shown in FIG. 7 will be described.

FIG. 6 is a flowchart showing a conversion processing routine by the Kanji / Kana conversion unit 7.

In step S11, the body character code string 30
The character code in is searched one by one. Step S
At 12, it is determined whether or not the character code is an identifier (this is 2 bytes). If it is not an identifier, the process proceeds to step S13 to determine whether the character code is a kanji code. If it is not a kanji code, the process skips to step S17. If it is a kanji code, the process proceeds to step S14 to perform kanji-kana conversion. That is, the kanji code is converted into a hiragana code.

When it is determined in step S12 that the character code is an identifier, that is, when the character code is found to be a kanji code with ruby characters, the process proceeds to step S15, and based on the identifier, as shown in FIG. The ruby character code string 40 shown in (b) is searched. Then, in step S16, the code of the ruby character following the corresponding identifier in the ruby character code string 40 is determined as it is as a hiragana code for reading.

In step S17, it is determined whether or not one page is the end, and the above step S1 is performed until the end.
1 to S16 are repeatedly executed, and when one page is finished, the process moves to the next page. This is because the body character code string 30 and the ruby character code string 40 are managed on a page-by-page basis.

As described above, at the location of the kanji with ruby characters, the synthesized speech data is generated not by reading the kanji but by reading the ruby characters in the speech synthesis, and the sound of reading the ruby characters is output from the speaker 13. Will be done. It goes without saying that the reading voice of the ruby characters is output at the timing when the kanji corresponding to the ruby characters should be read out.

Therefore, it is possible to read aloud even in difficult kanji such as "disease" without saying that it is impossible to read aloud, as in the ruby characters, and to enjoy the nuances of the content of the novel as in the original text. You can

Further, for example, even when a special ruby is put on so that "poem" is read as "song" or "woman" is read as "person", it is faithful based on the reading of the ruby character. Can be read aloud.

[0039]

As described above, according to the present invention, with regard to the kanji with ruby characters, the sound is output not by reading the kanji but by reading the ruby characters. When reading a Kanji with ruby text, you can read it exactly as it is in the original text by reading it exactly as it is, without making the reading impossible or making a false reading. , The effect of being able to taste the contents of novels and poems faithfully along with the original text including their nuances.

[Brief description of drawings]

FIG. 1 is a block diagram showing an electrical schematic configuration of an automatic reading-type reading device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a detailed configuration of a layout processing unit and a Kanji / Kana conversion unit in the embodiment.

FIG. 3 is an exemplary diagram of a sentence including kanji with ruby characters.

FIG. 4 is a diagram showing a sentence layout of image data obtained by optically reading a sentence including kanji with ruby characters and a state of vertical and horizontal projection data thereof.

FIG. 5 is a flowchart showing a processing routine for kanji with ruby characters in the layout processing in the embodiment.

FIG. 6 is a flowchart showing an operation of performing language processing on a text character code string and a ruby character code string in the embodiment.

FIG. 7 is a sentence example diagram showing an example of a body character code string and a ruby character code string in the embodiment.

[Explanation of symbols]

1 optical character reader (OCR) 2 image sensor 3 layout processing unit 3a ruby character recognition means 3b kanji character recognition means for ruby characters 3c ruby kanji character association means 4 character recognition part 5 internal dictionary 6 language processing device 7 kanji kana conversion part 7a Kanji character extraction means with ruby characters 7b Rubi character hiragana code handling means 8 Morphological analysis part 9 Speech synthesis device 10 Phonological / prosodic control part 11 Speech synthesis processing part 12 Speech output control part 13 Speaker (speech output means) 21 Page area 22 Body line area 22a Ruby corresponding phrase area 23 Ruby area 30 Body character code string 40 Ruby character code string A Original text A ₀ Image data A _V Vertical projection data A _H Horizontal projection data

Claims (1)

[Claims] 1. An optical character reading device for performing character recognition from image data obtained by optically reading an original text, a language processing device for converting a character code determined by character recognition into a hiragana code, and a hiragana code. An automatic reading-type reading device including a voice synthesizing device for creating synthetic voice data from a voice, and a voice output means for outputting a voice corresponding to the synthetic voice data, wherein the optical character reading device recognizes ruby characters. Means, a means for recognizing the kanji in the text corresponding to the ruby characters, and a means for associating the ruby characters with the corresponding kanji, wherein the language processing device is time-sequential. A method to extract kanji with ruby characters from the character codes to be processed in and instead of converting the kanji with kanji when extracting kanji with ruby characters. Automatic reading type reading apparatus characterized by being configured to that and means for handling as it opened up a code ruby character code to respond.