JPH0434189B2 - - Google Patents

Description

[Detailed Description of the Invention] [Summary] When the notation of a sentence input to a text-reading proofreading device includes misuse of kanji with the same reading, okurigana, etc., correct proofreading cannot be performed depending on the content read out loud. Therefore, we provided a means to check for typographical errors such as typos and misuse, making it possible to correct the notation.

[Industrial application field]

The present invention relates to a text reading proofreading device that converts a written text into audio for proofreading and automatically reads the text out loud.

[Conventional technology]

Text-to-speech proofreading devices that input written text into characters, automatically determine the pronunciation, output it aloud, and use it for proofreading have recently been attracting attention as a useful means of streamlining document proofreading. .

In Japanese sentences, a single kanji often has multiple readings, such as onkun, and the usage may differ depending on the meaning or the relationship between consecutive words. In order to determine the correct "reading" for a text, a multifaceted analysis of the text is required.

FIG. 5 shows the basic configuration of a conventional text reading proofreading device.

In the figure, 51 is a text input section, 52 is a text analysis section, 53 is a reading dictionary, 54 is a reading sequence storage section, 5
5 is an audio output section.

The text input unit 51 converts Japanese text input using a keyboard or the like into text data expressed in Kanji code.

The text analysis unit 52 uses the reading dictionary 53 to identify (recognize) each word in the input text data.
Then, the "yomi" is stored in the reading sequence storage section 54. Identification of words in a sentence is performed by extracting candidate words from the reading dictionary 53, performing DP matching with the sentence data, and selecting a combination of words with the minimum distance.

In the reading dictionary 53, the spelling of words, their pronunciations, usage, etc. are registered. Words include kanji words, katakana words, kanji kana-mixed words, etc.

The audio output unit 55 takes out the pronunciation of a sentence from the pronunciation storage unit 54, performs speech synthesis, and outputs the result as a voice.

In this way, if any written text is input to the text reading proofreading device, an appropriate reading sequence will be automatically created and the text will be read out.

[Problem that the invention seeks to solve]

With conventional text reading proofreading devices, errors such as `` kaishin '' (error of ``kaishin'') and `` kaibar '' (error of ``kaicho'') may occur in the text due to errors during manuscript creation or input. If there are misspellings or incorrectly used words such as “analogue” (an error in “analogue”), if they have the same pronunciation (homophone) as the correct word, the proofreader should correct the manuscript by himself/herself. These errors had to be detected by comparing them with the display screen.

[Means for solving problems]

In order to solve the above-mentioned problems, the present invention includes an error checking dictionary in which incorrect spellings and correct spellings of words that are easily made are registered in association with each other, and an error spelling correcting unit in a text reading proofreading device. It detects misspelled words and displays correct spellings as candidates, allowing for easy correction.

To this end, the configuration of the present invention includes a pronunciation dictionary that gives the pronunciation of written words, and a text analysis unit, and the text analysis unit receives the written text as input and analyzes the text using the pronunciation dictionary. A sentence reading proofreading device that generates a pronunciation sequence of a sentence, converts it into an audio signal, reads it out and proofreads it, is provided with an error checking dictionary that associates error-prone spellings with correct spellings, and an error spelling corrector, and is equipped with a spelling correcting unit. The notation correction unit searches the error checking dictionary using search data of the reading dictionary when the text analysis unit performs sentence analysis using the reading dictionary, and corrects the incorrect spellings registered in the error checking dictionary. When a match is detected, the reading operation is stopped before the incorrect spelling, and the corresponding correct spelling is retrieved from the error checking dictionary and displayed and output as a candidate word. In accordance with the instructions, a correction process is performed to replace the incorrect notation with a correct display, and at that time, control is performed to skip a location that has once been the target of an error check so that it will not be the target of an error check again after that error check. Features.

FIG. 1 shows the basic configuration of the present invention.

In the figure, 11 is a text input unit, 12 is a text analysis unit that identifies words in the text and creates a pronunciation sequence;
13 is a reading dictionary that associates notation and pronunciation for each word; 14 is a reading string storage unit; 15 is an audio output unit; 1
Reference numeral 6 denotes an error correction unit for detecting and correcting errors; 17 is a dictionary for checking incorrect expressions that associates incorrect expressions with correct expressions; and 18 is a display output unit.

The functions of the block elements 11 to 15 in FIG. 1 basically correspond to the functions of the block elements 51 to 55 in the conventional device shown in FIG.

The notation data of the sentence inputted from the sentence input section 11 is compared with the notation of many words registered in the reading dictionary 13 in the sentence analysis section 12, and each word constituting the sentence is identified.

The sentence analysis unit 12 creates pairs of spellings and pronunciations for each of the series of identified words, and sequentially stores them in the pronunciation storage unit 14.

Next, the misspelling correction section 16 reads out the spellings of each word stored in the pronunciation sequence storage section 14, and checks the presence or absence of misspellings by comparing them with the contents of the misspelling checking dictionary 17. If an incorrect notation is detected here, the reading sequence storage unit 14
, add its position and correct spelling candidates from the dictionary for checking incorrect spellings.

After this, the pronunciation of each word in the pronunciation sequence storage section 14 is
It is sent to the audio output section 15, and the corresponding notation (excluding correction candidates) is sent to the display output section 18. The audio output unit 15 converts the pronunciation into an audio signal through sequential speech synthesis and outputs the converted audio signal. Further, the display output unit 18 displays the notation in text format on the display.

The operator (proofreader) compares the read out content and the content displayed on the display with the manuscript, etc., and proceeds with proofreading.

Every time the pronunciation of a word is read out from the pronunciation sequence storage unit 14 to the audio output unit 15, the error correction unit 16 checks whether the pronunciation of a word is in an incorrect spelling position or not, and if it is in an incorrect spelling position, it reads it out from the audio output unit 15. The operation is stopped, and at the same time, a correction candidate notation corresponding to the incorrect notation is read out from the pronunciation sequence storage section 14 and displayed on the display output section 18.

When the operator determines that the notation of the displayed correction candidate is valid, the operator gives an instruction for correcting the incorrect notation and replacing it. Based on the instruction, the misspelling correcting unit 16 replaces the misspelling in the pronunciation sequence storage unit 14 with a correction candidate.

[Effect]

According to the present invention, incorrect spellings such as so-called misspellings and correct spellings that exist in an input sentence are automatically detected within the device using a dictionary for checking incorrect spellings, and the operator is notified. Since corrections to incorrect spellings are made based on judgment, highly reliable corrections are possible.

In the typographical error checking dictionary, any examples of incorrect notation can be registered in advance, and it is possible to uniformly check not only incorrect spellings, incorrect spellings such as correct characters, and errors in sending letters, but also specific usage specified by the user. It is also possible to check whether it is protected or not.

〔Example〕

FIG. 2 shows details of the processing procedure in one embodiment of the present invention. In addition, in FIG. 2, reference number 1
1 to 18 correspond to the elements with the same numbers shown in FIG. 1, and the explanations given for the basic functions of these elements in FIG. 1 also apply to FIG. Ru.

First, the processing procedure of the text analysis section 12 will be explained.

In the reading dictionary 13, a number of words are registered with corresponding spellings 131, readings 132, and usages 133. When the text analysis unit 12 is given a sentence to be proofread from the text input unit 11 or a document file (not shown), it uses a word identification process 121 to classify the notation of the input sentence into various categories. The written notation is compared with the notation 131 of each word in the reading dictionary, and a series of words constituting the sentence is identified (recognized). For this identification process, information on the usage 133 in the reading dictionary 13 is used.

Next, the sentence analysis unit 12 extracts the corresponding reading 132 from the reading dictionary 13 for each word notation based on the identified word string, pairs it with the notation 131, and stores it in the reading sequence storage unit 14.

The pronunciation sequence storage unit 14 arranges data of a series of words constituting a sentence in the order of their pronunciation, and each data has a pronunciation 141, a pronunciation 142, and an incorrect spelling flag 14.
3. Contains correction candidates 144.

Next, the processing procedure of the spelling error correction unit 16 will be explained. The processing of the misprint correction unit 16 is divided into two: a misprint check process 161 and a misprint correction process 162.

In the error spelling check process 161, the spellings 141 of each word are sequentially read out from the pronunciation sequence storage unit 14, and
It is compared with the erroneous notation of the erroneous notation data in the erroneous notation checking dictionary 17. In the error spelling check dictionary 17, incorrect spellings 171 and correct spellings 172 are registered for words that are likely to be mistaken.

As a result of the matching, if a matching incorrect notation 171 is found, the incorrect notation flag 143 in the reading string storage section 14 is set.
ON (“1”), and furthermore, the correct notation 172 of the error notation check dictionary 17 is written in the correction candidate 144.

In this way, after checking all the notations 141 in the pronunciation sequence storage unit 14, the audio output unit 15 displays the pronunciations 141 in the pronunciation sequence storage unit 14.
Executes a reading action based on the column. Also,
The column with the notation 141 is displayed by the display output unit 18.

Here, the error correction unit 16 performs error correction processing 1.
62, during the reading operation, the reading sequence storage unit 14
The incorrect spelling flag 143 is monitored in synchronization with the reading 142 read from the
When ON, stops the reading operation. At the same time, the corresponding correction candidate 144 is read out and displayed on the display output section 18. There may be multiple correction candidates.

In response to this, the operator decides whether to adopt or not adopt the correction candidate and gives a corresponding instruction.

The error correction process 162 of the error correction unit 16 is as follows:
When the adoption of a correction candidate is instructed, the notation 141 in the reading string storage section 14 is replaced with the correction candidate 144,
Resume reading operation.

To give a specific example, if the sentence to be proofread is "This is Kaishin's work," the result of word identification by the text analysis unit 12 will be as shown by the following underline.

This is the work of Kaishin.The error correction unit 16 compares the above word string with the error spelling check dictionary 17, and uses the word ``Kaishin''.
is detected to be an error, and "Kishin" is written in the reading string storage section 14 as a correction candidate.

The reading operation is performed, and when "Kaishin" is uttered (or even before it is uttered), the reading is stopped and the correction candidate "Kaishin" is displayed on the display.

If replacement with the correction candidate "Kaishin" is instructed here, the proofread text will be as follows.

``This is a masterpiece.'' By the way, as described above, when a typographical error is corrected, the typographical error check process 161 is re-executed.

However, when the error checking process is re-executed each time a correction is made, the incorrect spelling check dictionary displays "for example" (incorrect) → "for example" (correct), as shown in the figure.
When there is a designation such as "for example" (incorrect) → "for example" (correct), "for example" in the sentence should be "for example".
(Incorrect) → “For example” (Correct) → “ For example ” (Incorrect) →
Checking and correcting creates a loop, such as `` for example '' (correct) → ``for example'' (incorrect) → ``for example'' (correct), which hinders proofreading work.

To prevent this, once a part of a proofread text is subject to correction of spelling errors, it is displayed as checked with a flag, regardless of whether it has been replaced with a correction candidate, and when the spelling error check is restarted, the flag is set. It is possible to exclude the parts that are checked and to skip them. This checked flag can be set for each character in the pronunciation string storage section 14 or the like.

A specific example is shown in FIG. is the sentence to be proofread, and the words ``for example'' and ``feelings'' in it are checked in the dictionary for checking typographical errors. First, when reading aloud, it stops at ``for example,'' and is replaced by ``for example,'' as in . Here, a flag ("1") is set in the "for example" part. Next, the spelling error check is executed again, but since the flag is set for ``for example,'' this part is skipped. However, the incorrect spelling of "feelings" is checked, so when reading is restarted next time, it is stopped at "feelings" and replaced with "feelings", as in . As shown here, the "feeling" part is flagged, and the next time the check is run again,
``For example'' and ``feelings'' are skipped. This prevents "for example" and "for example" from going round and round, and "feelings" in "feelings" is modified to "feelings" and becoming "feelings," and then "feelings" becomes "feelings." Chichi”,
...can be prevented from continuing.

FIG. 4 shows an example where correction candidates are not replaced.

In the figure, is the text to be proofread, and ``stamping'' is checked in the dictionary for checking errors. Therefore, when reading out the text, the reading stops at "seal" and "seal" is displayed as a correction candidate. However, in this case, no replacement is specified, and a flag is set on the first character "oshi" of "oshiki" as shown in . After this, the check is executed again, but the check for words starting with "push" is skipped. Therefore, when reading is resumed, it is no longer stopped by "pressing".

〔Effect of the invention〕

According to the present invention, it is possible to automatically detect incorrect notations, misspellings, or usages different from specified ones in a text, and present correction candidate notations to the operator (proofreader) to replace them. Work efficiency can be significantly improved, and calibration accuracy can be improved.

[Brief explanation of drawings]

FIG. 1 is a basic configuration diagram of the present invention, FIG. 2 is a configuration diagram of one embodiment of the present invention, and FIGS. (Part 2) FIG. 5 is a basic configuration diagram of a conventional device. In Figure 1, 11: text input section, 12: text analysis section, 13: reading dictionary, 14: reading sequence storage section, 1
5: Audio output section, 16: Misprint correction section, 17: Dictionary for checking misprints, 18: Display output section.

Claims (1)

[Scope of Claims] 1. A pronunciation dictionary that gives the pronunciation of a written word and a text analysis section are provided, and the text analysis section receives the written text as input and analyzes the text using the pronunciation dictionary. A sentence reading proofreading device that generates a pronunciation sequence, converts it into an audio signal, reads it out, and proofreads the text, which is provided with an error checking dictionary that associates error-prone spellings with correct spellings, and a misspelling corrector, to correct the misspellings. The correction unit searches the error checking dictionary using search data of the reading dictionary when the sentence analysis unit performs sentence analysis using the reading dictionary, and searches the error checking dictionary to compare the incorrect spellings registered in the error checking dictionary. Performs an error check for a match, and when a match is detected, stops the reading operation before the incorrect spelling during the reading operation, extracts the corresponding correct spelling from the error checking dictionary, and displays and outputs it as a candidate word; A correction process is performed to replace the erroneous notation with a correct display in accordance with an instruction, and at that time, control is performed to skip a part that has once been the target of an error check so that it will not be the target of an error check again after the error check. A text reading proofreading device.