JPH0442382A - Word reader - Google Patents

Abstract

PURPOSE:To accurately decide a word at all times by tentatively converting all the KANJI (Chinese characters) in a candidate characters to a HIRAGANA (cursive form of Japanese syllabary writing) when there is a possibility that the HIRAGANA is included in the inputted words, collating them with a HIRAGANA word dictionary, and deciding the inputted word while using the result of collation together. CONSTITUTION:First, the collation is performed to collate the candidate word obtained from a character recognition means 3 in the order named with a KANJI word dictionary 6. The obtained words are outputted as they are when the collation with the specific KANJI words is succeeded and the candidates of an upper grade than individual candidate characters forming the word are all KANJI on the obtained KANJI words, and when they are not, all the KANJI in the candidate characters are tentatively converted to the HIRAGANA by a KANJI/KANA conversion table 8 to collate them with a HIRAGANA word dictionary 7, and the word with the highest accuracy is outputted from among the obtained HIRAGANA words and the KANJI words decided here. Thus, the word with orthography altering variously by the KANJI, the KANA, and the one mixed with the KANJI and the KANA, etc., can be accurately decided independent of its orthography.

Description

[Detailed Description of the Invention] [Industrial Application Field] This invention relates to a word reading device for reading words written on a form, and in particular for reading words that have variations in written expression such as kanji, kana, kanji, and kana. This relates to a reading device.

[Conventional technology]

FIG. 7 is a configuration diagram of a conventional word reading device shown in, for example, "Utilization of word information in handwritten kanji recognition" at the 1981 National Conference of the Institute of Electronics and Communication Engineers (1341). 2 is a form on which words are written; 2 is a scanning means for optically scanning and photoelectrically converting the words on the form I;
3 is a character recognition means for cutting out and recognizing each character of the photoelectrically converted word and outputting recognition candidate characters; 4 is a word dictionary storing words to be entered in the form 1; and 5 is an output from the character recognition means 3. This is a word determining means that determines an input word by comparing the selected candidate characters with a plurality of words included in the word dictionary 4.

FIG. 8 shows an example of an input word and its candidate characters.

9 is an input word, and 10 is a candidate character corresponding to each character of the input word 9. As shown in the figure, the candidate characters 10 are ranked in order starting from the first place.

FIG. 9 is an example of a word dictionary. 11 is the word "Ishikawa," 12 is the word "Ishida," 13 is the word "Ishida," 14 is the word "Migita," and 15 is the word "Migaya."

Next, the operation of this conventional example will be explained. When the input word 9 shown in FIG. 8 is written on the form 1, the scanning means 2 photoelectrically converts it and outputs it to the character recognition means 3. The character recognition means 3 cuts out and recognizes individual character patterns from the obtained pattern, and outputs a plurality of candidate characters 10 ordered in descending order of similarity. In Figure 8, the first candidate character °'right for the first character of the input word,
The 2nd place candidate character "stone" is output, and the 1st place candidate character "Pain", the 2nd place candidate character "田", and the 3rd place candidate character "dan" are output for the second character of the input word. Word determination The means 5 collates the candidate character 10 with the word in the word dictionary 4.As a specific method of collation, for example, first, the word dictionary 4 is
jth word A.

Calculate the evaluation value PJ. Here, n indicates the number of characters in the word.

PJ-Σp.

If there exists A1 such that Pk/n = min (P, /n) < T (T is a threshold), then A is taken as the determination result.

An example in which the candidate character 10 shown in FIG. 8 is compared with the word dictionary 4 shown in FIG. 9 will be explained in detail. In this example, the above judgment condition 1=1
0. Let T=5. Furthermore, since the input word 9 is a two-character word, n=2. The word determining means first compares it with the first word "Ishikawa" 11.

“Ishi” is in second place, and “kawa” is not a candidate, so PI = 12. Next, it is compared with the second word “Ishida”, which is 12. Since “1 stone” is in second place and “da” is in second place, P2 = 4. Thereafter, "Ishida" 13, "Migida" 14, and "Migaya" 15 are compared in the same way, and P3=12°P4=3.P5=11 is obtained.

Here min (P, H/n) = Pa /n < T (
-5) holds, so the fourth word "Migita" 14 is output as the determination result.

Without the word determination means 5, an incorrect result such as "right figure" would have been obtained, but by providing the word determination means 5 and comparing the words with the words in the word dictionary 4, the result is " The correct word "Migita" came out.

[Problem to be solved by the invention]

Conventional word reading devices are configured as described above, so the words written on the form must be stored in a word dictionary including kanji, kana, mixed kanji and kana, etc. However, there are problems in that, although it is possible to correctly determine the notation stored in the word dictionary, it cannot be determined correctly with the notation that is not stored.

This invention has been made in order to solve the above-mentioned problems, and it is an object of the present invention to provide a word reading device that can correctly determine words whose notation varies in various ways due to kanji, kana, mixture of kanji and kana, etc., regardless of the notation. The purpose is to

[Means to solve the problem]

The word reading device according to the present invention includes a scanning means 2 for reading words on a form 1, a scanning means 2 for reading words on a form 1, recognizing each character pattern constituting the read word, and outputting a plurality of candidate characters ordered in descending order of similarity. a kanji word dictionary 6 that stores the words written on the form 1 in kanji, a hiragana word dictionary 7 that stores all the contents of the kanji word dictionary 6 replaced with hiragana, and the kanji characters. The kanji-kana conversion table 8 indicating the corresponding hiragana character by character for all kanji in the word dictionary 6 and the candidate characters 17 output from the character recognition means 3 are input, and the kanji-word dictionary 6 is
, word determining means 5 for determining a word by referring to the hiragana word dictionary 7 and the kanji-kana conversion table 8.

[Effect]

The word determining means 5 in this invention first compares the candidate characters 17 obtained from the character recognition means 3 and their order with the Kanji word dictionary 6, and if the matching with a specific Kanji word is successful and the For a Kanji word obtained, if all of the candidates higher than the individual candidate characters that make up the word are Kanji, the obtained Kanji word is output as is. All kanji are temporarily replaced with 27 hiragana and compared with the hiragana word dictionary 7, and the word with the highest accuracy among the obtained hiragana words and the kanji conflict words determined above is output.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.
Scanning means 22 Character recognition means 3. The word determining means 5 shows the same components as the conventional example. .

In the figure, 6 is a kanji word dictionary that stores the words written in the form 1 in kanji, 7 is a hiragana word dictionary in which all the contents of the kanji word dictionary 6 are replaced with hiragana, and 8 is a kanji word dictionary that stores the kanji in the kanji word dictionary 6. This is a kanji-kana conversion table that converts hiragana to hiragana.

Figure 2 is an example showing an input word and its candidate characters.
17 is an input word, and 17 is a candidate character. In the example shown in the figure, for the first character of input word 16, the first candidate character is "Miya" and the second candidate character is "Ei", and for the second character, the first candidate character is "Za". The second candidate character is "gi°", the third candidate character is "", and for the second character, the first candidate character is "sa" and the second candidate character is "ki".

Figure 3 shows an example of the kanji word dictionary 6, where 18 is the word "Miyanohara", 19 is the word "Yamanoue", and 20 is "
The word is Haseyo.

Figure 4 shows an example of Kanji-kana conversion table 8, with 21
is the kanji “gu”, 22 is the kanji “ei”, 23 is the kanji “之”
It shows. Zarani 24 is the hiragana "miya" corresponding to "miya" 21, and 25 is the hiragana corresponding to "ei" 22.
``Ei'', 26 is the hiragana ``no'' corresponding to ``之'' 23.

FIG. 5 shows candidate character 17 shown in FIG. 2 in which all the kanji are replaced with hiragana, and 27 is a candidate character consisting only of hiragana.

FIG. 6 shows an example of the hiragana word dictionary 7, where 28 is “
29 is the word ``Enomoto,'' and 30 is the word ``Miyazaki.''

Next, the operation of this embodiment will be explained. After performing the same processing as in the conventional example, the word determining means 5 obtains the candidate characters 17 of the input word 16, and first compares them with the Kanji word dictionary 6. As a specific example, consider the case where the input word 16 in FIG. 2 is compared with the kanji word dictionary 6 in FIG. 3. The matching method uses the word evaluation value PJ as in the conventional example, and the values of the constants used therein are also the same as in the conventional example. However, since the input word 16 has three characters, n=3. In this comparison example, first, a comparison is made with the three-letter word "Miyanohara" 18 in the Kanji word dictionary 6. “Miya” is in first place, “No” is in second place, and “Hara” is not a candidate, so Pl=13. Similarly, “Yamanoue” 19. Comparison with “for Hase” 20 is performed, and P2 = 22. Pa=
Get 30. Here min (PJ/n) =P+
Since /n < T (-5), the matching with "Miyanohara" 18 is successful, but since the hiragana "za" is among the top candidates for "no", the word determination means 5 outputs "Miyanohara" 18 as the determination result. Instead, the process proceeds to the next step and converts all the kanji in the candidate characters 17 into hiragana using the kanji-kana conversion table 8 shown in FIG. Specifically, Miya” 21. “Ei” 22.
The hiragana ``miya'' 24 temporarily corresponds to ``no'' 23. Convert "Ei" 25.6 to "26".

As a result, candidate character 17 in FIG. 2 becomes candidate character 27 in the form shown in FIG. Next, this candidate character 27 is compared with a hiragana word dictionary 7 as shown in FIG. The matching method here is basically the same as in the conventional example, but for the candidate character 27 converted from the kanji candidate character 17 to multiple hiragana characters (for example, "miya" and "ei" in Figure 5). In this case, only when all the hiragana characters corresponding to that kanji character match, the rank of the kanji character before conversion is used as the rank of each individual hiragana character, and in other cases, it is not a candidate.

This will be explained using an example in which the candidate characters 27 in FIG. 5 are compared with the hiragana word dictionary 7 in FIG. 6. Word determination means 5
first compares the candidate character 27 with "Aizawa" 28. Here, the first two characters of the candidate character 27 are converted from one Kanji character, so if both characters do not match, the candidate character 27 is not a candidate. The first two characters of “Aizawa” 28 do not match any of the characters, so they are both non-candidates, the third character “’za” ranks first, and the fourth character “wa” is a non-candidate, so it is P.

It becomes 1-31. Next, there is a comparison with "Enomoto" 29, but all of them are not candidates, so P. It becomes 2-40. ``Miyazaki'' 30, the first two characters ``Miyapa'' are among the candidate characters 27, and the original kanji candidate character is 1.
Therefore, both the first and second characters are ranked first. 3
The letter ``zapa'' ranks first, and the fourth letter ``ki'' ranks 2nd, so P.3=5.

Finally P. 1 to PO3, and P, which was successfully matched as a kanji word, min (P, /n) −PO3/n <T (=
Since 5) holds true, "Miyazaki" is determined as a candidate word. However, when outputting the result, the ``miya'' 24, which was temporarily replaced with hiragana, is changed back to the kanji character ``'宮'' 21 and output as the word ``Miyazaki''.

In addition, in the above embodiment, one set of hiragana is shown to correspond to one kanji in the kanji-kana conversion table, but
A plurality of sets of hiragana may be associated with one kanji.

〔Effect of the invention〕

As described above, according to the present invention, a kanji word dictionary, a hiragana word dictionary, and a kanji-kana conversion table are added to the word determination means, and when there is a possibility that an input word contains hiragana, all candidate characters are Kanji are temporarily converted to hiragana using a kanji-kana conversion table, and then compared with a hiragana word dictionary, and the results of the comparison are also used to determine the input word. This has the effect of allowing words to be determined correctly without being affected by fluctuations.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention, FIGS. 2 and 8 are example diagrams showing input words and their candidate characters, FIG. 3 is a diagram showing an example of a kanji word dictionary, and FIG. 4 is a diagram showing an example of a kanji word dictionary. The figure shows an example of a kanji-kana conversion table, Figure 5 shows the candidate characters in Figure 2 with all kanji characters replaced with hiragana, and Figure 6 shows an example of a kanji-kana conversion table.
The figure shows an example of a hiragana word dictionary, FIG. 7 is a block diagram of a conventional device, and FIG. 9 is a diagram showing an example of a word dictionary used in a conventional device. DESCRIPTION OF SYMBOLS 1... Form, 2... Scanning means, 3... Character recognition means, 5... Word determining means, 6... Kanji word dictionary, 7
...Hiragana word dictionary, 8...Kanji-kana conversion table. In addition, in the figures, the same reference numerals indicate the same or corresponding components. Agent Patent Attorney Jun Miyazono

Claims (1)

[Claims] a scanning means for reading words on a form, a character recognition means for recognizing each character pattern constituting the read word and outputting a plurality of candidate characters ordered in descending order of similarity; A kanji word dictionary that stores words in kanji, a hiragana word dictionary that stores all the contents of the kanji word dictionary replaced with hiragana, and kanji that indicate the hiragana that corresponds to each character for all the kanji in the kanji word dictionary. comprising a kana conversion table, and word determination means that receives candidate characters output from the character recognition means and determines a word by referring to the kanji word dictionary, the hiragana word dictionary, and the kanji kana conversion table; From the candidate characters and their order obtained from the character recognition means, the word determination means first compares them with a Kanji word dictionary, and for the Kanji words that are successfully matched with a specific Kanji word and obtained, If the candidates higher than the individual candidate characters that make up the word are all kanji, the obtained kanji word is output as is. If not, the Kanji-kana conversion table above is used to temporarily convert the kanji among the candidate characters. A word reading device characterized by replacing all hiragana words with hiragana words, comparing them with a hiragana word dictionary, and outputting the word with the highest accuracy among the obtained hiragana words and the above-determined kanji words.