JPS63284676A - Character string processor - Google Patents

Abstract

PURPOSE:To effectively process numerals by recognizing the whole of a character string as a numeral in case said character string consists of numeric characters and auxiliary numeric characters such as prefixes or suffixes. CONSTITUTION:A numeric character table 20 stores Arabic numerals, Chinese numeric characters, numeric characters showing digits, the equivalent characters of the Chinese numeric characters, and flags showing the sections of these numeric characters. An auxiliary numeric prefix table 22 stores the auxiliary numeric prefixes like 'OYOSO', 'ZU', 'DAI', etc., and an auxiliary numeric suffix table 24 stores the auxiliary numeric suffixes like 'KAI', 'KO', etc. When the characters corresponding to these prefixes and suffixes are inputted, the auxiliary numeric prefixes and suffixes are regarded as a group and undergo the analysis of numerals through a numeral processing part 14. While such words as 'ICHININ MAE', 'GORI MUCHU', etc., are stored in an inhibition table 26 and undergo no analysis of numerals through the part 14.

Description

TECHNICAL FIELD The present invention relates to a character string processing device, and more particularly to a device for processing character strings including numerals.

BACKGROUND ART When analyzing Japanese character strings for purposes such as translation, it is necessary to analyze numerals as preprocessing for morphological analysis and syntactic analysis.

Conventional devices were able to analyze numbers such as l, 2, and 3, but they were unable to analyze character strings containing numbers, such as "Table 1."
Only ``rlJ'' was recognized as a number word, and the entire ``1 Table 1'' could not be recognized as a number word. Similarly, the ``billion'' in ``200 million yen'' and the ``number'' in ``several tens'' cannot be recognized as number words, so ``two'' and ``billion yen'' and ``number'' and ``junin'' are used as numbers.
He was unable to recognize the words ``200 million yen'' and ``several tens of people'' as whole numbers.

Furthermore, since ``yen'' was registered in the dictionary along with words representing the digits of number words, a large dictionary capacity was required.

Furthermore, expressions containing both arithmetic numerals and Chinese numerals could not be recognized as numerals at the same time, making it impossible to perform appropriate analysis.

OBJECTS It is an object of the present invention to provide a character string processing device that can eliminate the drawbacks of the prior art and can efficiently detect numerals in Japanese character strings.

Configuration In order to achieve the above object, the present invention includes an input means for inputting a character string, a number storage means for storing numbers and characters equivalent to Chinese numerals, and a number storage means for storing number characters that constitute a number word when used together with the numbers. and a number word processing means that searches the number storage means and the song number character storage means for a character string input from the input means and determines whether or not the character string is a number word. , the number word processing means is
Even if the character string consists of numbers and number letters, the character string is recognized as a number word as a whole.

Hereinafter, a detailed explanation will be given based on one embodiment of the present invention.

FIG. 1 shows an embodiment of a character string processing device according to the present invention.

This device has an input unit 10, into which a Japanese character string containing kanji and kana is input. For example, the input unit 1O may be a keyboard having character keys, function keys, etc., an optical character reader (OCR) that reads Japanese character strings recorded on paper, and Japanese text stored on a storage medium such as a magnetic disk. It may include a file storage device for reading, etc.

The input character string file 12 temporarily stores Japanese character strings input from the input unit 10. The numeral processing unit 14 converts the numerals included in the Japanese kanji/kana mixed character string read from the input character string file 12 into a number table 20,
Number words are processed by referring to the song number prefix table 22, the song number suffix table 24, and the prohibition table 26, and information on the number words is added and output to the output unit IB.

An example of data stored in the number table 20 is shown in FIG. As shown in the figure, the number table 20
Arithmetic numerals, Kanji numerals, Kanji numerals representing digits, and characters equivalent to Kanji numerals are stored in , and a flag of 0 to 3 is set for each. Kanji numeral equivalent characters are characters used in place of numbers, such as ``what'', ``number'', ``number'', etc., as shown in Figure 2.

An example of the song number prefix table 22 is shown in FIG.

As shown in the figure, the song number prefix table 22 stores characters used at the beginning of numerals including numbers.

An example of the song number suffix table 24 is shown in FIG.

As shown in the figure, the song number suffix table 24 stores characters used at the end of number words.

An example of the prohibition table 26 is shown in FIG.

As shown in the figure, there are some strings of characters that are registered that are idioms or idiomatic expressions that have unique meanings and should not be subjected to number processing. The prohibition table 28 is accessed during prohibition processing, which will be described later.

The number word processing in the number word processing unit 14 converts the number word part into the following four
It is classified into two patterns.

(a) Song number prefix + number + song number suffix example: 5th, approximately 120 years (b) Number + song number suffix example = 15 people, several hundred (C) Song number prefix + number example : Table 1, Figure 2 (d) Number example 26, 1900 Therefore, when the number word processing unit 14 detects a number or a song number prefix, it enters number word processing, skips reading as many numbers as they last, and reads the numbers when there are no more numbers. up to the number part. If there is a song number suffix immediately after that, it will be summarized as a number part. Further, by referring to the prohibition table 26 for phrases that include number words that should be treated as a group, the number words are not analyzed.

The output unit 16 includes, for example, a printer, a display, and a file storage device such as a magnetic disk.

Next, the operation of number word processing in the number word processing section 14 will be explained with reference to the flowcharts shown in FIGS. 6(a) and 6(b).

The pointer P indicating the beginning of the character string to be analyzed and the beginning Q of the numeric part are initialized (102), and the presence or absence of the character string is checked (104), and if not, the process is terminated.

If there is a character string, prohibition processing is performed (106). That is, the prohibition table 26 is searched to check whether an idiom containing a number word is at the beginning. Details of this prohibition processing will be described later.

As a result of the prohibition process, if it cannot be parsed, that is, it is a group of idioms that should not be processed with number words, it returns =O, and if it can be analyzed, that is, it can be processed with number words, it returns =1.
return it. Next, determine whether K is O or not +08)
, K = O, the process returns to step 104 without performing numeral processing and proceeds to the next character. If K=0, numeral processing is performed.

First, by checking the number table 20, it is determined whether or not there is a number at the beginning of the character string (110).

If the beginning of the character string is a number, this part is regarded as a numeral part, the value of pointer P indicating this position is stored in H (124), and the process proceeds to step 128.

If the beginning of the string is not a number, determine whether there is a song number prefix at the beginning of the string (+12), and if there is no song number prefix, advance the pointer to the next character (114).
, return to step 104.

If there is a song number prefix, it is considered the beginning of the number part,
Save the value of pointer P indicating this position in H (11B
). The pointer is advanced by the length of the prefix (118), and the position of the first character resulting from advancing the pointer is stored in Q (120). The first character is determined whether it is a number or not (122), and if it is not a number, it is recognized that it is not a numeral part, the position Q of the first character is initialized, and the pointer P is set to 1.
Advance one character to the next character (12B), Step 1
Return to 04.

If the first character is a number, advance the pointer P by one,
Proceed to the next character (128). Further, even after the value of pointer P is stored in H in step 124, pointer P is similarly advanced by one (128) to proceed to the next character.

Determine whether the next character is a number by checking the number table 20 (130); if it is a number, step 1
Return to 28, advance pointer P by one, advance to the next character,
Similarly, it is checked whether it is a number or not (130).

If the character is no longer a number, the loop is exited, and whether or not that character is a song number suffix is determined by searching the song number suffix table 24 (132). If it is a song number suffix, determine whether Q>H or not.134)
, If Q>H, move from H to Q-1 with the starting position saved.
is recognized as a prefix, the number part from the first position Q to pi is recognized as a number part, and the part from the first position P of the suffix to the end is recognized as a suffix (13B). That is, the number words in the pattern (a) above are recognized.

If Q>H, there is no prefix, so the part from H to P-1, where the first position of the numeral part is saved, is recognized as the numeric part, and the part after P is recognized as the suffix (13B). In this case, the number words in the pattern (b) above are recognized.

If there is no song number suffix in step 132, then Q
>H or not (140), if Q>H, the prefix is from H to Q-1 where the starting position is saved,
The part from the beginning position Q of the number part to P-1 is recognized as the number part (13B). That is, the number words in the pattern (C) above are recognized.

If Q>H, there is no prefix, so the area from H to P-1, where the leading position of the numeral part is saved, is recognized as the numeric part (138). In this case, the number words in the pattern (d) above are recognized.

After the numeral part is detected in this way, after steps 138 and 138 where the suffix was present, the pointer is advanced from P at the beginning of the suffix by the length of the suffix, and the process returns to step 104 again. After steps 142 and 144, the process directly returns to step 104.

Next, the prohibition process in step 10B will be explained with reference to the flowchart of FIG.

The prohibition process detects, before detecting the number word part, phrases that are made up of Chinese numerals or characters equivalent to Chinese numerals and should be treated as a group, and prevents them from being processed as a number word. Examples of words that should not be processed in this way include ``Shimen Souka'', ``Gori Kirinchu'', and ``Shiragami Sanzenjo''.

Number processing must not be performed on the input character string, that is, analysis is prohibited in the following four patterns.

(1) A kanji character with a number or song number suffix immediately after it.
...Analysis possible Example: Quenching, number tens (2) Kanji numeral kanji with no numeral or song number suffix immediately after, and no kanji numeral kanji... Example where analysis is prohibited :Number, where, geometry (3) Numbers other than those in the room corresponding to Chinese numerals are on the prohibited table.
...Analysis prohibited Example: Shimen sanka (4) Numbers other than the Chinese numeral equivalent room and not in the prohibition table...Analysis possible Example: In the figure 7 for four men and five people, the beginning of the character string is The number table 20 is checked to determine whether or not the character is a Chinese numeral character such as "what" or "number" (202), and if it is a Chinese numeral character, the pointer is advanced by one (204).

Check the song number suffix table 24 to determine whether the next character is a song number suffix (20El), and if it is a song number suffix, determine that it can be analyzed, and move the pointer back by one 210
), returns =1 to indicate that it can be analyzed (212
).

If it is not a song number suffix in step 206, it is checked by the flag of the number table whether it is a Chinese numeral or not. 210), returns =1 to indicate that it can be analyzed (212).

If it is not a Chinese numeral, it is determined that analysis is prohibited, the pointer is advanced by one (214), and =O is returned to indicate that analysis is prohibited (21B).

In step 202, if the beginning of the character string is not a kanji kanji character, it is checked whether this character string exists in the prohibition table 2B (21B), and if it is not in the prohibition table 26, the pointer is left as is and the analysis is performed. =1 is returned to indicate that it is possible (220).

If it is in the prohibition table 2G, analysis is prohibited, so
The pointer is advanced by the length of the character string in the prohibition table 26 (222), and =O indicates that analysis is prohibited.
is returned (224).

In this way, prohibition processing is performed.

Next, the operation of this device will be explained by giving a specific example.

The input sentence is ``I am the 25th...'' as shown in Figure 8.
When the sentence "・" is input, the pointer P is set to 1 and the pointer Q is initialized to 0 (102).

Since there is a character string (104), we move on to the prohibition processing in Figure 7 and write, ``I J is not a Chinese numeral konduji (202), and it is not in the prohibition table (2+8), so it can be analyzed (220) .$ 6 Return to 7o- in Figure (a), K=
Not an O (+08), not a number at the beginning (+10)
, since it is not a song number prefix (112), the pointer is set to 1.
Advance one step (+14), advance to the next character "wa" (114)
.

``Ha'' is also processed in the same way as ``Washi'' and proceeds to the next character ``No.'' (+14). In prohibition processing 10B, the process moves to Figure 7, and ``No.'' is a Chinese numeral konduji ((202) ), it is not in the prohibited table (218), so it can be analyzed (220
), returning to the flow in Figure 6(a), if K=0 instead of
(108), since the beginning is not a number (110) but a song number prefix (112), the number of pointer P indicating this position is stored in H (116). That is, H=3
It is said that Then advance the pointer by the length of the prefix (
+18). In this case, the prefix "No." is a single character, so
Advance pointer P by one.

As a result of advancing the pointer, position 4 of the character "2" which becomes the first character is stored in Q (120). The first character "2" is a number (122), so the pointer advances by one (12,
8) , move on to the next character. The next “ten” is a number (1
30), proceed to the next character (128), and since "five" is also a number (130), proceed to the next character (12B). When the pointer reaches 7 and reaches the character ``times'', it is not a number (130), so we exit this loop, and since ``times'' is a song number suffix (132), we proceed to step 134 and select Q and H.
Compare with. In this case, Q=4, H=3, and Q>
Since it is H, it is the pattern of (a) above (13B)
.

Therefore, the prefixes are from H to Q-1, that is, from 3 to 4-1, and only the "th" of pointer 3 is the prefix. The numerical part is from Q to P-1, that is, from 4 to 7-1, and "25" from pointers 4 to 6 is the numerical part. The suffix starts from P, that is, from 7, and the suffix starts from "time" of pointer 7.

As a result, the numeral part is detected, so the pointer is advanced by the length of the suffix (14B), and the same process is performed for the next pointer 8, "no".

Next, the input sentence as shown in Figure 9 is [He studies geometry]. ” will be explained below.

"He" is not a Chinese numeral equivalent (202) in the prohibition process in Figure 7, similar to "I am" in the above example.
Since it is not in the prohibition table (218), it is analyzable (220), and in Figure 6 (a), = On' is not (+
08), since the beginning is not a number (110) nor is it a song number prefix (112), the pointer is advanced by one (11).
4), proceed to the next character "Iku" (114).

In the prohibition process in Figure 7, "Iku" is a character equivalent to a Chinese numeral (202), so the pointer is advanced by one (204).
, if you look at the next character "what", you will see that it is not a song number suffix but (2
0Ei), since it is not a kanji numeral equivalent (208),
Determines that parsing is prohibited and advances the pointer to the next character (
214), returns K=O (218). Figure $6 (
Returning to a), since K=O (10B), numeral analysis is performed and the next character is processed (104).

According to this embodiment, it is possible to analyze a character string including a numeral in consideration of patterns specific to numerals. In other words, prefixes and suffixes that were not traditionally treated as numbers,
For example, by treating characters equivalent to Chinese numerals and Chinese numerals representing digits as part of number words, the entire string of characters consisting of these and numbers can be recognized as a number word, and appropriate number words can be processed. Therefore, there is no need to individually register characters corresponding to Chinese numerals or Chinese numerals representing digits in the dictionary, and the dictionary capacity can be reduced.

Furthermore, character strings containing a mixture of arithmetic numerals and Chinese numerals can be correctly analyzed.

In addition, the prohibition process detects words that should be treated as a group, such as idioms and idiomatic expressions that include number words or number parts, and prevents number word analysis. It also prevents idiomatic expressions such as verbs from being mistakenly treated as numerals.

Effects According to the present invention, even when a character string consists of numbers and number characters such as prefixes or suffixes, the entire character string can be recognized as a number word. Therefore, number words can be processed efficiently.

Furthermore, since it is not necessary to individually register characters corresponding to Chinese numerals and Chinese numerals representing digits in the dictionary, the capacity of the dictionary can be reduced.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram showing an embodiment of a character string processing device according to the present invention, FIG. 2 is a diagram showing an example of data stored in the number table shown in FIG. 1, and FIG. Figure 4 is a diagram showing an example of data stored in the song number prefix table. Figure 4 is a diagram showing an example of data stored in the song number suffix table in Figure 1. Figure 5 is the prohibition table in Figure 1. 6(a) and 6(b) are flowcharts showing the operation of the device in FIG. 1, and FIG. 7 is a flowchart showing the operation of the prohibition process in FIG. 6(a). , FIG. 8 is a diagram showing an example of an input sentence inputted into the device shown in FIG. 1, and FIG. 9 is a diagram showing another example of an input sentence inputted into the device shown in FIG. 1. Explanation of symbols of main parts 10, Input section 12, Input character string file 14, Number processing section 16, Output section 20, Number table 22, Song number prefix table 24, ,, Song number suffix table 2B, , Prohibited table

Claims (1)

[Scope of Claims] 1. An input means for inputting a character string; a numeric storage means for storing numbers and characters equivalent to Chinese numerals; and an arbiter for storing an arbiter character that constitutes a numeral when used together with the numerals. character storage means; and numeral word processing means for searching the numeral storage means and the arbiter character storage means for the character string input from the input means and determining whether or not the character string is a numeral word. A character string processing device characterized in that the numeral processing means recognizes the character string as a numeral as a whole even when the character string consists of the numeral and the fractional character.