JPH0267684A - Calibration supporting system and dictionary retrieving system - Google Patents

Abstract

PURPOSE:To automatically uniform the deviation of notation which may generate a trouble when plural documents are written by using a corresponding table previously storing the deviation of notation as a rewriting rule to rewrite a character string. CONSTITUTION:A table writing notation with low using frequency on the left side and writing notation with high using frequency on the right side is previously stored in a rewriting rule storage means 108 as a rewriting rule. A dictionary retrieving means 105 retrieves a dictionary by using a retrieving key transferred from a text editing means 102. In addition, the means 105 commands a character string rewriting means 107 to rewrite a retrieving key and executes dictionary retrieval by using the rewritten retrieving key. The rewriting and retrieval are properly repeated and the rewritten contents of the retrieving key and the retrieved result are returned to the means 102. The means 107 applies the rewriting rule to the retrieving key on the basis of the command from the means 105 to rewrite the retrieving key.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to natural language processing technology, and is applicable to technology that analyzes languages with variations in the spelling of words, such as Japanese-English automatic translation and proofreading support. The present invention relates to a suitable proofreading support method and dictionary search method.

[Conventional technology]

In conventional techniques, for example, the following methods are used to cope with variations in notation regarding dictionary search methods. That is, a record is created in the dictionary using each notation as a key. the grammar and meaning of the word.

Information attached to a word, such as translation, is written in one of these records, and a pointer to the record in which the information is written is written in the remaining records.

Furthermore, as described in Japanese Patent Application Laid-Open No. 61-223979, the delimiter characters of compound words are rewritten to standardized characters before dictionary searches are performed.

In addition, as described in Japanese Patent Application No. 107677/1983, by rewriting kanji into kana character strings representing their pronunciations and registering and searching from ``C'', we can deal with the fluctuation that allows both kanji and kana to be written. It had become.

Also, "Direct search for foreign words written in katakana" (Nakamura et al., Information Processing Society of Japan, 35th National Conference (late 1986), p.
, 1223-p., 1224), the English spelling is restored from the katakana notation of the foreign word and the English dictionary is searched directly.

On the other hand, regarding the 1 proofreading support method, a product is on sale that sorts by using katakana characters as a key and displays the results in a quick (KWIC) list format.

[Problem to be solved by the invention]

Regarding the dictionary search method in the above-mentioned conventional technology, the method of writing a pointer in the dictionary is versatile, but it is necessary to comprehensively register variations in notation. If there are two or more places where one spelling varies, such as in compound words, it is necessary to consider them in combination.

For example, in the word "Multi-window J",
Whether you want to separate the text with , or with a space, or write it with a space, or write the word ``rai'' in ``window'' as ``rai,'' or write ``do'' as ``do'' or ``do.'' There are 18 ways to write it, depending on how you write it. If we were to cover all of these spelling variations for all foreign words, the amount of data would be enormous. Furthermore, there is a possibility that some items may be omitted from registration, making it difficult to guarantee the quality of the dictionary.

In the method described in JP-A No. 61-223979, the delimiter characters for compound words are standardized, and the same problem as above occurs when there is a variation in the notation of the constituent elements of a Pli compound word.

The method described in Japanese Patent Application No. 107677/1987 can deal with variations such as being able to write both kanji and kana, but
There is a problem in that it cannot deal with other variations in notation.

The above-mentioned method of "direct search for foreign words written in katakana" cannot be applied to foreign words whose source is a language other than the previously assumed foreign language, and in order to restore the English spelling, it is necessary to search for foreign words without any fluctuation in the spelling. Treatment is required for the location. Also,
There is a problem in that it is not possible to uniquely restore English spelling from the Japanese notation, which complicates the process.

Regarding the proofreading support method, displaying it in the form of a KWIC list is effective for finding spelling errors and input errors in one word, but it is effective for finding spelling errors and input errors between different words. There is also a problem that automatic correction is not sufficient.

For example, the text may contain "...activ data...
” and “...cognitive science...
・Suppose there is a notation ノ. There are variations in the spelling of syllables that have the same origin, such as ``tig'' and ``tig.'' However, in the case of "active" and "cognitive", the character strings "aku" and "cogni" before the part where the 1 notation deviates are different, so in the above-mentioned conventional technology based on the KWIC list, these two notation will be displayed in a separate area. For this reason, there is a problem in that it is difficult to detect variations in notation. Furthermore, even if such variations in notation are discovered, there is not a sufficient function to automatically unify them.

A first object of the present invention is to provide a proofreading support method that avoids the above-mentioned problems and detects variations in notation. A second purpose is to provide a proofreading support system that automatically unifies detected variations in notation.

Further, the third object of the present invention has a variation in notation of 1. To provide a dictionary search method that eliminates the need for exhaustive registration of spelling variations in the creation and use of language-related dictionaries and deals with spelling variations in a simple manner.

[Means to solve the problem]

First, the dictionary search method, which is the third objective, is a # number search method that uses a word dictionary storage means and a dictionary search means for searching the word dictionary using a search key. This is achieved by providing a processing step that utilizes a character string rewriting means that rewrites the string according to the following.

In addition, the first and second objects of the proofreading support method are one in the proofreading support method that utilizes a text editing means, a text storage means, a display means, and an input means.The third object of the present invention is a dictionary search method. This is achieved by providing a processing step of rewriting a predetermined portion of the text string with the rewritten string.

[Effect]

The rewriting rule storage means stores in advance a rewriting rule in which a less frequently used notation is written on the left side and a notation more frequently used than the notation on the left side is written on the right side. These rules are classified according to the frequency of use on the left side. What is the dictionary search method?

A dictionary search is performed using the search key passed from the text editing means. Further, the dictionary retrieval means indicates JH to rewrite the search key to the one character string 7F rewriting means, and performs a dictionary search using the rewritten search key. The 0-character string rewriting means repeats this rewriting and searching as appropriate and returns the search results, including the rewritten contents of the search key, to the text editing means.Based on instructions from the dictionary search means, the 0 character string rewriting means applies the rewriting rules to the search key and performs the search. Rewrite the key. Rewriting rules are applied starting from the least frequently used notation on the left side.

The text storage means stores the text to be proofread in advance. The text editing means cuts out a portion of the text stored in the text storage means where it is determined that there is a possibility that there is a possibility that there is a variation in the notation, and passes it to the dictionary search means as a search key.

Further, the text editing means receives the rewritten contents of the search key and the search results returned from the dictionary search means. The text editing means makes a final judgment as to whether or not there is any deviation in the 5 notation based on the rewritten contents of the search key and the search results. If it is determined that there is a fluctuation, the notation is automatically unified or the fact that there is a fluctuation is displayed via a display means.

〔Example〕

An embodiment of the present invention will be described below.

FIG. 1 is a diagram showing an example of an apparatus for realizing the present invention.

FIG. 2, FIG. 3, FIG. 4, and FIG. 5 are flowcharts showing an example of the calibration support method according to the present invention. The text editing means 102 in FIG. 1 operates according to this flow.

FIGS. 11 and 12 show data structures that maintain character correspondence when the length of a character string changes due to application of rewriting rules. The details of FIG. 11 and FIG. 12 will be explained later along with the explanation of the flow.

FIG. 13 shows the structure of the search result staff.

This will also be explained together with the explanation of the primary flow.

Hereinafter, first, the flows of FIGS. 2 to 8 will be explained using examples.

At step 201 in FIG. 2, attention is paid to the first character of the text stored in the text storage means 101 in FIG. 1, and this is designated as the character of interest. In step 202, the character of interest is compared with the character indicating the end of the text. If they match (step 203), the process ends. If they do not match, the process advances to step 301 in FIG. Here, check the type of character of interest. If the character of interest is neither katakana nor nakaguro [・], in step 303, the character next to the character of interest is set as the character of interest and
8 are flow diagrams showing an example of a dictionary search method according to the present invention. The dictionary search means 105 in FIG.
It operates according to the flowcharts shown in FIGS. The character string rewriting means 107 shown in FIG. 1 operates according to the flow shown in FIG.

FIG. 9 is an example of a rule group stored in the rewriting rule storage means 108 of FIG. 900 in FIG.
``Level'' is a number representing a classification based on the frequency of use of the notation on the left side of the rewriting rule.

The smaller the rule number, the less frequently the notation on the left side of the rule is used.

FIG. 10 is an example of a word dictionary stored in the word dictionary storage means 104 of FIG. Figure 10 1001.1
In 002, 1003, and 1005, "content" is "琐", which specifies that the heading is to be used as the correct notation. In addition, if you particularly want to use a notation that is not normally used, please refer to Figure 10 1004.
All you have to do is register the notation you want to use in "Contents" like this.

Return to chip 202. 202, 203, 301° 302 until the noted character becomes Katakana or Nakaguro.
Repeat each step of 303.

Now, in step 301, it is assumed that the character of interest, for example, is the first character "ma" of "multi-window system..." in the text. This is determined to be Katakana in step 302, and the process proceeds to step 304. Here, the range "Multi-Window System" starting from the noted character and followed by Katakana or Nakaguro is cut out from the text and substituted for the key (KEY) (0). In steps 305 and 306, preparations are made to call the flow of the dictionary search method. The search result stack (FIG. 13) is used to store information obtained from the dictionary as a result of dictionary searches. REWRITTEN indicates whether the search key has been rewritten by the character string rewriting means. In step 307 after going through steps 305 and 306, the flow in FIGS. 6 and 7 is called, and step 601 in FIG.
Proceed to.

In step 601 of FIG. 6, the length of KEY (0) is assumed to be L. Array variables 0RIGLOG (1) to 01 (IGL
OG (L), in order, natural numbers 1, 2, . ... Substitute L. Currently, KEY (0) is "multi-window ° system", so L is 14, and 0111GLO
The contents of G are as shown in FIG. 11th
In the figures, corresponding characters are written in parentheses to make the explanation easier to understand.

In the following step 602, KEY (0), that is, "multi-
Perform a dictionary search using the longest match using "Window System" as the search key. If the search is successful, write the search results to the search results stack. Furthermore, the number of characters that match the dictionary heading and KEY (0) is written in the "correspondence" item of the search result stack. However, the contents of the dictionary are the first
As shown in Figure 0, the current dictionary search is K.
There are no headings that match EY(0). Therefore, nothing is written to the search result stack.

Next, in steps 603 to 606 and the flows shown in FIGS. 7 and 8, a dictionary search is performed with the search key rewritten. In step 603, the maximum number of levels of rewriting rules stored is assigned to N. In the example shown in FIG. 9, the maximum number is 4, so 4 is substituted for N. At step 604, 1 is assigned to level counter I. End determination step 605
After that, the process proceeds to step 701 in FIG.

In step 701, KEY (1-1) is changed to KEY (1).
Assign to . Since I is 1 at the moment, KEY (0)
In other words, "Multi-window system" is KEY (1
). At step 702, the subroutine REWR
Call ITE KEY, that is, transfer control to the flow shown in FIG.

FIG. 8 shows search key rewriting processing based on rewriting rules. At step 801, the first one of the stored rewriting rules is assigned to RULE. At this moment, ".→(null)J" shown in FIG. 9 901 is substituted.

In step 802.

Determine whether the level of RULE is equal to ■.

As shown at 901 in FIG. 9, the current level is 1°, so the process advances to step 803. At step 803,
Apply RU L E to KEY (1).

At present, KEY (1) changes from "Multi-Window System" to r-Multi-Window System J. As a result, the length of KEY (1) has changed, so at the same time,
Modify the contents of 0RIGLOG. 0RIGL after modification
The contents of the OG are shown in FIG. In Figure 11, 1104
．． As shown at 1110, there was a portion corresponding to Nakaguro, but in FIG. 12, that portion has been deleted and packed in. Thereafter, the process proceeds to step 805 via end determination step 804 . In step 805, the next rule after the rule assigned to the current RULE is newly assigned to RULE. At this point, the next rule 902 "Duyush" in FIG. 9 901 is substituted. After that, the process advances to step 802. Since the level of the rule 802 assigned to RULE at this moment is 2, it does not match 1 in ■, and the process proceeds to step 804. Hereinafter, each rule of FIG. 9 903 to 908 will be explained in order
It is assigned to LE, but in this example, not all levels match 1, so it is not applied. After the rule 908 is assigned to RULE, the end determination in step 804 is established, and control is returned to step 703 in FIG.

At present, as a result of the processing shown in FIG. 8, "Multi-window system" is assigned to KEY (1). In step 703, it is checked whether the search key has been rewritten or not. At this point, KEY (0) and KEY (1) are different, so the process advances to step 704. In step 704, the KEY
(1) That is, at present KEY (1) A dictionary search is performed using the longest match using "r multi-window system" as a search key.

When data is obtained through a search, the data is added to the search result stack. However, the contents of the dictionary are as shown in FIG. 10, and there is no heading that matches KEY (1). That is, the dictionary search fails and nothing is added to the search result stack. In step 705, 1 is assigned to a variable REWRITTEN indicating whether the key has been rewritten.

Next, in step 606, 1 is added to I. At the moment I
becomes 2. Thereafter, the processes shown in FIGS. 7 and 8 are performed again at step 605. That is, since I has become 2, in the process of FIG.

Application of rules 902, 903, 904, 905 (7) in FIG. 9 is attempted. Here, rule 905 is successfully applied. As a result, KEY (2) changes from "multi-window system" to "multi-window system." A dictionary search is performed using this KEY (2) as a search key, but this dictionary search also fails like the previous time. Thereafter, the process proceeds to step 606 via step 705.

In step 606, 1 is further added to ■. This result I is 3
Then, substantially the same processing as above is performed. Since ■ is currently 3, rules 906 and 907 in FIG. 9 are attempted for KEY (3).

Rule 907 is applied successfully. As a result, KEY(3)
changes from "multi-window system" to "multi-window system." As a result of a dictionary search using this KEY (3) as a search key, "multi-window" shown in record 1001 in FIG. 10 matches. Therefore, record 1001 is added to the search results stack. Since the length of the heading "Multi-Window" of this record 1001 is 8, the value "9" of 0RIGLOC(8) shown in FIG. 12 1208 is written in the "Correspondence" item of the search result stack. In other words, the last letter of "Multi-window" is KEY
(0) indicates that it corresponds to the 9th character of [Multi-Window System]. As a result, the contents of the search result stack become as shown in FIG.

Next, in step 705, 1 is assigned to a variable REWRITTEN indicating whether or not the search key has been rewritten.

After this, 1 is further added to I in step 606.

At this moment, I is 4. Therefore, application of the rule 908 in FIG. 9 is attempted but fails. As a result, KEY (
4) is the same as KEY (3), "Multi-window system J" remains. Therefore, the condition does not hold in step 703 of FIG. 7, which returns from the process of FIG. 8. Therefore, step 704 Step 606 without going through 705
Proceed to.

In step 606, 1 is further added to summer. As a result,
■ becomes 5, the termination judgment condition of step 605 is satisfied, and step 30 of FIG. 3, which is the calling source of the process of FIG.
7, and proceed to step 401 in FIG.

In step 401, it is checked whether the contents of the search result stack are empty. 6. If the contents are empty, the process proceeds to step 403 via step 402. If the key has been rewritten, a warning is displayed in step 404. after that,
Proceed to step 405 and process the rest of the text in the same way. At this moment, in step 401, the contents of the search result stack are not empty, as shown in FIG. Therefore, the process advances to step 501 in FIG.

In step 501, one notation to be used is selected from the search result stack. At present, the search result stack is as shown in FIG. 13, so select "multi-window" as the notation to be used and substitute it for 5TRI. In step 502.

Since the "correspondence" of the search result staff is "9", K
The first 9 characters of EY (0), that is, “Multi-
window" to 5TR2.

If 5TRI and 5TR2 are different, proceed to step 504. In step 504, 5T in the text
Modify the notation that matches R2 to 5TRI notation. At this time, "Multi-window" will be modified to "Multi-window".

Thereafter, the rest of the C text is processed via step 505. In other words, the text that was originally "Multi-Window System..." was modified to "Multi-Window System..." and then changed to "Multi-Window System...".
Processing continues with the first character "・" as the character of interest.

By repeating the process as described above, the expression "multi-window system" in the text is corrected to the expression "multi-window system."

In this embodiment, rewriting and searching are continued regardless of the success or failure of the dictionary search corresponding to each stage of rewriting, but it is also possible to proceed to the next process immediately if the search is successful.

In addition, rewriting and searching are repeated step by step, but if there is sufficient consistency between dictionary registration and rewriting rules, all rewriting rules can be applied to one search key at the same time, and dictionary search can be performed. can be completed once.

In the implementation means and embodiments described above, the dictionary search method is used in the proofreading support method, but this dictionary search method can also be used in other natural i' word interpretation methods. For example, when using the @machine translation method,
Window system” “Multi-window system”
[Multi-window system] is searched in the dictionary with the notation "Multi-window system" in the colon, so the last word "Multi-window system" is included in the dictionary for machine translation.
By registering the translated word as a heading, it is possible to search from the other three types of notation.

In addition to the implementation means and embodiments described above, the following implementation means are also conceivable.

For the first purpose of detecting discrepancies in orthography between different words, in a proofreading support method that uses a text storage means, a text editing means, a display means, and an input means, a spelling deviation information storage means is used. This is achieved by providing a step of using record storage means for sorting and sorting means.

In the orthographic variation information storage means, orthography is classified and stored in advance for each syllable having the same source. The text editing means compares the text stored in the text storage means with the notation stored in the notation variation information storage means based on instructions from the input means. A sorting record is created for the matched portion and stored in a sorting record storage means. When the comparison process is completed, the records stored in the sorting record storage means are sorted by the sorting means. The text editing means displays the sorting results via the display means.

An example of the above configuration will be described below.

FIG. 14 is a diagram showing an example of a device that implements the present invention. FIG. 15 is a flow diagram showing an example of the calibration support method according to the present invention. Text editing means 14 shown in FIG.
01 operates according to the flow shown in FIG. FIG. 16 is a diagram showing the contents of the notation variation information stored in the storage means 1405 of FIG. 14, and as shown in the figure, notations having the same source are classified according to item numbers.

The following description will be made according to the flow shown in FIG. 15. In step 1501 in FIG. 15, the contents of the sorting record storage means 1406 in FIG. 14 are emptied.

In step 15o2, the text and each notation of the notation variation information are compared, and a sorting record is created for the matched portion, and is stored in the sorting record storage means 1406.

An example of sorting records is shown in FIG. For example, assume that there is an expression "...activating data..." in the fifth line of the 20th page of the text. Since the "Tig" part of this matches the "Tig" shown in FIG. 16 1604, a record shown in FIG. 17 1701 is created. In other words, "4" in the "item number" of record 1701 is the first
The "item number" of record 1604 in Figure 6 is "4". Furthermore, “Tig” in the “notation” of record 1701 is the notation that matches in the comparison process of step 1502. Also,
"Qua..." in "Forward text" of record 1701
is a reverse order of the character string ``Aku'' that precedes ``Tig'' in ``...Actig data...'' which was in the fifth line of page 20 of the text. Also, 17
Similarly, the "data" of the "rear text J" of 01 is a character string in normal order after "chii".

Also, the "20p" of "Text position" of record 1701 is
, 5Q'' indicates that the content of the record 1701 is related to the notation in the 5th line of the 20th page of the text.

In addition, as another example, on page 22, line 29, “...
If there is an expression "cognitive science...",
A record as shown in record 1702 in FIG. 17 is created.

After such comparison and record creation processing, the 15th
At step 1503, records are sorted. When sorting, use the "item number" in Figure 17 as the first key and the "notation"
is the second key, "forward text" is the third key, "backward text" is the fourth key, and "text position j is the fifth key. By setting the sort keys like this, you can sort Then, the records (1701 and 1702) regarding variations in the spellings of ``Active Data'' and ``Tig'' in Cognitive Science J, which are completely different words located far apart, will be close to each other.

This sorting result is displayed at step 1504 in FIG. 15 via the display means (1403 in FIG. 14).

Two types of means and embodiments have been described above. In the above explanation, the fluctuating notation was written in katakana, but it is also possible to improve the performance of pattern matching by expressing it in hiragana, romaji, regular expressions, and state transition diagrams. can. It can also be spelled in other languages. For example, misspellings between English and American can be considered as variations in one spelling and can be detected or automatically corrected.

〔Effect of the invention〕

According to the first implementation means of the present invention, there is an effect that it is possible to automatically unify the variations in notation that become a problem when a plurality of documents are generated. Furthermore, if there are enough rewriting rules to be used, in order to create a dictionary, it is sufficient to provide the notation to be used.

That is, it is not necessary to investigate possible spelling variations for each word and register all spelling variations. Also, it is sufficient to give the notation to be used for each component of the compound word. This has the effect of significantly reducing the effort required to create a dictionary. Also 1
By combining the dictionary search method of the present invention with other natural β word analysis methods, it is possible to deal with variations in the notation of the input text correctly without having to register the variations in the notation comprehensively. .

According to the second implementation means of the present invention, it is possible to easily realize a display in which variations in spelling between different words can be easily discovered.

[Brief explanation of the drawing]

FIG. 1 is a functional configuration block diagram representing a first embodiment of the present invention, and FIGS. 2, 3, 4, and 5 are processing flows of a calibration support method according to the embodiment of the present invention. 6, 7, and 8 are processing flow diagrams of the dictionary search method according to the embodiment of the present invention, and FIG. 9 is an explanatory diagram showing an example of rewriting rules used in the dictionary search method, Fig. 10 is an explanatory diagram showing an example of the structure of a dictionary; Figs. 11 and 12 are explanatory diagrams showing a data structure showing the correspondence of each character when the character string length changes due to the application of rewriting rules; FIG. 14 is an explanatory diagram of the structure of a search result stack that stores dictionary search results. FIG. 14 is a functional configuration block diagram representing a second embodiment of the present invention.
FIG. 15 is a processing flow diagram of the proofreading support method of the second embodiment, and FIG. 16 shows the steps used in the proofreading support method of the second embodiment.
FIG. 17 is an explanatory diagram showing the structure of the notation deviation information, and FIG. 17 is an explanatory diagram showing the structure of the sorting record used in the above-mentioned proofreading support method.

Claims (1)

[Scope of Claims] 1. A proofreading support method characterized in that character strings are rewritten using a pre-stored correspondence table of notation deviations as a rewriting rule. 2. A proofreading support method that rewrites character strings by using a pre-stored correspondence table of notation variations as a rule for rewriting from non-standard notation to standard notation. 3. The proofreading support system according to claim 1 or 2, characterized in that a desired character string in katakana notation is rewritten into another katakana notation based on the correspondence table. 4. Claim 1 or 2, characterized in that in rewriting where the length of the character string before rewriting is different from the length of the character string after rewriting, the correspondence of each character between character strings is preserved. Calibration support method described in Section 3. 5. Rewrite the character string that serves as the search key for the character string to be searched by hand using a previously stored correspondence table of spelling variations as a rewriting rule, and search the dictionary based on the rewritten search key. A dictionary search method that is characterized by performing the following. 6. Claim characterized in that the rewriting of the search key is carried out in stages according to rewriting rules divided into several stages, and a dictionary search is performed based on the search key rewritten in each stage. Dictionary search method described in Section 5. 7. A proofreading support method, characterized in that the dictionary search method according to claim 5 or 6 is used to perform common rewriting for common spelling variations between different words. 8. A proofreading support method characterized by detecting and displaying notation deviations in text by comparing the notation of a text character string with a correspondence table of notation deviations stored in advance.