JP2001092830A - Device and method for collating character string - Google Patents

Abstract

PROBLEM TO BE SOLVED: To collate a character string on the basis of an element larger than a character or term by realizing comparison on sentences composed of terms or documents composed of sentences. SOLUTION: Character string groups constituting a document are arranged in the form of a table and, after coincident or resemble character strings are adjoined to each other by sorting the character strings, detailed verification is performed. Then more detailed results are obtained by repeating a procedure for changing the dividing size or order of the character strings.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an apparatus and a method for collating a plurality of character strings of a document converted into an electronic file by a character code.

[0002]

2. Description of the Related Art When collating a document or a sub-element of a document, a character string obtained after setting a retrieval condition based on a keyword or a key sentence and performing an automatic retrieval is collated. The application of natural language analysis for setting and searching has been implemented.

[0003]

However, there is a case where the selection of the key and the setting of the search condition cannot be properly performed in some cases, and there is a problem that the number of objects to be collated becomes excessive and the collation becomes impossible.

[0004] It is an object of the present invention to provide a collation method that does not require a search in order to solve the problems of selecting keywords and setting search conditions.

[0005]

According to a first aspect of the present invention, there is provided a collating apparatus comprising: a single document including a character string and a group of character strings; When determining the mutual agreement or the difference, the document is input from the input document storage device, and the character string characteristic table generating means extracts the character string from the component and determines the position of the component in the document. By including a document identifier such as the document name together with the identifier and a relation expression code between the constituent element and the document, an input document character string characteristic table expressing the character string characteristics is generated, and the character string sorting means The document character string characteristic table is sorted by character strings to create an output document character string characteristic table, and the document restoring means generates a document from the output document character string characteristic table by adding a matching expression to the document format. And outputs whether or output document string property table output to the output document storage device and to the output document storage device.

According to the above means, since a document stored in the information processing apparatus is digitized by a character string and a character code of a character string group, a sorting method is performed on the character string and the character code group of the character string group. Is applied, and as a result, the matching relationship between adjacent character strings and character string groups is determined, and a document identifier such as the document name and the constituent element are added to the character string and the character string group together with the position identifier in the document. After performing a general table sort by making the table format accompanied by information such as the expression code related to the document, add a matching expression to the document format based on the accompanying information. It is characterized by outputting or outputting the sort result itself.

A collating apparatus according to a second aspect is a method for improving the output format of the collation result by adding processing means for converting the document into a format conforming to the rules for specifying a matching relationship expression and outputting the same to the document restoring means according to the first aspect. I will provide a. Regarding the setting of the rules of the present invention and the use of the set rules, it is possible to incorporate the processing conforming to the rules into the implementing means, and to make the setting, storing, and using the rule processing means independent. It is possible to use general methods that are possible and widespread.

According to a third aspect of the present invention, in order to determine not only a matching relationship but also a similar or conflicting relationship, the collating apparatus according to the input document character string characteristic table based on the similar conflicting term database is used. Is added to the similar character string characteristic table, and the document processing / reconstructing means refers to the similar contradiction term database and conforms to the similar relation expression format specification rules. An output document is generated and output.

The above means replaces a character string with a reference character string based on a similar contradiction term database based on a matching relation collation method, and then sorts the table using the original character string or the like as accompanying information. Apply to provide similarity matching methods.

According to a fourth aspect of the present invention, the above-described means is applied to collate a plurality of documents, the number of the documents is input, the input document character string characteristic table of the number of the documents is generated, and After converting to the similar character string characteristic table of the number of documents, a characteristic table combining means for converting the similar character string characteristic table of the number of documents into a single combined characteristic table is added. Are sorted by a character string to create a single output document character string characteristic table, and the document processing / restoring unit generates a single or the number of output documents from the single output document character string characteristic table to output the document. The data is output to a storage device.

[0011] The means provides a method for collating a plurality of documents by using a single table associated with the identifiers of the plurality of documents to collate the plurality of documents.

According to a fifth aspect of the present invention, a means for changing a character string in conformity with a document characterizing rule is added to the character string characteristic table generating means, and only the character string is converted for similarity matching. In addition, there is provided a method for improving by changing the structure of a character string such as an order. In other words, if a document stringing rule that changes the arrangement order based on the terms that make up a character string is set, the order of the strings is changed in terms of terms, so the position is moved forward by the stringing process When the term of the document is prioritized and adjacent in the sort result, and if the document string conversion rule of reversing the character order is set, the characters of the character string are changed in reverse order and the sorting is executed and terminated. Output an output document character string characteristic table in which similar character strings are adjacent.

According to a sixth aspect of the present invention, in order to further improve the result of the collation by the above-mentioned means, an output document character string characteristic generated by the character string sorting means by adding an adjacent character string correlation rule extracting means. The adjacent character strings in the table are compared with each other to extract the adjacent character string mutual relation rules in accordance with the extraction rule of the adjacent character string mutual relation rules. When extracted, the collation procedure is continued within the limited number of repetitions while changing the setting of the document character string conversion rule.

In order to improve the matching performance, the above means
The adjacent character strings obtained by sorting are compared with each other to extract the rule of the mutual relationship between adjacent character strings according to the extraction rule of the adjacent character string correlation rule, and the adjacent character string correlation rule is written in a document. A method is provided in which the collation procedure is repeated by additionally setting a string conversion rule.

[0015]

Embodiments of the present invention will be described with reference to the principle diagrams. In FIG. 1 corresponding to claim 1, in order to use a general tabular sorting method, the character string characteristic table generating means 12 includes a character string and a single character code of a character string group. The input document is read from the input document storage device 11 which stores the file of the document, a character string is extracted by a method suitable for the collation purpose, and the character string is sorted by a predetermined sorting criterion in the input document character string characteristic table 13. At the same time as stored in the location, according to the format of the character string characteristics table, the identifier of the position of the character string in the document, a document identifier such as the document name, or a relation expression code between the character string as a component of the document and the entire document For example, the input document character string characteristic table is generated by storing information expressing the characteristics of the character string, such as the associated character string, and the character string sorting unit 14 compares the input document character string characteristic table with the character string to be compared. A result in which the relation between the accompanying information and the reference character string is stored by executing the sorting in the table format is obtained, and the result is stored as the output document character string characteristic table 15. In the case of a request for expressing a matching relationship in a format, the sort result is returned to the input document format, a document to which the expression of the matching relationship is added is generated, output to the output document storage device 17, and matches as the sorting result. When a table in which character strings are adjacent is desired, an output document character string characteristic table is output to the output document storage device.

In FIG. 2 corresponding to claim 2, when there is a request for the expression form of the matching relation, the request is expressed as a matching relation expression form specification rule 18, and the character strings that match by sorting are adjacent. By paying attention to this, the matching of the character strings is determined, and the output document character string characteristic table 15 of the sorting result is modified according to the request to generate a document to which the expression of the matching of the request is added, and the output document storage device 17 is added to the document restoring means 16 and the document processing / restoring means 19 is added.
To improve the output format. It is also possible to add the function of the document processing / restoring unit without expressing the demand for the expression format of the matching relationship as the matching expression format designation rule.

In FIG. 3 corresponding to claim 3, in general, when character strings such as a sentence or a paragraph which are constituent elements of a document completely match, it is necessary to determine a similarity or a contradiction. Therefore, instead of using the input document character string characteristic table 13 by the character string sorting unit 14 as shown in FIG. The similar conflicting relation setting means 21 converts the input document character string characteristic table into the similar character string characteristic table 23 based on the similar conflicting term database 22, sorts the similar character string characteristic table, and sorts the output document character string characteristic table. The document processing / restoring unit shown in FIG. 2 which generates the output document 15 from the output document character string characteristic table obtained by sorting is referred to the similar contradiction term database and conforms to the similarity relation expression format designation rule 24. Has been changed to a document processing restoring means 25 for adding the generated and function of outputting the output document Te. It should be noted that the similarity relation expression form specification rule also includes the coincidence relation expression form specification rule and the contradiction relation expression form specification rule, thereby realizing matching, similarity, and contradiction. Further, by including the input document character string before conversion by the similar reciprocal relation setting unit in the similar character string characteristic table, it is possible to conceal the similar reciprocal relation and output only the input document character string.

FIG. 4 corresponding to claim 4 shows the principle of a means for comparing a plurality of documents using a single document matching means. The character string characteristic table generating means 32 is different from the character string characteristic table generating means 12 in that it processes a plurality of documents and stores the number of documents corresponding to the input document. 33, and the similarity reciprocal relation setting unit 34 differs from the similar rebellion relation setting unit 21 in that the input document character string characteristic table of the document number is based on the similarity reciprocal term database 35 applicable to the document number of the document. Then, the characteristic table combining means 37 converts the similar character string characteristic table of the document number into a single combined characteristic table 38, and the character string sorting means 39 converts the similar character string characteristic table of the document number into a single combined characteristic table 38. The entirety of the combined property table that has become one is sorted by the character string sorting means 14.
A single output document character string characteristic table 40 is created by sorting by a character string in the same manner as described above. In accordance with the relation expression format specification rule 41, a single or the number of output documents can be generated using the identifiers of a plurality of the documents associated with the property table and the number of the documents can be stored. Output to the output document storage device 43.

In FIG. 5 corresponding to claim 5, in order to realize a collation that does not depend on the difference in the order of the character string elements and the like, the character string characteristic table generating means 52 uses an input document. A means for changing the character string of the input document input from the storage device 31 in accordance with the document character string conversion rule 51 to generate the input document character string characteristic table 33 is added.

In FIG. 6 corresponding to claim 6, an output document character string characteristic table 40 created by the character string sorting means 39 in order to further improve the executed collation result.
The adjacent character string correlation rule extraction unit 53 compares the adjacent character strings in the output document character string characteristic table created by the character string sorting unit with the adjacent character string correlation rule. The adjacent character string mutual relation rule 54 is extracted in accordance with the extraction rule, and the character string conversion rule setting means 55 changes the setting of the document character string conversion rule 51 when the adjacent character string mutual relation rule is extracted. The input document character string characteristic table 33 generated by the character string characteristic table generation means 52 is changed by continuing the collation within the limited number of repetition times of the collation procedure, and is changed to the output document character string characteristic table of the next sorting result. Repeat the procedure of reflecting.

An embodiment of the document processing according to the present invention will be described with reference to the drawings using the specification of the present invention as an example. FIG.
When the input document is single, the character string characteristic table generating means 12 must match the format of the document and extract a character string which is a component of the document so as to meet the purpose of collation. The input document is read from the device 11,
Generates a character string according to the file format of the document.

For example, if the same part is distinguished from the different part by collation to clarify the mutual relationship between the twelve claims in the specification, each claim is described in one paragraph. Since there is no possibility of matching if collated in paragraph units, claim 1 is divided by reading points as shown in FIG. 7, and a character string is generated by dividing claim 1 into a total of seven document components. I do. In the example of the input document character string characteristic table 13 in FIG. 8, the number of rows in the table is the total number of divisions, the predetermined sorting reference point is the second column, and the character strings divided by the reading point are numbers (the numbering method is arbitrary). Is assigned here as a serial number starting from 101), and the identifier of the position in the document in the numerical format stored in the first column is used. It is also effective to express the relationship between the sentence element represented by the character string and the entire document by adding “claims” which is a column name of a framework wider than the claim number. A general document has a table of contents such as a chapter or a section, which is used as a word processor control code, and can be used. In addition,
The document name may be, for example, the “character string collation apparatus and method”, which is the name of the present invention. However, if the file name in the input document storage apparatus is used, it is usually simplified.

Even when the input document is a drawing as shown in FIG. 1 of the present case, it is possible to extract a character string from the digitized file and generate a characteristic table as shown in FIG. it can. If the input document is in the form of a table, the means of the present invention can be specified by copying the character string portion of the table into the input document character string characteristic table. This is possible by setting a plurality of keys by means.

FIG. 10 is a flowchart showing an example of the procedure of the character string characteristic table generation means 12 shown in FIG. 1. A document character string group division rule for dividing a document at a reading point is set (S1), so that a document is converted to a character. It is possible to determine the total number of character strings by dividing the character string into column groups (S2). Subsequently, a simple rule such as a parsing rule of a natural language or a method of extracting terms separated by kana and kanji is set as a character string group term extraction rule for extracting terms from a character string group (S3).
Then, it is possible to use the character string characteristic table generating means 12 for similarity search or the like, but it may be omitted as long as the collation of character strings is simply collated.

From S5 in FIG. 10, in the procedure for generating the input document character string characteristic table 13 from the extraction result, the character string number is initialized (S4), the character string is started (S5), and the divided document is obtained. Is converted into a character string (S6), and if necessary, supplementary data of the character string is added together with a character string, a document identifier, or a character string position identifier, and stored in the character string number row of the characteristic table ( S7), and if the character string number has reached the total number of character strings, the process is terminated (S8), but if not, the number is updated (S8).
9) Then repeat steps S6 and S7. If the environment at the time of executing this procedure is suitable for processing the division of the document and the conversion of the character string in parallel, the total number of character strings is set as shown in FIG. Instead of repeating the processing, the procedure may be started from the beginning of the document, divided into document elements while repeating the procedures from S4 to S9, and the procedure may be terminated when the document ends.

In FIG. 1, when the character string characteristic table generating means 12 generates the input document character string characteristic table 13 as described above, the character string is digitized numerically by the character code. In this case, the sorting method for the table is widespread. Therefore, the character string sorting unit 14 sets the entire input document character string characteristic table as a matching target, and determines a character string to be matched. Use a table column that contains a string to be matched as a key to a regular tabular sorting method. Since the accompanying information and the reference character string are stored in the same line as shown in FIGS. 8 and 9, the result of storing the relation between the reference character string and the accompanying information by performing a table sort is obtained. Then, it is set as the output document character string characteristic table 15.

FIG. 11 is a flowchart showing an example of the procedure of the document restoring means 16 shown in FIG. 1. The sentence of claim 1 is processed for the sentence of claim 2 in the same manner as the characteristic table shown in FIG. After the characteristic table of the two claims is collectively implemented and the means of the present invention is implemented, for example, a character string group document restoration rule of “express a matching relationship in the original document format” is set (S11). 12 is generated. In this case, for example, it corresponds to the setting of the adjacent character string group collation rule of “determine only a perfect match” (S12), and the procedure proceeds to the following procedure.

Since the number of lines in the output document character string characteristics table 15 of FIG. 1 is the same as the number of lines in the input document character string characteristics table 13, it is requested that the input document format be used to express a matching relationship. As an example, S4, S5, S5 in FIG.
8, S13, S14, S17, and S in FIG.
In step 18, an output document is generated from the output document character string characteristic table as the sorting result. 2. A character string number is initialized (S13), document restoration is started (S14), and the character string is collated with an adjacent character string group in step S15 according to an adjacent character string group collation rule. Is determined to be the same as the second element of claim 2 and so on, the collation result is added in step S16, and the character string group document is restored based on the character string position identifier and the supplementary data. When the character string is restored to a document, each line is as shown in FIG. 12. If the character string number has reached the total number of character strings, the processing ends (S17). If the character string number does not reach the number, the number is updated (S18) and the processing of S15 and S16 is repeated. Repeat steps.

When a table in which matching character strings are adjacent to each other is desired as shown in the sorting result, an output document character string characteristic table as shown in FIG.

For example, when there is a request for the expression format of the matching relationship, "the same character string is placed on the same line and the different character string is placed on a different line between the claims", the request is expressed by the matching relationship expression shown in FIG. It can be used as the format specification rule 18 to generate an output example such as the document in the correlation table format of FIG.

In the case where the same collation as that of the above claims 1 and 2 is performed for all the claims in this specification, there are a plurality of matching character strings. The rows of the column will be adjacent to each other in a plurality of rows.
From this result, it is easy to count the number of matching lines, and by expressing it in the output document, the matching relationship can be clarified. If the expression of the number of matching is requested, the matching is performed according to the request. The output document is improved by generating and outputting a document in which the number of character strings is added to, for example, the left end of the table in the correlation format.

Since the present specification is not an example for explaining the process of the similar contradiction in FIG. 3, it is assumed that the similarity between the "device for controlling the temperature of hot water" and the "temperature controller for hot water" is determined. Considered as an example, in the similar term column of hot water relation in the similar contradiction term database 22, “hot water” and “hot water” are defined as being similar and based on “hot water”, and “control” and “adjustment” are similar. Defined based on "control", defined as "device" and "vessel" are similar and defined based on "device", for example, "ignore all kana in similarity matching" The conflicting relation setting means 21 generates a "hot water temperature control device" of the reference character string for the "device for controlling the temperature of hot water" and the "temperature controller for hot water", A device that controls the temperature of hot water and a hot water temperature controller Is stored, and the "hot water temperature control device" is stored in another column of each row as a reference column, and a code indicating that similar conversion has been performed in another column as supplementary information is stored. A similar character string characteristic table 23 is generated, and a reference column is set as a key, and a table sort is performed.
"Device for controlling hot water temperature" and "Hot water temperature controller"
Get the result that the two rows of are adjacent.

That is, the reciprocal relation setting means 2 shown in FIG.
FIG. 15 is a flowchart of an example of the first procedure, and the similar conflicting matching rule such as “Ignore all pseudonyms in similarity matching” or, for example, “apply a similar term field related to hot water” is set (S21). Then, the newly created similar character string characteristic table is initialized (S22), and the line number of the characteristic table is adopted as the character string number to be processed in the input document character string characteristic table and the similar character string characteristic table. Is initialized (S23),
The conversion process for each character string from the input document character string characteristic table to the similar character string characteristic table is started (S24). According to this procedure, the character string to be converted in the input document character string characteristic table is selected, and the line of the input document character string characteristic table corresponding to the character string is transferred to the similar character string characteristic table (S25), and the character string is converted. Is decomposed into terms by the common term extraction database structure and the common term extraction method, a term number is assigned for term processing of the character string, and the term number is initialized (S26).
Then, the conversion process for each term in the character string is started (S27). If it is determined in step S28 that the term is included in the similar contradiction term database, a new character string is generated in accordance with the similar contradiction matching rule, and the new character string is stored in the similar character string property table. Add to a column different from the column (S2
9) Store and determine whether the term in the character string is finished (S30), and if not finished, update the term number (S3)
1) and return to step S28. If it is determined in step S30 that the term in the character string has been completed, it is determined whether the character string in the characteristic table is completed (S32). If not, the character string number is updated (S33). Then, the process returns to step S28.

Similarly, in a document in which, for example, "power on" and "power off" are described, for example, the document element opposite to "power on the device A" is collated. In the conflicting term column of the similar conflicting term database, "input" and "disconnection" are defined as conflicting and based on "input", and the similar conflicting relation setting means, for example, "turns off the power of the apparatus A." Is converted to "turn on the power of the device A". In this processing, columns such as “turn on the power of the device A” and “turn off the power of the device A” which are originally included in the input document character string characteristic table are stored, and another column of each row is stored. Is stored as a reference column, and “Apparatus A is turned on” as supplementary information.
By storing a code indicating that contraversion was performed in another column of the row of “Power off the device” and executing a table-type sort by setting the reference column as a key, The result is that the two rows of “turn on the power of the device A” and “turn off the power of the device A” are adjacent.

The similar character string characteristics table generated in this way is sorted in a table format to create an output document character string characteristics table 15, and the similarity relation is also determined from the output document character string characteristics table obtained by sorting. In order to generate an output document to be represented, the document processing / restoring unit 25 in FIG. 3 outputs, for example, “a character string converted in a similar reciprocal relation together with a character string of an input document”.
An output document is generated in accordance with the similarity expression format specification rule 24 such as whether or not. In addition, it is also possible to store, in the similarity contradiction term database, supplementary information such as the similarity relation and the contradiction term definition together with the similarity relation priority order, and reflect the supplementary information in the output document.

A plurality of documents can be collated according to the principle of FIG. 4 by using the above-mentioned single document collating means.
The processing function of the character string property table generating means 32 is different from the processing function of the character string property table generating means 12 in that the character string property table generating means 12 processes a single input document whereas the character string property table generating means 12 processes a single input document. The generation means 32 processes a plurality of input documents. Since a matching character string or a similar character string exists in a single input document, the principle is that a plurality of input documents are similar to the processing of a single input having many components.
Accordingly, the input document character string characteristic table 13 and the similar character string characteristic table 23 shown in FIG. 3 differ only in the number of input documents in FIG. 4, so the character string sorting means 39 in FIG. In order to make it the same as the sorting method of the third character string sorting unit 14, the characteristic table combining unit 37 of FIG. 4 changes the similar character string characteristic table of the number of input documents into a single combining characteristic table 38. However, when it is necessary to simplify the processing of the similar character string characteristic table and the combination characteristic table due to the large number of input documents, the input document character string characteristic table 33 is stored and the document processing / restoring means is stored. Used at 42.

FIG. 16 is a flowchart showing an example of the procedure of the property table combining means 37 shown in FIG. 4. The newly created combined property table is initialized (S41), and the document number of the input document to be processed is initialized (S42). Starts a process of generating a single decision characteristic table by combining similar character string characteristic tables of the number of input documents (S
43). Since the similar character string characteristic table of the input document to be combined is selected by this procedure, the similar character string characteristic table is read (S44), and a position identifier indicating the relationship between the similar character string characteristic table and the input document is added. (S45),
It is determined whether or not the processing of the total number of input documents has been completed (S47), and if not completed, the document number is updated (S48) and the process returns to the step S44. When it is determined in step S47 that the processing of all similar character string characteristics has been completed, the combining processing ends.

The character string sorting means 39 shown in FIG. 4 sorts the entire combined property table by a reference character string in the same manner as the character string sorting means 14 shown in FIG. A table 40 is created, and the document processing / restoring unit 42 determines whether or not the above-mentioned “the character string converted in a similar conflict is also output together with the character string of the input document” which is commonly requested for the input documents of the number of input documents. Is used as the similarity relation expression format specification rule 41, and the identifier of each of a plurality of documents associated with the property table is used in accordance with the rule 41 to determine the number of single or input documents from a single output document character string property table. An output document is generated and output to the output document storage device 43. Note that even if there are a plurality of input documents, the output of a single document in the form of a correlation table as shown in FIG. 14 is effective, so that the function of the document processing / restoring means 42 of FIG. The same may be applied.

Since various documents have their own subtle characteristics, there are cases where the order of terms in a sentence is different even though they have the same or similar contents. A situation arises in which it is desired to ignore such various expressions in matching because there is no need to discriminate. When a reference character string is generated in this situation, for example, a document character string conversion rule 51 such as “specify the order of terms in the character string” is set as shown in FIG. By changing the character string of the input document input from the input document storage device 31 in accordance with the document character string conversion rule and generating the input document character string characteristic table 33, Even if it is an expression, the sorted result can be adjacent. Note that if it is difficult to set “define the order of terms in a character string”, the terms can also be sorted by character code. This objective can be achieved simply by applying the sorting method used in the present case to a single character string by replacing the rule of "standardizing the order of terms with a coding system". Also, by setting a document character string conversion rule that reverses the order of all characters in a character string, a character string having a similar end portion can be extracted.

After obtaining the result of collating the document element character strings as described above, the result is re-determined, and the process of dividing the document into document elements and changing the character string is performed. May be further adapted for matching purposes. Taking the collation of claim 1, claim 2 and claim 3 as an example, the last character string of claim 1 and claim 2 "a collation device for matching the character strings of constituent elements of a single document based on sorting"
It is determined that the character string “the collation device for the similarity of the component character strings of the single document based on the sort” is different in the division based on the reading point, but the part before the character string “sort Since the components of a single document based on a match up to "", they are adjacent in the sorted result. For example, if the adjacent characters match at least 20 characters from the front, the character string is replaced with the new document component. Extract as
Accordingly, it can be determined that "the character string of the constituent element of the single document based on the sort" matches in claim 1, claim 2, and claim 3, and claim 1 and claim 2 are "matching apparatus for matching relation".
Can be obtained, but only claim 3 is different from the "similarity matching device".
The collation result becomes more detailed.

In order to make the collation result more detailed,
In FIG. 6, adjacent character strings in the output document character string characteristic table 40 created by the character string sorting unit 39 are compared with each other to extract the adjacent character string mutual relation rule 54 in accordance with the extraction rule of the adjacent character string mutual relation rule. Equipped with extraction means 53 for the adjacent character string correlation rule,
If more than characters match, extract the character string as a new document component "and set the extraction rule of the adjacent character string correlation rule to the above" Single component character string of single document based on sorting " Extract the matched string.

In order to perform more detailed collation than the previous collation result using the character string extracted in this way, the output document character string characteristic table 32 of FIG. 4 is converted into the set document character string of FIG. By replacing the character string characteristic table generating means 52 that changes the character string in accordance with the rule 51, for example, the rule that “if there is a registered character string, extract it from the document and extract it” as the document character string conversion rule By setting and setting the extracted character string “of a component character string of a single document based on sorting” as its registered character string, the following collation procedure is performed in the next collation procedure. The character string "Single-based single-document component character string matching relationship matching device" is "sort-based single-document component string."
And a "matching device for a matching relationship", and the last character string "matching device for similarity of components of a single document based on sorting" in Claim 3 is "a component of a single document based on sorting." Since it is divided into "character string" and "similarity collation device" and stored in the input document character string characteristic table, the character string "" which matches claim 1, claim 2, and claim 3 in the sorted result It is possible to output a collation result indicating that there is a component character string of a single document based on sorting.

In the repetition of the collation procedure, when changing the character string conversion by the character string conversion rule setting means 55, the repetition is continued because the adjacent character string correlation rule is newly extracted and the number of repetitions is limited. By setting a constraint of limiting to a number, unintended excessive repetition is avoided.

In the case where claims 1 and 2 are collated, the second character string divided by the reading point of claim 1 is used to determine whether the character strings constituting the document are mutually identical or different. The "when" is the "single" of the first character string "at the time of judging the mutual agreement or difference between the character strings that are the components of a single document" Since it matches the part excluding, the rule that "all the character strings extracted and extracted are set as registered character strings" is adopted and the above "registered character string, if any, is divided from the document. The first character string of claim 1 is defined as "in a single document composed of a character string and a character string group" by setting the rule of "extract by extracting" as a document character string conversion rule.
And the second character string is "when judging the mutual agreement or difference between the character strings which are the components of the document", and the first character string in claim 2 is "single" and the second character string. Is "when judging the mutual coincidence or difference between character strings that are constituent elements of the document". Therefore, it is determined that the second character strings in claims 1 and 2 match, and Are compared to recognize the difference.

FIG. 17 is a flowchart showing an example of the procedure of the adjacent character string correlation rule extracting means 53 of FIG. 6 for performing detailed collation. Since this example corresponds to the case where the extraction rule of the adjacent character string correlation rule is a single, the setting that the extraction rule of the adjacent character string correlation rule is used for the processing of the output document character string characteristic table 40 of the sorting result (S51), and
The line number of the characteristic table is adopted as the character string number to be processed and initialized (S52), the generated adjacent character string relation rule number is initialized (S53), and the adjacent character string mutual relation rule is extracted. Is started (S54), and the character string ends (S5).
5) It is determined whether or not, if not completed, adjacent character strings are compared with each other, and the process proceeds to a related rule extraction process.

Since the adjacent character strings are compared with each other, the next line number is set to the adjacent character string number based on the line number of the target character string (S56), and the character string of the character string number is set. And whether the character string of the adjacent character string number has a common term (S57).
If the process proceeds to step 1, it is determined whether or not the extraction rule of the adjacent character string relation rule is satisfied (S58). If the extraction rule is not satisfied, the process proceeds to step S61, and if the extraction rule is satisfied, extraction of the adjacent character string relation rule is performed. And execute the storing process.

In step S59, the adjacent character string relation rule number is updated. In step S60, the established adjacent character string relation rule is stored based on the adjacent character string relation rule number. The adjacent character string that is not satisfied is set as a new reference character string, and step S
Return to 56.

FIG. 17 is an example of a flowchart in the case where the extraction rule of the adjacent character string correlation rule is a single and the number of adjacent character strings having a common term is two or less. Numbers are assigned to rules, and when there are three or more adjacent character strings having mutual relation, numbers are assigned to the adjacent character strings so that the second becomes the initial value, and those numbers are added after step S57. , And is updated after step S58, it is determined whether or not the condition is satisfied for all extraction rules and all adjacent character strings. If none of the rules is satisfied, the process proceeds to step S61. However,
Since the determination of whether or not the adjacent character strings have a common term and the extraction rule in step S57 are closely related, the determination of the initialization, update, and end of the number and the determination in step S5 are performed.
7 and S57 are various depending on the contents of the extraction rule.

FIG. 18 shows the character string conversion rule setting means 55 of FIG.
In the flowchart of the example of the procedure, it is determined whether or not the number of collation repetitions is excessive (S71). If it is excessive, the repetition is terminated. If not, is the adjacent character string relation rule number updated (S72)? If it has not been updated, the repetition is terminated because there is no extracted character, and if it has been updated, the adjacent character string relation rule is additionally set to the document character string conversion rule (S73). Step 6 is repeated from the character string characteristic table generation means 52. According to the flow of FIG.
Even if the adjacent character string relation rule number is updated, the problem remains that if the number of repetitions is excessive, the repetition will be aborted. If the character string relation rule number is to be discontinued despite being updated, a countermeasure such as outputting an unprocessed adjacent character string relation rule may be taken.

In the above example, the rule "extract the character string as a new document component if the preceding character matches at least 20 characters in the adjacent character string" is set as the extraction rule of the adjacent character string correlation rule. Then, the character string extracted by the character string characteristic table generating means 52, which has set the above-described rule of “if there is a registered character string, and divide it from the document and extract it” as a document characterizing rule, is used as the registered character string. Thus, a means for making the generation of the input document character string characteristic table more detailed than the previous processing is realized. As an extended form of this example, means for storing a plurality of registered character strings in the form of a storage table and using a large number of registered character strings in a document character string conversion rule can be realized.
In addition, collation is performed by setting a document string conversion rule that reverses the characters, and a matching character string is extracted and set as a registered character string. It is also possible to extract a matching character string that is longer than the number of characters. In addition, all the character strings that match in the collation result are adopted as the registered character strings of the document character string conversion rules, and the characteristic table is converted into a character string as a component of the character string, and the iterative procedure of FIG. 6 is repeated. As a result, a means for extracting the same or similar character string at a plurality of positions with a certain number of characters or more is realized.

In the above description, the character string collating means based on the present invention has been described for a case where a character string larger than a term is used as a document constituent element. However, the present invention can be used even if a divided element is reduced and a term is used as an element. it can. As an example of such a case, the above-mentioned character string “a collation device for matching the character strings of the constituent elements of a single document based on sorting” in claim 1 and the above-mentioned character string “single based on sorting” in claim 3 FIG. 1 shows a collation device for a similarity relation between component character strings of one document.
If an output example such as a document in the form of a correlation table shown in FIG. 4 is desired, FIG. 19 is used. Since the terms are short and can be arranged side by side, the output can be performed as shown in FIG.

[0052]

According to the present invention, since the character string itself can be collated as described above, the following effects can be obtained.

In a single document, the content often changes slightly with the progress of the description, so that it is necessary for the writer himself to check the content, and for the reader of another person, the content is not changed. Before entering the interpretation of change, it is necessary to clarify the same and different parts of the written text itself and the terms used, and then move on to examining and understanding the differences in content. However, the burden of collating a large number of terms or sentences in a long document in detail requires a large amount of time, and if done manually, it is affected by various situations and involves negligence.
By means of the present invention, by automatically collating and obtaining an output in which a match and a disagreement and a similarity and a dissimilarity are separated, a large work load is reduced, a long time is reduced, and errors caused by manual work are eliminated. be able to.

In order to collate a document, it is necessary to extract a collation target portion. There is a conventional technique of extracting a feature by setting a keyword or extracting a sentence by parsing a sentence. Since the character string characteristic table can be generated and collated simply by dividing the input document, such a work becomes unnecessary.

This is equivalent to the conventional means for searching a document by setting a keyword or a characteristic value of a keyword. A character string corresponding to a document element or a sentence is set and collated to generate an output document in a format that is easy to use. Then, a document element or document that matches or is similar to the set document element or sentence can be extracted, so that it can be used as a search unit using the document element or sentence as a key.

When the result of collation after setting the purpose is large, the work load for organizing the result becomes large. However, according to the format of the character string characteristic table, the identifier of the position of the character string in the document or the document name such as the document name is used. It stores information that expresses the characteristics of character strings, such as the relationship between a document identifier, a character string as a component of the document, and the entire document, and generates an output document by reflecting the information. Can be reduced and the time can be shortened, and negligence caused by the work can be avoided. Also, if there is a request for the expression format of the output document, the request can be reflected in the generation of the output document by the document processing / restoring means by describing the request in the matching relation expression format specification rule. Can output the result.

In the collation of similarity relations, when a general dictionary is defined as similar, a document in a specific area in which terms are used in a discriminative manner, or conversely, a document is created in a domain of a document user. Some documents aim to collect information by judging terms discriminating in the domain of the user as similar, and the similarity conflict setting means is similar based on the similarity conflict database applicable to the input document to be collated. Since the character string is converted into the character string characteristic table 23, the similarity can be collated by utilizing the special circumstances of each area.

By using the similar contradictory term database storing terms having the opposite meaning, it is possible to set a positive sentence and extract a negative sentence opposite thereto.

In the case of collating a plurality of input documents, the property table combining means sorts the entire combined property table into a single combined property table instead of a single combined property table. Eliminate limited issues.

Since the input document character string characteristic table can be generated by changing the character strings in accordance with the document character string formation rules, the character strings have similar meanings even though the order of the character string elements is different. Can be matched without depending on the difference in order. Also, by reversing the order of all the characters in the character string, it is possible to perform a collation that emphasizes the end of a sentence unique to Japanese, such as extracting a character string having a similar end portion.

If there is a lot of common part in the long character string of the executed collation result and it is determined that the difference is due to a part of the difference, the adjacent character string correlation rule is extracted. Therefore, the collating procedure is repeated while changing the setting of the document characterizing rule, so that a common part can be extracted.

Since the number of times of extracting the adjacent character string correlation rule and repeating the collation is limited to the limited number, unnecessary repetition can be avoided.

When collating an input document with a table of contents,
Since the control code of the heading of the word processing document file can be used as the identifier of the position in the document or the identifier of the expression of the relationship between the sentence element and the entire document, an output document in a format utilizing the structure of the input document is generated. be able to.

For a document file mainly composed of a drawing, a character string can be extracted from the file and a characteristic table can be generated. A document in a text format and a document in a text format can be collated, and the collation target of a document file can be expanded.

Since the specialty in the character string conversion and the number of character strings that match the sort result can be added to the characteristic table as accompanying information, the expression of the collation result in the output document can be clarified. .

The process of dividing a document and extracting document components can be executed in units of arbitrary character strings, so that detailed matching can be performed in terms of terms, and the matching target can be changed from a document unit to a sentence unit. Thus, differences in composition and terminology between sentences can be compared.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of a matching apparatus for matching a single document;

FIG. 2 is a principle diagram of a document restoring unit based on a matching expression format designation rule.

FIG. 3 is a diagram illustrating the principle of a similar document collation apparatus for a single document.

FIG. 4 is a principle diagram of a collation device for a similarity relationship between a plurality of documents.

FIG. 5 is a principle diagram of a character string characteristic table generating means based on a document character string conversion rule.

FIG. 6 is a principle diagram of an apparatus for repetitively refining collation of similarity between a plurality of documents.

FIG. 7 is a diagram showing an example of a character string divided by a reading point when claim 1 is processed.

FIG. 8 is a diagram showing an example of an input document character string characteristic table when processing claim 1;

FIG. 9 is a diagram showing an example of an input document character string characteristic table when processing FIG. 1;

FIG. 10 is a flowchart illustrating an example of a processing procedure of a character string characteristic table generation unit (12).

FIG. 11 is a flowchart illustrating an example of a processing procedure of a document restoration unit (16).

FIG. 12 is a diagram showing an example in which a collation result in processing claims 1 and 2 is output in the form of an input document.

FIG. 13 is a diagram showing an example in which a collation result in processing claims 1 and 2 is output as a sort result.

FIG. 14 is a diagram showing an example in which the result of collation in processing claims 1 and 2 is output in the form of a correlation table.

FIG. 15 is a flowchart illustrating an example of a processing procedure of a similar conflicting relationship setting unit (21).

FIG. 16 is a flowchart illustrating an example of a processing procedure of a characteristic table combining unit (37).

FIG. 17 is a means for extracting an adjacent character string correlation rule (53).
6 is a flowchart showing an example of the processing procedure of FIG.

FIG. 18 is a flowchart illustrating an example of a processing procedure of a character string conversion rule setting unit (55).

FIG. 19 is a diagram showing an example in which a result of collation using terms as constituent elements is output in the form of a vertically arranged correlation table.

FIG. 20 is a diagram illustrating an example in which a collation result having terms as constituent elements is output in the form of a horizontally arranged sentence.

[Explanation of symbols]

 11, 31 Input document storage device 12, 32, 52 Character string characteristic table generating means 13, 33 Input document character string characteristic table 14, 39 Character string sorting means 15, 40 Output document character string characteristic table 16 Document restoring means 17, 43 Output document storage device 18 Matching relationship expression format designating rules 19, 25, 42 Document processing / restoring means 21, 34 Similar conflicting relationship setting means 22, 35 Similar conflicting term database 23, 36 Similar character string characteristic table 24, 41 Similarity relationship expression format Specification rule 37 Characteristic table combining means 38 Binding characteristic table 51 Document character string conversion rule 53 Adjacent character string mutual relation rule extracting means 54 Adjacent character string mutual relation rule 55 Character string conversion rule setting means S1 to S73 In the flow showing the processing procedure Step Y: Branch if the condition is satisfied (Yes) N Branch if the condition is not satisfied (No)

Claims (12)

[Claims] In a single document composed of a character string and a character string group, when judging a mutual match or a difference between character strings which are constituent elements of the document, the input document storage device reads the document from the input document storage device. And the character string property table generating means extracts the character string from the component, and a document identifier such as the document name together with the position identifier of the component in the document and the relationship between the component and the document. Generates an input document character string characteristic table that expresses the character string characteristics by including the expression code, etc., and the character string sorting means sorts the input document character string characteristic table by character string and creates an output document character string characteristic table Then, the document restoring means generates a document in which the expression of the matching relation is added to the format of the document from the output document character string characteristic table and outputs the document to the output document storage device, or outputs the output document character string characteristic table. A collation device for matching the character strings of constituent elements of a single document based on sorting, which is output to an output document storage device 2. A method according to claim 1, wherein the character string is a constituent element of a single document, and the character string is generated from the input document storage device. Express the character string characteristics by extracting the character string from the element and including a document identifier such as the document name and a relation expression code between the component and the document along with the position identifier of the component in the document The input document character string characteristic table is generated, the character string sorting means sorts the input document character string characteristic table by character strings to create an output document character string characteristic table, and the document processing / restoring means generates the output document character string characteristic table. A collating unit for matching the character strings of the constituent elements of a single document based on sorting, wherein the collating unit generates an output document in accordance with a rule for specifying a matching relationship expression format and outputs the generated document to an output document storage device 3. When determining the mutual agreement or difference between character strings that are constituent elements of a single document, the document is input from an input document storage device, and a character string characteristic table generating unit outputs Express the character string characteristics by extracting the character string from the element and including a document identifier such as the document name and a relation expression code between the component and the document along with the position identifier of the component in the document A similar conflicting relationship setting means converts the input document character string property table into a similar character string property table based on a similar conflicting term database, and the character string sorting means generates a similar character string property table. The output document character string characteristic table is created by sorting the table by character strings, and the document processing / reconstructing means uses the output document character string characteristic table to convert a document in a format easily associated with the document into a similar contradictory term database. Collating similarity of component character strings of a single document based on sorting, wherein an output document is generated in accordance with the rules for specifying a similarity relation expression format with reference to a source document and output to an output document storage device apparatus 4. In order to match a plurality of documents, when judging a mutual match or a difference between character strings as constituent elements, the same number of documents are input from the input document storage device and the characters are input. The column characteristic table generation means extracts the character string from the component, and includes a document identifier such as the document name, a relation expression code between the component and the document, together with a position identifier of the component in the document. The input document character string characteristic table of the number of documents expressing the characteristics of the character string is thereby generated, and the similar contradiction relation setting means converts the input document character string characteristic table of the relevant document number to the document in accordance with the similar contradiction term database. Number of similar character string characteristics tables, the characteristic table combining means converts the number of similar character string characteristics tables into a single combining characteristic table, and the character string sorting means converts the single combining characteristic table into a character string. sort To create a single output document character string characteristic table, and the document processing / restoring means uses the single output document character string characteristic table to convert the single Generating a number of output documents, referring to a similar contradiction term database and an input document character string characteristic table, performing a generation process in conformity with a similarity relation expression format designation rule, and outputting the generated output documents to an output document storage device. Device for collating similarity between component character strings of multiple documents based on sorting 5. When collating a plurality of documents, when judging a mutual agreement or a difference between character strings as constituent elements, the number of the documents is inputted from the input document storage device and the characters are inputted. The column characteristic table generating means extracts the character string from the component, changes the character string in accordance with the document characterizing rule, and a document identifier such as the document name together with the position identifier of the component in the document. An input document character string characteristic table of the number of the documents expressing the character string characteristics by including a relation expression code between the component and the document is generated, and the similar conflict setting means conforms to the similar conflict database. The input document character string characteristic table of the number of documents is converted into a similar character string characteristic table of the number of documents, and the characteristic table combining unit changes the similar character string characteristic table of the number of documents into a single combined characteristic table, and Row saw The output means sorts the single combined property table by character string to create a single output document character string property table, and the document processing / reconstruction means outputs the same number of documents from the single output document character string property table. When generating a single or the number of output documents in a format that can be easily correlated with each other, refer to the similar contradiction term database and the input document character string property table, and perform the generation process based on the similarity expression format specification rules Collating device for similarity relation among component character strings of a plurality of documents based on sorting, wherein the collation is output to an output document storage device 6. In order to collate a plurality of documents, when judging a mutual match or difference between character strings as constituent elements, the same number of documents are input from the input document storage device and the characters are input. The column characteristic table generating means extracts the character string from the component, changes the character string in accordance with the document characterizing rule, and a document identifier such as the document name together with the position identifier of the component in the document. An input document character string characteristic table of the number of the documents expressing the character string characteristics by including a relation expression code between the component and the document is generated, and the similar conflict setting means conforms to the similar conflict database. The input document character string characteristic table of the number of documents is converted into a similar character string characteristic table of the number of documents, and the characteristic table combining unit changes the similar character string characteristic table of the number of documents into a single combined characteristic table, and Row saw The sorting means sorts the single combined property table by the character string to create a single output document character string property table, and the extracting means of the adjacent character string correlation rule uses adjacent character strings in the output document character string property table. The character string is compared and the adjacent character string correlation rule is extracted in accordance with the extraction rule of the adjacent character string correlation rule. When generating a single or the number of output documents in a format that can be easily correlated with each other, refer to the similar contradiction term database and the input document character string property table, and perform the generation process based on the similarity expression format specification rules Then, when the adjacent character string correlation rule is extracted, the character string conversion rule setting means changes the setting of the document character string rule and repeats the collation procedure by a limited number of times. To continue within Wherein, matching device similar relationship component string plurality of documents based on the sorting 7. In a single document file in which a character string and a character string group are digitized by a character code, a character string is used when judging a mutual match or a difference between character strings which are constituent elements of the document. And sort by the original form of the character string group, and determine the matching relationship between adjacent character strings and character string groups based on the result. The document identifier such as the document name is added to the character string and the character string group together with the position identifier in the document. And output the discrimination result by adding the expression of the matching relationship to the format of the document by attaching the relation expression code between the component and the document, or output the characteristic table itself of the sorting result Characteristic Matching Method for Component Strings of Single Document 8. A method for matching a matching character string of a component of a single document, wherein the result of the matching method for matching is converted to a format conforming to rules for specifying a matching relationship expression format and output. 9. According to the matching method for matching, the character string is replaced with a reference character string in accordance with a similar contradiction term database, and then sorted and output in accordance with a similarity relation expression format designation rule. Matching Method for Similarity Relationships 10. A method for collating a plurality of documents based on the collation method for similarity of a single document, wherein all character strings of the plurality of documents are sorted and output documents are output. Collation method 11. A collation method, comprising: changing the arrangement order of character strings based on the collation method according to a document character string conversion rule; 12. An adjacent character string interrelation rule is extracted in accordance with an adjacent character string interrelation rule extraction rule by comparing adjacent character string groups obtained by sorting based on the collation method. A collation method characterized by setting a string correlation rule to a document string conversion rule and repeating a collation procedure.