JP3501240B2 - Document creation support device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document creation support apparatus capable of detecting a contradiction between the contents of a table and the contents of a sentence in a document including a table.

[0002]

2. Description of the Related Art Conventionally, there is a document proofreading device for proofreading a document created by a word processor, a desktop publishing system or the like. Such a conventional document proofreading device detects and corrects an error in the kana kana kana notation in a created document (for example, a method of constructing proofreading knowledge in a document proofreading support system, IPSJ). Research Report Vol.92, No.21, pp.79-86 (Information Processing Society of Japan
Published March 13, 1992)).

[0003]

However, the above-mentioned conventional techniques have the following problems. That is, when the contents shown in the table are explained in sentences or the contents explained in the sentence are collectively shown in the table, the contents explained in the sentences often conflict with the contents of the table. For example, when explaining the sales situation for each branch of a certain retail store in a table and a sentence, while “Sales” of “Osaka Branch” is “20 billion yen” in the table, it is “10 billion yen” in the sentence. This is the case when it is a "yen". Such a contradiction occurs because either the table or the sentence is wrong, but conventionally, such a contradiction cannot be resolved unless the user reviews both the table and the sentence and finds an error. could not.

[0004]

The document creation support device of the present invention is characterized by the following points in order to solve the above-mentioned problems. The present invention translates documents including tables and sentences into sentences.
It is read into the memory as calligraphy data, and the table and sentences are
A text area consisting only of table area data and natural language text
Document input means for converting to data and storing the data, and the memory
In the tablespace data from the tablespace data stored in
The columns and rows are detected by the frame of the
Recognized characters and identify headings and attributes for each recognized character.
The registered table word dictionary is generated and registered in the table word dictionary.
Based on the position of each recognized character on the table and the part of speech,
Generates a combination of multiple consecutive characters and stores them in a predetermined storage area.
Table meaning conversion means to be stored and the above-mentioned stored in the memory
For sentence area data, word notation and part of speech for each word
Morphological analysis is performed for each sentence using the word dictionary defined by
To a large number of words with parts of speech
And the part of speech of the recognized character registered in the front word dictionary
Morphological analysis means for selecting a plurality of matching words and the selection
The plurality of selected words are written in the predetermined meaning of the table meaning conversion means.
Syntactic meaning to combine in correspondence with the combination stored in the memory area
Of the characters generated by the taste analysis means and the table meaning conversion means.
Combination of combinations and words generated by the syntactic and semantic analysis means
Semantic matching is used to match characters and detect unmatched characters and words.
The user can correct the matching means and the detected characters and words.
And a contradiction candidate display means for displaying in Noh .

In the present invention, the contradiction candidate display means
The user can use any of the letters and words displayed by
Of the document data in the memory when
Document modification means for generating document data with this part corrected
It is further characterized by being provided .

[0006]

(1) The table area data of the input document data is read from the memory by the table meaning conversion means ,
A table word dictionary in which the recognition characters, their headings and attributes are registered is created , and multiple recognition characters in this table word dictionary are created.
And a combination of recognition characters related based on the position and part of speech
Is generated and stored in a predetermined storage area . On the other hand, the text area data of the input document data is the morphological analysis means, read from the memory, a word dictionary
Referring to the analysis is performed, the recognized character and elegance of the table word dictionary
A plurality of words having the same words are selected. Then, the syntactic and semantic analysis means analyzes the syntactic meaning according to, for example, a syntactic and semantic analysis rule, and a combination sentence in a predetermined storage area
Word combinations that match the letters are generated. These combinations are created in the same format, and as a result, it is possible to match the two by the meaning matching means. If the characters and words of both combinations do not completely match and only a part of them does not match, the matching means determines that one of them may be erroneous and considers it as a contradiction candidate. These contradiction candidates are displayed on the display by the contradiction candidate display means. As a result, the user can complete the proofreading work only by checking only the portions listed as these contradiction candidates.

(2) In (1), when the inconsistency candidates are displayed by the inconsistency candidate display means, a serial number is attached and a predetermined field of the display device is added.
By inputting the serial number to and inputting the changed data, it is possible to easily perform the calibration work.

[0008]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram of an embodiment of the document creation support device of the present invention. The illustrated apparatus comprises a document input means 1, a table meaning conversion means 2, a morpheme analysis means 3, a syntactic meaning analysis means 4, a meaning collation means 5, a contradiction candidate display means 6, a document change means 7 and the like.

The document input means 1 is an optical system such as a lens, C
It is composed of a CD, a scanning system, etc., and optically reads a document written on a paper surface, converts it into digital data, and reads it into a memory (not shown). It is assumed that the document has various areas such as a document area consisting only of natural language sentences, a table area, a drawing area, and a graph area. Then, the document input means 1
It has a function of separately extracting the elements constituting these documents from the memory. That is, the document input means 1 uses a sentence, a table,
Each data of figure and graph can be stored separately and retrieved separately. Such a document input means 1 may be configured together with a means for loading a memory from a secondary storage device such as a magnetic disk or a magneto-optical disk, or a document for converting a document printed on paper into code information. It may be configured together with the recognition device.

The table meaning conversion means 2 is composed of a processor or the like, and converts the contents of the table area stored in the memory into a table word dictionary 9 and a table meaning expression 10. That is, the columns and rows are detected from the table by a frame, the characters written in each column and row are recognized,
The table format is converted into the table word dictionary 9.
Further, it is converted into a format of a table meaning expression 10 representing a certain set of meaning contents. The table meaning expression 10 created in this way is stored in a predetermined storage area such as a disk device. The morphological analysis unit 3 is composed of a processor or the like, and divides a sentence in a document area, which is stored only in a natural language and is stored in a memory, into words using a word dictionary 8 in which word notation and parts of speech are defined for each word. To do. That is, the morpheme analysis unit 3 divides the sentence stored in the memory into words, searches each word in the word dictionary 8, and specifies the part of speech.

The syntactic and semantic analysis means 4 is composed of a processor and the like, and uses the syntactic and semantic analysis rule 11 to convert an input sentence divided into words into a sentence and semantic representation 12 having a unit of meaning. That is, even if the input sentence is a so-called complex sentence having a complex meaning, the complex sentence analysis rule 11 converts the complex sentence into a simple sentence having one group of meanings. This simple sentence is expressed in the form of sentence meaning expression 12.

The meaning collating means 5 is composed of a processor or the like, and collates the table meaning expression 10 and the sentence meaning expression 12 with each other. That is, by comparing both constituent elements, if all constituent elements are in agreement, it is regarded as the same meaning content, and if all elements are not in agreement, they are regarded as separate meaning content. Then, when only a part of the data matches and a non-matching part is included, it is determined that there is a possibility of erroneous writing of the contents that have the same meaning content, and they are considered as contradiction candidates. The contradiction candidate display means 6 displays the matching result of the meaning matching means 5 to the user by a display device such as a display. For example, an underlined portion or the like is added to a portion which is cited as a contradiction candidate, and a portion which may be erroneous is clearly indicated to the user. The document changing unit 7 displays a predetermined input field on a display or the like, and the user performs a process for changing the content of the document by designating a correction portion for the portion pointed out by the contradiction candidate displaying unit 6. That is, the position of the corrected portion in the document is detected according to the designation from the user in the input field, and the content of the table or the sentence is changed.

Next, the operation of the above-mentioned device will be described. First, with reference to FIG. 2, an outline of a processing flow in the apparatus of FIG. 1 will be described. In step S21, the document input means 1 reads a document into a memory (not shown). Then, in step S22, if there is no table in the read sentence, the process ends. In step S23, the table meaning conversion means 2 converts the table area in the document in the memory into the table meaning expression 10 and the table word dictionary 9. In step S24, the morpheme analysis unit 3 uses the word dictionary 8 and the front word dictionary 9 to perform morpheme analysis on the sentence in the document area in the document in the memory. Details of this analysis will be described later. In step S25, the result of the morphological analysis is converted by the syntactic and semantic analysis means 4 into the sentence and semantic representation 12 using the syntactic and semantic analysis rule 11. In step S26, the meaning collation means 5 collates the table meaning expression 10 with the sentence meaning expression 12. In step S27,
As a result of the collation, when the meaning contents of the table and the meaning contents of the sentence are inconsistent, the inconsistency candidate display means 6 displays the inconsistency candidates in step S28. If no contradiction candidate exists, the process ends without displaying.

Next, the operation will be described using the actual data shown in FIGS. An example of the input document is shown in FIG.
In step S21 of FIG. 2, the document of FIG. 3 is read into the image memory. The document shown in FIG. 3 is composed of sentences and tables. The text and the table data 31 and 32 are separately stored in the document. In step S22, the presence or absence of the table is determined depending on whether the data read into the image memory includes the table data. Can be detected. If the document contains a table, the table is referred to as a table semantic representation 10
To the front word dictionary 9. The result of converting the table of FIG. 3 into the table word dictionary 9 is shown in FIG.

The table is generally composed of the names of the items in the table and the data corresponding to those items. Portions 33 and 34 describing the names of the items in the table of FIG.
Indicates that the first column (hereinafter referred to as "column 1") is "branch", column 2 is "sales", and the description specifying the unit of column 2 is "100 million yen".
Is. The table word dictionary 9 shown in FIG. 4 is obtained by extracting these descriptions and converting them into a predetermined format. That is, the table word dictionary 9 is composed of the headline 401 and the attribute 402, and it is determined that the “branch” described at the top of the column 1 is the name 403 of the column 1 according to the description format of the table. Similarly, the “sales” at the top of column 2 is the name of column 2 404.
Is determined. On the other hand, “100 million yen” enclosed in parentheses at the top of column 2 is the unit of data in column 2 405.
Is determined.

The data 35, 36, 37, 38 of each column are described below the second column of each column. For example,
For “branch” in column 1, data such as “Osaka”, “Kyoto”, “Shiga”, and “Mie” are described. Then, for “sales” in column 2, “200”, “20”
"0", "100", and "100" are described. In this case, the unit is "100 million yen". These data are stored after the unit 405 in the table word dictionary 9 of FIG. That is, first, for the data in column 1, "Osaka",
“Kyoto”, “Shiga”, and “Mie” are stored in that order, and their attributes are column 1 data 406, 407, 408, and 409. Next, for the data in column 2, "200",
“200”, “100”, and “100” are stored in order,
These attributes are column 2 data 410, 411, 412, 4
It is supposed to be 13.

FIG. 5 shows a table semantic representation 10 created from the table of FIG. The left side of “=” is the column name 51, and the right side is the data 52 of the column. By this "=", "|", and "|", one table meaning expression 10
Is configured. That is, the table meaning expression 53 expresses the meaning that “the sales of the Osaka branch is 20 billion yen”.
The table meaning expression 54 expresses the meaning that “the sales amount of the branch in Kyoto is 20 billion yen”. The table meaning expression 55 is
It expresses the meaning that “the sales of the Shiga branch is 10 billion yen”. Further, the table meaning expression 56 expresses the meaning that “the sales of the Mie branch is 10 billion yen”.

Next, in step S24, morphological analysis is performed on the text area in FIG. 3 using the word dictionary 8 and the front word dictionary 9. An example of the word dictionary 8 is shown in FIG. The word dictionary 8 includes a word headline 61 and a part of speech 62.
First, the sentence “Osaka branch has sales of 10 billion yen” in the sentence of FIG. 3 is divided into words. Then, the word dictionary 8 is searched using the word "Osaka" in the first row 63 as a key. Then, "Osaka" is judged to be a noun part of speech. On the other hand, the front word dictionary 9 is searched using "Osaka" as a key. Then, "Osaka" is determined to be column 1 data in the table. Hereinafter, for “branch”, “ha”, “,” and the like, the morphological analysis is performed using the word dictionary 8 and the front word dictionary 9. The result is shown in FIG. 7. According to FIG. 7, for example, “Osaka” 71 is a noun 72, and column 1 data 7
It is represented as 3.

In step S25, the sentence area shown in FIG. 3 is subjected to syntax analysis and semantic analysis. The rules used in the syntactic analysis and the semantic analysis are shown in FIG. Each rule includes a rule ID 81, a condition 82 for applying the rule, and a conversion operation 83 when the condition is satisfied. In FIG. 8, N1 and N2 represent noun words. V1 represents a word whose part of speech is a numerical value. U1 represents a word whose unit of speech is a unit of speech. Places enclosed in quotation marks ““ ”and“ ”” represent character strings in the document. Also, the part of speech is shown after ":". The syntactic and semantic analysis means 4 applies the syntactic and semantic analysis rule 11 as shown in FIG. 8 to the result of the morphological analysis shown in FIG. That is, the condition of each rule is compared with the result of morphological analysis to check whether they match. If they match, the conversion operation of the matched rule is executed. Apply these rules until there are no more applicable rules.

For example, when the rule of FIG. 8 is applied to the result of the morphological analysis of FIG. 7, first, the rule R1 is applied to the “Osaka branch” part. That is, "Osaka" of "Osaka branch"
Corresponds to N1, and “branch” corresponds to N2. As a result, the conversion operation determines that "branch" is the name of the column and "Osaka" is the data. This can be expressed in the form of FIG. 5 as follows. | Branch = Osaka | In addition, rule R3 is applied to the portion of “sales is 10 billion yen” in the result of the morphological analysis in FIG. That is, when the rule R3 is applied, "sales", "10"
"0" and "100 million yen" correspond to N1, N2, and U1, respectively. As a result, the conversion operation determines that “sales” is a column name and “10 billion yen” is a value.

This can be expressed in the form of FIG. 5 as follows. | Sales = 20 billion yen | In addition, the result of the morpheme analysis in FIG. As a result, by the conversion operation, the following sentence semantic expression is written. | True / False value = True | When the above three sentence meaning expressions are put together, a sentence meaning expression as shown in FIG. 9 is obtained. In the sentence semantic expression of FIG. 9, except for the “true / false value”, the column name 91 shows the column name 51 of the table semantic expression, and the value of the data 92 of the column is the data 52 of the table semantic expression. Indicates the value.

In step S26, the sentence meaning expression and the table meaning expression are collated. In the present embodiment, with respect to the sentence meaning expression whose truth value is true, if the corresponding values are the same for all column names, the sentence meaning expression 93 and the table meaning expression 5
3 is considered to be an exact match. Otherwise, it is considered not to match. Then, of the sentence meaning expressions and the table meaning expressions that do not match, those having a different value by one are collected as contradiction candidates. That is, in FIG. 5, three table meaning expressions 53,
Collect 55 and 56 as contradiction candidates. On the other hand, those having two or more different values are not considered as contradiction candidates. That is, in the table meaning expression 54 of FIG. 5, since the “branch” and the “sales amount” are different, the table meaning expression 54 does not become a contradiction candidate and is not a collection target.

In step S27, it is determined whether or not there is a contradiction candidate in step S26. In the illustrated example, there are some inconsistencies, and the result is displayed in step S28. FIG. 10 shows a display example of contradiction candidates. Underlines (1) and (2) in the display result indicate the locations corresponding to the values of the sentence meaning expression 93 in FIG. The underline (3) is the sentence semantic representation 93 of FIG.
6 shows an inconsistent part of the table meaning expression 53 of FIG. Similarly, underlines (4) and (5) indicate parts that contradict the table meaning expressions 55 and 56 in FIG. The user of this device sees this display and indicates the number of the underlined portion to be corrected in the field 101.
, And the correct value can be entered in field 102. After that, by pressing the change button 103, the change is reflected in the document by the document changing means 7. Then, pressing the end button 104 ends the process.

FIG. 11 shows the document after the correction is completed when the underline (3) is selected in FIG. 10 and "100" is input as the changed value. In this way, it is possible to automatically find a portion where there is a contradiction between the contents described in the text of the document and the contents of the table, and it is possible to create a correct document.

The present invention is not limited to the above-mentioned embodiments, and it goes without saying that various modifications can be made. For example, the syntactic and semantic analysis rule 11 is not limited to the illustrated example, and any other rule can be added. This makes it possible to analyze more complicated sentences. Further, in the above-described embodiment, the contradiction candidates collected by the meaning matching means 5 are different in value by one, but the value is not limited to this, and two or more different ones may be included. This allows a wider range of contradiction candidates to be collected. Further, the name of the part of speech in the word dictionary 8 can be changed appropriately. Further, the word dictionary 8 may include information other than headwords of words and parts of speech. Further, the user may use a pointing device such as a mouse, voice recognition, a touch panel, or the like as means for selecting the underlined number of the inconsistent portion from the display of the inconsistency candidates.

When the names of the items in the table are composed of a plurality of words, a morphological analysis similar to the morphological analysis for the sentence in the above-mentioned embodiment is performed on such a plurality of words, and then, The table meaning conversion means 2 may create the meaning expression of the table.

[0027]

As described above, according to the document creation support apparatus of the present invention, the recognized characters in the table and the words of the sentence data are listed.
Create a combination and combination of words recognized character representing unified manner in conjunction bets. Thus matching letters and words of both combinations, detecting the contents of a table, the inconsistency of the contents of the text You can Therefore, it is possible to automatically detect a contradiction that may occur when the contents shown in the table are explained in sentences or the contents explained in the sentence are collectively shown in the table. As a result, the user can be prompted to correct an error in either the table or the sentence, and such a contradiction can be resolved.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a document creation support device of the present invention.

FIG. 2 is a flowchart illustrating a processing procedure of the apparatus of FIG.

FIG. 3 is an explanatory diagram of an example of a document to be proofread by the apparatus of FIG.

FIG. 4 is an explanatory diagram of an example of a table word dictionary created from the table of FIG.

5 is an explanatory diagram of an example of a table meaning expression created from the table of FIG.

FIG. 6 is an explanatory diagram of an example of a word dictionary prepared in advance.

FIG. 7 is an explanatory diagram of an example of a morpheme analysis result.

FIG. 8 is an explanatory diagram of an example of a syntactic and semantic analysis rule prepared in advance.

FIG. 9 is an explanatory diagram of an example of a sentence semantic expression created as a result of syntactic and semantic analysis.

FIG. 10 is an explanatory diagram of an example of a display example of contradiction candidates.

FIG. 11 is an explanatory diagram of an example of a document correction result.

[Explanation of symbols]

2 Table meaning conversion means 3 Morphological analysis means 4 syntactic and semantic analysis means 5 Meaning matching means 6 Contradiction candidate display means 7 Document change means

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hisaaki Matsushita 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (58) Fields investigated (Int.Cl. ⁷ , DB name) G06F 17 / 21 550 G06F 17/21 548

Claims (2)

(57) [Claims] 1. A document input means for reading a document including a table and a sentence into a memory as document data, converting the table and the sentence into sentence region data consisting of only the table region data and the sentence of natural language, and storing the document region data. From the table area data stored in the memory, the columns and rows are detected by the frame of the table in the table area data, the characters written in each column and row are recognized, and the heading and the attribute for each recognized character are identified. A table for generating a registered table word dictionary, generating a combination of a plurality of related characters based on the arrangement position on the table of each recognized character registered in the table word dictionary, and a part of speech, and storing the combination in a predetermined storage area. With respect to the sentence area data stored in the memory, the meaning conversion means performs a morphological analysis for each sentence using a word dictionary that defines word notation and part-of-speech for each word, and adds a large number of words To Morphological analysis means for selecting a plurality of words that match the part of speech of the recognized character registered in the table word dictionary from the plurality of words, and the selected plurality of words for the table meaning conversion means. The syntactic and semantic analysis means for combining in correspondence with the combination stored in a predetermined storage area, the combination of characters generated by the table meaning conversion means and the combination of words generated by the syntactic and meaning analysis means are collated. ,
A document creation support apparatus comprising: a meaning collating means for detecting unmatched characters and words; and a contradiction candidate display means for displaying the detected characters and words in a correctable manner by a user. 2. A document change for generating document data in which a corresponding portion of the document data in the memory is corrected when a user corrects any of the characters and words displayed by the inconsistency candidate display means. The document creation support apparatus according to claim 1, further comprising means.