JPH0242561A - Document doctrine structure extraction method - Google Patents

Abstract

PURPOSE:To realize the analysis of the logical structure of a document produced in a free form by preparing a specific processing step for a document editing device using a multi-window system. CONSTITUTION:In a 1st step a character string inputted by an operator according to the template specifications is stored in a logical structure file 4 as a logical structure template. In a 2nd step a list of names of logical structure templates which are previously registered are displayed on a template menu window according to an instruction of the logical structure analysis start by an operator. At the same time, the logical structure template is analyzed based on the template specifications and a hierarchical structure is stored in a logical structure memory 15. The operator points a character string subject on a document display window against each logical element name of the corresponding template and stores the pointing result as the contents of the hierarchical structure of the memory 15.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for extracting a logical structure from an existing document, and in particular, a method for extracting a logical structure from a document written without format restrictions. This invention relates to a logical structure extraction method suitable for extraction.

[Conventional technology]

When a computer processes a document, the document can represent a logical structure with a hierarchical structure such as titles, chapters, and sections. In order to handle the logical structure, logical elements (units that are semantically divided, such as chapters, sections, sections, and charts of a document) must be specified in advance. As a way to specify traditional logical elements.

Information Processing Society of Japan 34th (first half of 1986) National Convention Proceedings (II) P1309 describes the hierarchical structure of chapters, sections, etc. based on Rule 1 to identify whether it is a heading section or a content section, and the relationship between spans. A method for automatically extracting the logical structure of one document by writing identification rule 2 inside the system is described.

In this method, the input document is a solid document that has not been layout edited, and includes the line feed code as one unit. It is also assumed that the sentences are arranged in a logical class. The logical structure of a document is automatically extracted from an input document in such a format. Rules 1 and 2 for automatic extraction are written inside the system. First, a unit-by-unit analysis is performed according to Rule 1 to identify whether the element of the - sentence is a heading or a content part. After that, the connection of sentences in the document is analyzed according to Rule 2, and the hierarchical structure of identifiers, chapters, sections, etc., such as parent-child relationships and sibling relationships of each sentence, is extracted. These processes are performed in batch mode.

[Problem to be solved by the invention]

The above-mentioned prior art analyzes a document using rules for identifying whether it is a heading section or a content section and rules for identifying a hierarchical structure such as chapters and sections based on the relationship between sentences.

Therefore, in order to extract a logical structure, it is necessary to write the document according to these rules, so it is not possible to extract a logical structure from an existing document that is not written according to the above rules, and when creating a new document. There were problems such as not being able to write documents in a free format.

An object of the present invention is to solve the problems of the conventional methods and to enable logical structure analysis of documents created in a free format.

[Means to solve the problem]

In a document processing system using multi-windows,
The logical structure of a document is defined by logical element names (Chapter 9).

Sections, references, etc.) are called logical structure templates.

As a first step, a character string input by an operator using a template registration window according to template specifications is stored in a logical structure file as a logical structure template.

As a second step, in response to the operator's instruction to start logical structure analysis, a list of names of logical structure templates 1 recorded in advance is displayed in the template menu window, and the operator can select an appropriate template from the menu. Clear the template menu window and display the template in the logical element menu window based on . Internally, according to the template specification.

Analyze the logical structure template and save 1 on the logical structure memory.
Memorize the hierarchical structure. For each logical element name of the template, the operator specifies the character string target on the document display window, and the result is stored as the content of the hierarchical structure in the logical structure memory. ,
The second step makes it possible to analyze the logical structure of a document created in a free format.

Further, as a third step, after displaying the document in the document display window and displaying the template in the logical element menu window, the character string of the logical structure template is modified based on the template modification by the operator.

Analyze the data according to the template specifications and
Store in memory.

By adding the third step above, the standard logical structure template 1- can be registered in the first step, modified and used, reducing the operator's work and achieving the purpose of this method. can be achieved.

[Effect]

A logical structure is created by analyzing a logical structure template according to template specifications to generate a hierarchical structure, and storing character string objects selected by an operator on the screen for each element name in the hierarchical structure.

As a result, one document data body is not analyzed according to rules, so a logical structure can be generated even if one document is written in a free format.

In addition, by making it possible to modify the logical structure template while editing the document, the operator can reduce the work involved in registering the template, since all he has to do is modify and use the standard logical structure template that has been registered in advance. .

〔Example〕

First, the overall configuration of a document editing system to which the present invention is applied will be explained with reference to FIG. This system consists of a main controller (microprocessor) that controls the overall system operation;1. Memory 2 for storing various programs executed by the main controller 1. 32 work memories for temporarily storing data generated during the execution process of these programs 49 template files for storing logical structure template data created and referenced by the system 49 for storing document data, logical structure data, etc. data file 16
．． 5. Document data memory 17 stores document data loaded from data file 16; refresh memory (bitmap memory) stores data to be displayed on display screen 7; A display control device 6 that sequentially reads out the contents of the refresh memory 5 and outputs them to the display screen 7. Data records 41 to 47 shown in FIG. 4 corresponding to each virtual screen
Virtual screen memory for storing 8. Window management table memory 99 for storing the window management table 30 shown in FIG. 3; Memory 10 for storing character fonts corresponding to character codes. A bitmap processor (BMP) 11 that develops display data on the virtual screen as bitmap data on the refresh memory 5.
BMP command memory 1 for storing various commands for operating the bitmap processor 11
2. Logical structure memory 1 for storing logical structure data
52 This system includes a keyboard 13° for inputting various control instructions and data, and a pointing device 14 such as a mouse (MOUSE) for specifying a position on a display screen with a cursor.

Next, the multi-window system used by one system will be explained with reference to FIGS. 2, 3, and 4.

FIG. 2 shows a virtual screen 20 stored in the virtual screen memory 8.
．． In this example, the position and size of the window 25.26 are determined by the upper left end point of the window rectangle. XY seat 41!1!
(Xl, Yt) and the lower right end point of the XY locus JIA CXx
', Yt').

Data located within the window display areas 23, 24 of the virtual screen 20, 21 is displayed in windows 25, 26. Window display area 23. ．． The position and size of 24 are expressed by the xy coordinates of the upper left end point and the lower right end point of the rectangle, similarly to windows 25 and 26.

The correspondence between windows 25, 26 and window display areas 23, 24 shown in FIG. 2 is stored in the window management table 3o shown in FIG.

FIG. 3 shows the data items of the window management table 3o. 31 is the display priority order when windows overlap. 32 is the XY coordinate of the upper left end point of the window with the origin at the upper left end point of the actual display screen, and 33 is the X'Y' coordinate of the lower right end point. 34 is the xya mark of the upper left end point of the window display area with the origin at the upper left end point of the virtual screen; 3
Similarly, 5 is the x'y' coordinate of the lower right end point.

FIG. 4 shows the data items of the virtual screen memory 8, where 41 is the horizontal and vertical size representing the size of the virtual screen, and 42 is the total number of areas within the virtual screen. Data within the virtual screen is managed in area units as described below. 43 is the horizontal and vertical coordinates of the area whose origin is the upper left corner of the virtual screen, and 44 is the size of the area. 45 is the data type (text/figure/image) of the area, and 46 is the data attribute of the area. If the data type 45 is text.

Horizontal/vertical writing type, 2-line pitch, character pitch, etc.

If the data type 45 is a figure, it is the number of figures, etc.

If the data type 45 is an image, the information includes the compression format, number of gradations, etc. 47 is area data. If the data type 45 is text, it is a character code string, if it is a graphic, it is a graphic command string, and if it is an image, it is image data.

FIG. 5 is a logical structure data table stored on the logical structure memory 15. The logical structure table 50 consists of records 59 for each object (document, title, chapter, section, etc.). Each record has a type 51, which indicates the top level, intermediate level, and bottom level of the hierarchical structure. Identifier 52 indicating the hierarchical relationship of objects; Dependent 53 indicating the number of subordinate objects.
Content section 54 indicating whether the target has document content.
User-visible name 551 indicating the logical element name of the small elephant (abstract, chapter 1, etc.) Layout style identifier 56 that links with layout information such as in which frame the document content is displayed, located on the document data memory 17 It includes a start pointer 57 and an end pointer 58 indicating the position of display data 59.

FIG. 6 shows a sequence of logical structures with identifiers 52 attached, and the meanings of the object identifiers will be explained. Identifiers have a hierarchy level of 1
Each time the level goes down, the number increases by one digit, and in the case of targets of the same level generated from the same target, the last digit increases by 1 in the order of occurrence.For example, 0.

If the object that appears next to the identifier "3210" is at a lower level, it will be the identifier "32100"'; if it is at the same level, it will be "3211". In this way, the identifier indicates the position of the object in the hierarchical structure. It plays a role in clarifying the

Next, it is shown in FIG. The control flow for the document logical structure extraction process performed using the logical structure data structure will be explained with reference to the program flowcharts shown in FIGS. 14, 7, 8, and 9. . 1st
0 shows the keyboard 13 used for processing, and FIGS. 11 and 12 show examples of the display screen 7 during processing.

First, FIG. 14 shows the main routine of this embodiment. It waits for a command to be input by the operator (step 141), and if the template registration key 101 has been input (step 142).

Step 14: Start the template registration control routine
3) If the logical structure analysis key 102 is input (step 142), start the logical structure extraction control routine (step 144), and if the end key 103 is input,
End control.

FIG. 7 is a program flowchart showing details of the template registration control routine 143.

First, a template registration window 111 is set in the window management table memory 9 (step 72).
When the operator writes the template name using the character code key 104 and inputs the input key 107 (step 73
), inputs the character code string of the template name into the work memory 3 (step 74).

Then, the operator uses the character code key 104 to
Enter the logical element name. Enter the line feed key 108 at the end of each element name, and input based on the template specifications for analyzing the template on a computer (
Step 75).

The document includes the title, author name, etc. at the beginning of the document, chapter 1, section, afterword,
It can be divided into three parts: the end part of the document such as references, etc. The template specifications consist of the following four items for these three parts.

The first item is the restriction on logical element names at the beginning and end of the document (
For example, the title and title are unified with the title).

For the second item, the names of logical elements in chapters and sections should be written using numbers and periods (for example, 3.1, 2.).

The third item is the name of the logical element in the chapter 1 section. In order to express the subordination relationship between the chapter and section, a space is added before the logical element name each time the level of the 1 section is lowered (for example, "1." - 1
, 1"-1.1.1").

For the fourth item, enter the logical element name in the logical expansion category.

When the operator inputs the end key 103 (step 76), the data is temporarily input into the work memory 3 (step 77), and the template is stored in the template file 4 with the previously input template name as the file name (step 77). Step 78).

FIG. 11 shows when an operator inputs a logical element.
This is an example of the display screen 7.

Next, set the message window 123 in the window management table memory 9 (step 79), and register the r-separate template in the virtual screen memory 8 corresponding to the message window 123? ” (step 710). If the operator inputs the YES key 105, the process returns to step 73; if the operator inputs the No key 106, the process returns to step 711.
).

Delete the template registration window and message window, and delete the management data of the window from the window management table memory 9 (step 712);
The above is the template registration process.

FIG. 8 is a program flowchart showing details of the logical structure extraction control routine 144.

In response to the input of the logical structure analysis key 102 by the operator, a template menu window is set in the window management table memory 9 (step 91).Next, a list of template file names is read into the work memory 3 from the template file 4. , input into the virtual screen memory 8 corresponding to the template menu window (step 92), read the file with the template name specified by the operator with the mouse from the template file 4 into the work memory 3 (steps 93 and 94), and , the data is analyzed by the logical structure analysis routine (see FIG. 9) (step 95).Next, the logical element menu window 12 is stored in the window management table memory 9.
2 is set (step 96), and the template on the work memory 3 is output to the virtual screen memory 8 corresponding to the logical element menu window 122 (step 97).

While waiting for a command input (step 918), when the operator specifies a logical element name using the mouse on the logical element window 122 (step 919), the logical structure table in the logical structure memory 15 is searched and the specified The identifier 2 of the logical element name is manually entered into the work memory 3 (step 99).

Then, the message window 123 is set in the window management table 9 (step 910).

In the virtual screen memory 8 corresponding to the message window,
An example of this is shown in FIG. 12, in which a character code string for a message prompting the user to specify a character object is input (step 911).

When the operator specifies the character string displayed on the document display window 121 (step 912).

The start pointer and end pointer of the character string on the document data memory 17 are searched (step 913).

The result is transferred to the start pointer 57 .corresponding to the identifier input into the work memory 3 earlier. Store in end pointer 58 (step 914) and continue to step 918.

When the operator inputs the end key 103, the message window 123. Logical element menu window 122
and deletes the management data of the window from the window management table memory 9 (step 917). FIG. 9 shows a flowchart of the logical structure template analysis routine 95.

First, the logical structure tape JL15 on the logical structure memory 15
0 (7), o in items 51-54, 56-58, 56''
& This is initialized by inputting the null character code (step 80). Next, the logical element name written in the template is input from the work memory 3 to another address in the work memory 3, with the part up to the line feed code as a single logical element name (step 82).

If the data has ended, the control flow returns to the one shown in FIG. 8 (steps 83 and 84).

Next, check the first character code of the input logical element name and divide it into three types: numbers, space characters, and other characters (
Step 85), if the first character is a space, check the character code to see how many characters are spaces (step 86), and follow the rules shown in Figure 6 in the same way as for numbers and other characters. An identifier is calculated according to the following (step 87).

The result is stored in the logical structure table on the logical structure memory 15, and the process returns to step 82 (step 88).6 and above are logical structure extraction processing.

According to the above embodiment, when extracting a logical structure.

Since there is no need to analyze the data on the document data memory 17 other than searching for character data pointers, there is an advantage that the logical structure can be extracted even if nine documents are freely written.

However, in the first embodiment described above, if the registered template and the logical structure of the document do not match, it is necessary to register a new template. In this case, the number of template files becomes very large, and template registration may become a burden for the operator.

Next, a second embodiment will be described in which this problem is solved by making it possible to modify the template.

The control flow shown in FIG. 13 is a program flowchart that is an improved version of the logical structure extraction control routine shown in FIG. This will be explained below with reference to FIG.

In FIG. 13, first, the message window 123 is set in the template management table memory 9 (step 131).
Do you want to use the l1ffi recorded template for virtual screen memory 8 that corresponds to ? ” (step 132).

When the operator inputs the No key 106 (step 133), the operator sets the template registration window 111 in the template management table memory 9, writes the template name using the character code key 104, and after finishing, presses the input key. Enter 107 (step 13
4,135).

The template name is input into the work memory 3.

Then, the operator uses the one-character code key 104 to
Enter the logical element name (step 131.6).

At this time, the line feed key 108 is pressed manually for each element. Also, the description is performed based on the template specifications (step 1316
), when the operator inputs the end key 103, the written data is temporarily input into the work memory 3, the template name input earlier is set as the file name, and the template is stored in the template file 4 (step 131).
7-1319).

After that, the process is the same as after step 94 in Figure 8. If
When the YES key 105 is input in response to a message (
Step 133): First, a template menu window is set in the window management table memory 9 (
Step 137).

Then, the list of template names is read into the work memory 3 from the template file 4, and manually entered into the virtual screen memory 8 corresponding to the template menu window (step 138).The operator indicates the template name with the mouse (step 139).

Next, the logical element menu window 122 is set in the window management table memory 9, and the data of the file specified by the operator is input from the template file 4 into the virtual screen memory 8 corresponding to the window 122 (steps 1310 and 1311). ), and inputs a message character code string indicating whether or not to modify the template into the virtual screen memory 8 corresponding to the message window (step 1312), and the operator presses the No key 1.
If 06 is entered, step 948 continues.

If the YES key 105 is input, the template registration window 111 is stored in the window management table memory 9.
, and manually input the file specified by the operator into the virtual screen memory 8 corresponding to the window 111 from the template file 4 (steps 1314 and 1315).
). Then, as in the case of new registration, step 1
Steps 313 to 1319 are performed and the process continues to step 94.

According to the second embodiment described above, in advance! : If you register an A$-like logical structure template, you can extract the logical structure just by modifying the template, and even if the document has a special logical structure, you can easily register a new template. Operator work is reduced and it is possible to respond flexibly to a variety of documents.

〔Effect of the invention〕

As is clear from the above description of the embodiments, according to one aspect of the present invention, the logical structure of a document is registered in advance using a simple method,
Since the correspondence between a sentence unit in a document and the mth registered logical element can be performed interactively, it is possible to provide a means for easily extracting the logical structure of documents in various formats.

[Brief explanation of the drawing]

Figure 1 shows a multi-window system according to the present invention.
FIG. 2 is a block diagram showing the overall configuration of the screen, FIG. 2 is a diagram showing the relationship between the window display area on the virtual screen and the windows on the real display screen, FIG. 3 is an explanatory diagram of the window management table, and FIG. 4 is an explanatory diagram showing the configuration of a virtual screen memory, FIG. 5 is an explanatory diagram of a logical structure data table, FIG. 6 is an explanatory diagram showing a logical structure with an identifier attached, and FIG. The program flowchart executed for template registration, FIG. 8, is the program flowchart executed for extracting the logical structure, FIG. 9.
The figure shows a detailed program flowchart of the logical structure template analysis routine in Figure 8, and Figure 10 is a plan view of an example of a keyboard used in a single system.
FIG. 11 is a diagram showing a display screen when registering a template, FIG. 12 is a diagram showing a display screen when extracting a logical structure, and FIG. 13 is a diagram showing a display screen when registering a template. The program flowchart shown in Figure 14 is the program flowchart shown in Figure 14.

Claims (1)

[Claims] 1. In a document editing device that uses an input means and a multi-window system in which a plurality of windows are set on a display screen, an operator can determine the hierarchical relationship of logical element identifiers using logical element identifiers. A first step of displaying the logical structure of the document input according to the description shown in the first window set on the display screen and storing the same input data in a logical structure file, and a first step of storing the data stored in the first step. The hierarchical structure of logical element names is read from the logical structure file and stored in the logical structure memory based on the specifications used when inputting data in the first step. Display the document in the window, and enter the logical element identification name in the first window and the logical element identifier in the second window by the operator.
a second step of generating a logical structure by storing the result in a logical element identification name area of a hierarchical structure on a logical structure memory based on an instruction for correspondence with a character string for each logical element of a document in a window; A document logical structure extraction method characterized by having the following. 2. Display the document in the window set on the first display screen, read the data stored in the first step from the logical element file, display the data of the logical element name in the second window, and display the data on the second window by the operator. Modify the data in the logical structure file based on the instruction to modify the logical element name data in , and when the modification is completed, read the data from the logical structure file and change the hierarchical structure of logical element identifiers to the logical structure based on the specifications. 2. The document logical structure extraction method according to claim 1, further comprising a third step of storing the document in a memory.