JP3206600B2 - Document generation device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer for performing document processing such as a word processor, and in particular, a user gives additional information to a document by a simple operation, thereby extracting information such as a summary sentence. TECHNICAL FIELD The present invention relates to a document generation device capable of easily obtaining a new document in which the execution has been performed.

[Conventional technology]

In recent years, with advances in computer technology, documents can be handled at high speed on computers, and with this, documents have been rapidly digitized in offices. In particular, once digitized documents are easy to process, research on machine translation and elaboration support is active. However, since there are many technical problems in practical use of these, their applications are easily limited, and they have not yet been widely used. Therefore, an information processing device or a dictionary lookup device has been developed in which words constituting an electronic sentence are checked using an electronic dictionary and annotations and markings are added.

[Problems to be solved by the invention]

However, a dictionary lookup device or the like is a supportive device for translation and comprehension of manuscripts, and has a problem that it is difficult to directly use an output result as a document.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a document generation apparatus capable of directly using an output result as a document and outputting a high value-added document effective for translation, understanding of a document, and the like.

[Means for solving the problem]

The present invention provides a storage unit for storing a document, a selection unit for sequentially selecting paragraphs of the document stored in the storage unit, a first input unit for inputting a predetermined format, and a second input unit for inputting a keyword. Input means, first determining means for determining whether the paragraph of the document selected by the selecting means is formed by one sentence, and determining that the first determining means is formed by one sentence In the second case, it is determined whether or not the character arrangement in the sentence matches the predetermined format.
The first extracting means for extracting the sentence and the first judging means are not formed of one sentence when the judging means and the second judging means judges that the sentence matches the predetermined format. If it is determined that the keyword exists in the paragraph, a third determining unit that determines whether or not the keyword exists, and if the third determining unit determines that the keyword exists, a sentence including the keyword is extracted. It is characterized by comprising a second extracting means, and a document generating means for generating one document sequentially holding one sentence extracted by the first extracting means and the second extracting means.

Examples of processing performed by the first determining means, the second determining means, the first extracting means, the third determining means, and the second extracting means include steps 102, 111, 11 in FIG.
The processes of 2, 118 and 119 correspond respectively.

[Action]

In the present invention, the storage means stores the document, and the selection means sequentially selects paragraphs of the document stored in the storage means,
The input means inputs a predetermined format, the second input means performs a keyword, and the first determination means determines whether a paragraph of the document selected by the selection means is formed by one sentence, When the second judging means judges that the first judging means is formed by one sentence, it judges whether or not the character arrangement in the one sentence matches the predetermined format, and performs the first extraction. When the means determines that the second determination means matches the predetermined format, the sentence is extracted, and the third determination means determines that the first determination means is not formed of one sentence. In this case, it is determined whether or not the keyword exists in the paragraph. If the second extracting unit determines that the keyword exists, the second extracting unit extracts one sentence including the keyword and generates a document. Means for extracting the first extraction means and the second extraction means; Means are adapted to generate a sequence retained one document sentence extracted by the.

〔Example〕

Hereinafter, an embodiment of a document generation apparatus according to the present invention will be described.

FIG. 1 is a block diagram showing a conceptual configuration of a document generation device according to the present invention. This document generation device includes an original text storage unit 11 that electronically stores an original document input by a user as document data, and a condition for creating a new document in the original document stored in the original text storage unit 11. A generation condition input unit 12 that is set based on the generation condition input unit 12, and a sentence extraction unit 13 that extracts a sentence having a specific syntax from the original document stored in the original text storage unit 11 according to the generation condition input from the generation condition input unit 12. And a document generation unit 14 for generating a target new document using the plurality of sentences extracted by the sentence extraction unit 13.

FIG. 2 is a block diagram showing a basic configuration of the document generation device according to the present invention. In the figure, the same reference numerals as those in FIG. 1 indicate the same parts.

The original text input unit 15 is a part that converts a document to be processed into a form of digitized document data so that it can be processed by a computer.After the original document is converted into document data by the original text input unit 15, It is stored in the original text storage unit 11. Information processing unit 10
Generates a new document based on the original document stored in the original text storage unit 11 and the generation conditions input from the generation condition input unit 12. The information processing unit 10 includes a microprocessor, an original sentence scanning unit 18 that cuts out words by the dictionary unit 16, a sentence extraction unit 13 that extracts a specific sentence from the original document stored in the original sentence storage unit 11 according to input conditions, and And a document generation unit 14 for generating a new document from the extracted sentences.
The document generated by the information processing unit 10 is temporarily stored in the result storage unit 17 and output from the display unit 19 or the printing unit 20.

It should be noted that a means for directly inputting an original sentence may be connected to the original sentence storage unit 11, and a result output means and a storage means may be separately provided for the document generation unit 14. Also, specific sentences extracted from the document include a statement having a specific syntax such as an item name,
Sentences having specific words such as keywords are included, and various electronic dictionaries may be provided for extracting these sentences and generating documents.

FIG. 3 is a system configuration diagram when the above-described document generation device is applied to a Japanese summary sentence generation system. This system is configured to output a summary sentence for various types of Japanese printed matter by specifying simple conditions, and the summary sentence is mainly used to promote understanding of a long sentence and to save time.

In FIG. 3, reference numeral 21 denotes an optical character recognition device (OC) which is an original input unit for reading an original A and converting it into document data.
R) and 22 are an edit / information input unit which is a generation condition input unit for inputting various data and commands such as generation conditions, and 23 is a printing unit for printing and outputting a generated new document on recording paper. Is a laser printer. Other configurations are the same as those in FIG. 2, and the same parts are denoted by the same reference numerals.

Next, the operation of the Japanese summary sentence generation system described above will be described.
The description will be made with reference to FIG. 2 to FIG.

First, the original text A is read using OCR21. In this example, the original text A is a printed matter, but any document that has already been digitized by a word processor or the like may be read via a storage medium such as a floppy disk. Also,
A document creating means capable of creating a document on the spot may be provided, and the edited result may be used as it is as the original text. In this case, the input means is not limited to typing using a normal keyboard. For example, if an online handwritten character recognition using a tablet is used, even a person who is unfamiliar with the computer can easily create the original text.

The original sentence A converted into the document data by the OCR 21 is temporarily stored on the magnetic disk of the original sentence storage unit 11, and then subjected to a summary sentence creation process in the information processing unit 10. The processing described below may be performed each time the original text is input, or may be selected from a plurality of stored documents.

First, the editing / information input unit 22 sets a generation condition. FIG. 4 shows an example of a case where a generation condition is input, and shows a display state on the screen of the display unit 19. In this example, the chapter names (1.
, 2 ...., etc.) and the names of the sections (2.1, 2.2, etc.), and the sentences containing the bullet points and keywords are extracted as a summary sentence. As shown in FIG. 4, the input of the generation conditions is to input the names of chapters and sections in the item section, and specific examples in the section and keyword sections. In the initial screen,
General conditions such as selection of items and bullets are preset.

These generation conditions may be input from the keyboard as long as the original text is relatively short, but can also be transcribed directly from the original text. For example, if you select the item “High-speed virtual machine of XXXXX” from the document holder 1 in FIG. 4,
In the window of the other area (not shown) on the screen,
Figure) is displayed. Here, for example, when inputting the name of an item in the original text, selecting "item" in the condition input menu opens the area on the right side, and the user sequentially selects the corresponding portion of the original text using a pointing device (mouse). And combine them. By similarly inputting the item of the bullet point and the keyword, the generation condition can be easily set. Since the loaded document is displayed in the document holder 1, after setting the generation conditions,
Select the document you want to summarize again and specify "Start"
The information processing unit 10 starts a sentence extraction and generation process.

Next, in the original text scanning unit 18 (FIG. 2) of the information processing unit 10,
The document is structured. First, the dictionary section
Cut out words while referring to 16. An example of word segmentation at the beginning of the original sentence is shown below.

| 1 || Introduction | | XXXXX | is | ** like || I || System || But | Execution speed | is |… After such word unit division is completed, the structure as shown in Fig. 6 Make the attachment. Here, focusing on a period (.) Or a period (,) and a line feed symbol, a sentence including other symbols in the sentence is put together as a sentence, and the sentence is divided into paragraphs by a portion where the line feed symbol is added. At this time, numbers are assigned to words, sentences, and paragraphs, respectively, so that paragraphs have the beginning and end sentence numbers, and sentences have the beginning and end word numbers. Based on this result, the sentence extraction unit 13 extracts a specific sentence according to a procedure described later according to the set generation condition,
The document is sent to the document generation unit 14. The document generation unit 14 generates a new document by combining the extracted sentences. The generated document is temporarily stored in the result storage unit 17, and is output to the CRT display of the display unit 19 or the laser printer of the printing unit 20 according to the designation at the start. If the output result is unsatisfactory, the contents can be edited by an editing device (not shown), and then a plurality of documents can be output together.

Next, the processing procedure for extracting a target sentence in the information processing section 10 will be described based on the flowchart of FIG. 7 with reference to FIG.

First, a sentence having a specific syntax (item name and itemized list) is extracted from the document structured by the original sentence scanning unit 18. Since these are generally only one line, and have the property of having a sequence mark at the beginning, they are extracted by comparing them with the format of the specific example entered by the user or the basic format set in the system itself. can do.

In FIG. 7, first, it is determined whether or not the cut paragraph is the last paragraph in order, that is, whether there is any remaining paragraph that has not been processed (step 101). Is determined as a paragraph having only one sentence (step 102). Here, it is checked from the structure of the paragraph whether the paragraph is composed of several sentences, and if the first and last sentence numbers match, it is determined that the sentence is one sentence.

Here, when the paragraph is a paragraph having only one sentence, the format of this sentence is obtained (step 103). The format is a description of the arrangement of symbols and the like in the sentence. It is expressed in a simple form. An example is shown below.

#: 1 numeric character ＄; 1 uppercase letter ＠; 1 lowercase letter%; 1 other symbolic character (eg, Iroha, Roman numeral, etc.) _; space →; tablation *; character string Other symbols ( ,>,-, Etc.) are written as they are. The above special symbols are For example, a format of “→ #. # _ *” Is determined from “1.2 example of format”.

Next, it is determined whether or not the item extraction is selected (step 104). If the item extraction is selected, the sentence number currently held is temporarily saved (step 105), and the format of the sentence and the format determined from the specific example input at the time of inputting the condition are changed. It is determined whether they match (step 106). If a specific example is not specified when inputting a condition, some general formats set in the system are used. Also, in the bulleted list, since it is written in the same format divided into clauses,
The following sentence can be regarded as a bullet if it has only one sentence and the same format.

If the formats match, the sentence number counter is advanced to check the next sentence (step 107), and the counter is advanced and repeated until no more matches are found. When the formats no longer match, it is determined whether or not the formats match for a plurality of sentences (step 108). If the formats match, the extraction flags are set for all the matching sentences (step 109).
By the above operation, the sentence number counter indicates the next sentence to be processed irrespective of the presence / absence of the item extraction specification.

Next, when extraction of an item is not selected, the process proceeds to step 114 described later. When it is selected, it is compared with the format of the item to determine whether it matches (step 11).
1). If the format does not match the item format, the process proceeds to step 114 described later. If the format matches, the extraction flag of the matching sentence is set as the item name (step 112), and the paragraph number counter is advanced (step 113). Return to step 101.

On the other hand, if the paragraph consists of a plurality of sentences in step 102, a sentence having a specific word (keyword, katakana, abbreviation) specified by the generation condition input unit 12 (FIG. 1) is extracted. First, the first sentence of the paragraph is set in the sentence number counter (step 114), and it is determined whether there is a last sentence of the paragraph, that is, whether there is a remaining sentence that has not been processed (step 115). If there are any remaining sentences, set the first word of the sentence in the word number counter (step
116), it is determined whether or not there is the last word in the sentence, that is, the remaining unprocessed word (step 11).
7). Here, if there are any remaining words, it is further determined whether or not the word is an extraction target by comparing the attribute of the result of the dictionary extraction at the time of word extraction with the input condition (step 118). If it is a word, the extraction flag of the current sentence is set (step 119), and the sentence number counter is advanced (step 120). At this time, if the statement number has been saved in step 105, the statement number is advanced.

If it is the last word in step 117, the process directly proceeds to step 120, and the above operation is repeated until the last sentence of the paragraph is reached. Step 118
If the sentence word is not the word to be extracted in
Advance the word number counter (step 121), step 117
Return to Further, when there is no remaining sentence in the paragraph in step 115, the process proceeds to step 113. And
The process ends when the last paragraph is reached.

In this example, the keyword input by the user is extracted. However, if the attribute included in the dictionary unit 3 is replaced, it can be made to correspond to a word in a specific specialty field.

After the above-described object sentence extraction processing is completed, a summary sentence is generated. First, the original document name is stored in the result storage unit 17 (second
Write in enlarged characters at the beginning of (Fig.). Then, the presence / absence of an extraction flag is checked sequentially from the top of the extracted document, and if the flag is set, the result storage unit 17 is sequentially checked.
Write to. At this time, if the keyword portion is emphasized with bold characters or underlines, or if a line feed symbol or an appropriate tabulation is inserted before and after the item name sentence, the output result becomes easier to see. Finally, page assignment is performed, and a page break symbol is inserted to complete creation of the summary document. As a result of such processing, a summary sentence as shown in FIG. 8 is completed for the original sentence shown in FIG.

If necessary, the result can be edited further to obtain a more natural and easy-to-understand summary sentence. In addition, the standard office document system (OD
If a document that conforms to A), that is, a document that can handle a structured document, is used, the document has a logical structure, which makes it easy to extract chapters and sections and handle figures described in Application Example (3) described later. , More advanced processing becomes possible.

In the above-described embodiment, an example in which a summary sentence of a Japanese manuscript is created has been described. However, the document generation device according to the present invention is not limited to the above-described embodiment, and various applications are possible. . An example is shown below.

(1) Japanese-English summary sentence creation system In the above embodiment, a Japanese summary sentence was created. However, not only are item names and bullet points easy to extract, but the syntax is simple and semantic analysis is unnecessary, so it is relatively easy. There is a feature that it is easy to translate. Therefore, by utilizing this, a Japanese-English summary sentence creation system as shown in FIG. 9 can be realized. This system extracts items and the like from a Japanese original sentence input from a manuscript input unit 31 according to a generation condition in a Japanese sentence abstract sentence extraction unit 32, performs a semantic analysis in a syntax analysis unit 33, and translates it into English in a conversion unit 34. Thereafter, the English summary sentence generation unit 35 generates an English summary sentence and outputs it on the output unit 36. The output English summary can be used for Japanese-English translation support.

(2) Table-of-contents creation system When creating a document, it is often necessary to change the table of contents due to changes in chapter chapters or additions to the contents. Also, when a document is written by a plurality of writers, it is easy to create a table of contents. However, if the document generation device according to the present invention is applied, it is possible to realize a table of contents creating system which can easily create a table of contents automatically as shown in FIG. In the table of contents creating system, a specific line such as a chapter, a section name, a “reference”, and an “index” is extracted in the item extracting unit 42 from the document input by the original input unit 41. Next, in the information adding unit 43, numbering is sequentially performed on the specific line, a page number corresponding to the right side of the item is written, and after appropriately paragraphing, writing such as “table of contents” is performed at the beginning. . This is output by the output unit 44. The table of contents created in this way is
If it is output at the beginning of the processed document, it can be used for reports etc. just by attaching a cover.

(3) Presentation support system Currently, when giving presentations such as research presentations,
Although it is common to use overhead projectors (OHPs) and slides, creating the originals for these devices is troublesome, and the originals are automatically converted from the general documents at hand, such as when the originals are needed suddenly. It is very convenient to be able to create it. Therefore, the presentation manuscript mainly consists of each item name,
Focusing on the fact that the configuration is centered on keywords and charts, a presentation support system as shown in FIG. 11 can be realized. In the presentation support system, first, a sentence and a figure are extracted from a document input by an original input unit 51 by a sentence extracting unit 52 and a graphic extracting unit 53 according to the input conditions of the original. Next, in the layout section 54, 10 per sheet is used for easy viewing during projection.
The layout is enlarged so as to be on the order of the row, and the enlargement is performed immediately after the page where the graphic appears so that the corresponding graphic can be accommodated in one sheet. Then, the created form is printed out as a completed manuscript directly on OHP paper by a laser beam printer or the like of the output unit 55. The user removes the printed document and leaves it
You can give a presentation by OHP. If the data of the manuscript is electronically stored in a result storage unit (not shown), the created manuscript can be further edited, and the edited manuscript can be directly output.

In addition, recently, a transmission type image projection device that directly projects an output screen of a computer onto a screen has been developed, so that it is possible to give a presentation without a hard copy by combining with these devices. .

〔The invention's effect〕

As described above, in the document generation device according to the present invention, a specific sentence is extracted based on a generation condition input in advance, and a new document is generated based on a plurality of extracted sentences. The document contains a large amount of specific information and can be effectively used as a high value-added document for translation and understanding of a manuscript. In addition, since the document is created by relatively simple processing, the realization of the apparatus becomes easier as compared with a conventional machine translation system or the like.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a document generation device according to the present invention, FIG. 2 is a block diagram showing a basic configuration of the document generation device according to the present invention, and FIG. FIG. 4 is an explanatory diagram showing a display state on a screen of a display unit, FIG. 5 is a diagram showing an example of an original text, and FIG. 6 is a structure of a document. FIG. 7 is a flowchart showing a procedure for extracting a target sentence in the information processing unit, FIG. 8 is a diagram showing an example of a summary sentence, and FIG. 9 is a basic configuration of a Japanese-English summary sentence creation system FIG. 10 is a block diagram showing a basic configuration of a table of contents creating system, and FIG. 11 is a block diagram showing a basic configuration of a presentation support system. 10 information processing section, 11 original text storage section, 12 generation condition input section, 13 sentence extraction section, 14 document generation section, 15 original text input section, 16 dictionary section, 17 ... Result storage unit, 18 ... Original text scanning unit, 19 ... Display unit, 20 ... Printing unit.

 ────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yoshifumi Matsunaga 3-57-6 Yoyogi, Shibuya-ku, Tokyo Granforet Fuji Xerox Co., Ltd. (56) References JP-A-62-271171 (JP, A) JP-A-63-89966 (JP, A) JP-A-63-298563 (JP, A) JP-A-60-138670 (JP, A) JP-A-63-163925 (JP, A)

Claims (2)

(57) [Claims] 1. A storage device for storing a document, a selection device for sequentially selecting paragraphs of the document stored in the storage device, a first input device for inputting a predetermined format, and a first input device for inputting a keyword. (2) input means, first determining means for determining whether the paragraph of the document selected by the selecting means is formed by one sentence, and determining that the first determining means is formed by one sentence The second determining means for determining whether the character arrangement in the sentence matches the predetermined format; and the second determining means determines that the character sequence matches the predetermined format. A first extraction unit for extracting the sentence, and a third determination for determining whether the keyword exists in the paragraph when the first determination unit determines that the keyword is not formed in one sentence. Means, and the third determining means is a key If it is determined that the keyword exists, a second extraction unit for extracting one sentence including the keyword, and a document that sequentially holds the one sentence extracted by the first extraction unit and the second extraction unit are generated. A document generation device, comprising: a document generation unit. 2. The document generation apparatus according to claim 1, wherein said document is a structured document.