JPH0229864A - Keyword extracting system - Google Patents

Abstract

PURPOSE:To extract a keyword from a diagram part by selecting constituting a diagram together with a graphic data or a table (ruled line) data and extracting a keyword from the selected character data. CONSTITUTION:A keyword extracting part 5 reads out document data from a document storage part 3 and searches character data constituting a diagram from a diagram corresponding part 6. The graphic corresponding part 6 executes display development to be executed by a document control part 2, finds out the coordinate position of the character data and finds out graphic data overlapped to the coordinate position. The keyword extracting part 5 extracts the character data of each group as a keyword, stores the extracted keyword in a keyword storage part 7 and displays the stored result on an output part 4. Thus, the keyword can be simply extracted from the diagram.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a keyword extraction method for extracting appropriate keywords from a document.

(Prior art) Conventional computer-based keyword extraction methods perform morphological analysis of character data that is part of document data, cut out words, eliminate words that are inappropriate for keywords (held in a dictionary), and if necessary, In other words, keywords were selected by the user.

By the way, the character data that makes up block diagrams, tables, etc. together with the figures and ruled line data is often in a different format from the character data that makes up the text, or has a different meaning in order to match the figure or table. Because the words are divided at odd places, the above method could not extract keywords from character data other than the original text.

Figures and tables are very important in terms of meaning to the entire document, as can be seen in the fact that even in papers, only the title, abstract, figures and tables are written in English, and translated journals are made for overseas markets. .

The keywords extracted from such diagrams and tables do not necessarily match the keywords extracted from the text. This tendency becomes stronger as there is less explanation for the figures and tables.

Therefore, if keywords cannot be extracted from the figures and tables, it does not mean that the keywords have been extracted from the entire document.

(Problem to be solved by the invention) As mentioned above, it is possible to analyze the character data that makes up charts and tables,
Extracting keywords has not been done in the past.

In order to solve these problems, the present invention provides a keyword extraction method that analyzes character data constituting charts and tables and extracts keywords.

(Structure of the Invention) (Means for Solving the Problems) The present invention includes an input section for inputting document data, instructing editing and keyword selection, and storing character data input from the input section. A document storage unit for analyzing document data and converting figures and tables to graphical data or tables (ruled lines)
It is characterized by comprising a keyword extracting section that selects character data constituting the data and extracts keywords therefrom, and an output section that displays the extracted keywords and character data.

(Function) According to the present invention, keywords can be extracted even from character data that has a format different from the original text or has different word cuts, so keywords can be extracted correctly from the entire document, so the extracted keywords can be used appropriately. The feeling increases.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of an embodiment of the present invention.

The input unit 1 is an input unit consisting of, for example, a mouse or a floppy disk driver, and is used to input document data and keyword selection instructions. Document data input from the input section 1 is given to a document management section 2 that controls the entire apparatus, and is displayed on an output section 4 as shown in FIG. 2, for example.

This data is stored in the document storage section 3.

The keyword extraction section 5 reads document data from the document storage section 3, and first, the chart correspondence section 6 searches for character data forming the chart.

Document storage unit 3 for graphic data, ruled line data, and character data
There are various storage formats.

The operation of the diagram corresponding section 6 will be explained using a format as shown in FIG. 3 as an example.

In the format shown in FIG. 3, character data is arranged in the − dimension. On the other hand, graphic data has individual types and coordinate positions regardless of this.

In order to display the data in FIG. 3 as shown in FIG. 2, the diagram correspondence section 6 performs the display development performed by the document management section 2, obtains the coordinate position of the character data, and obtains the graphic data that overlaps with it. , convert it into the format shown in Figure 4. The character data for “Display” is “Display” in the character data in Figure 3.
It was divided into "Dis" and "Play", but in Figure 4,
It is integrated into one character data surrounded by a rectangle. Character and graphic data is often stored in the format shown in FIG. 3 in older types of document processing devices, whereas in recent document processing devices, it is stored in the format shown in FIG. In this case, the chart corresponding section 6 as described above becomes unnecessary.

The keyword extraction section 5 extracts the character data of each group as a keyword from the storage format shown in FIG. For example, they are sorted and displayed.

At this time, if the user selects "FIG. 1" and "system configuration" from the input unit 1 and instructs to delete them, these will be removed from the keywords.

Terms related to figure/table numbers, such as ``1 Figure 1'', or terms related to table numbers, such as ``system configuration,'' are inappropriate as keywords. If such terms are added to the keyword extraction section 5 as an inappropriate word dictionary, these terms will be automatically removed from the keyword extraction.

Also, as mentioned above, if terms that are deemed inappropriate by the user's instructions are automatically added to the inappropriate word dictionary for each user,
The same term will no longer be extracted as a keyword from next time onwards.

Furthermore, if the keyword extraction unit 5 cuts words based on different character types such as hiragana, katakana, numbers, and alphabets, the keywords will be as shown in FIG. 5(b).

If a rule is added to the keyword extracting unit 5 that does not extract one-letter and two-letter words as keywords, "1", "Figure", and "Structure" will be automatically excluded.

In the above embodiments, the formats of graphic data and character data are described using the examples shown in FIGS. 3 and 4, but the formats are not necessarily limited to these.

It is also possible to add rules and dictionaries to improve the accuracy of keyword extraction in the keyword extraction section.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to easily extract keywords from charts and tables with important meaning, which have not been extracted conventionally, so the effect is great.

[Brief explanation of the drawing]

1 to 5 are diagrams for explaining a keyword extraction method according to an embodiment of the present invention. FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention, and FIGS. 1 is a diagram showing an example of a display screen, and FIGS. 3 and 4 are diagrams showing an example of a data storage format. 1... Input section, 2... Document management section, 3
...Document storage section, 4...Output section, 5...
Keyword extraction section, 6... Diagram correspondence section, 7... Keyword storage section. Agent Patent Attorney Nori Ken Ken Yudo Matsuyama Kosoku (Hage 1 Figure 2 Figure 1 C shape: (7 external shapes, -F, S 5 Z2. >)r\
, )) Figure 3 (Conspiracy) (Bage'1 Haruji 21-7t et al.) (4shi@%,: χI, yi) (fiM'l: -& Houji) Plzu:,)( Sei system:
Conspiracy〕 Sun (Female: Tisfu Ray) (Ri)shi-7 Otsu (Standing: Figure 1 Jimatam 60,000k)

Claims (3)

[Claims] (1) an input section for inputting a document consisting of characters, figures, tabular data, etc. and for instructing the selection of keywords; a document storage section for storing the document input from this input section; A keyword extraction section for analyzing a document and extracting text data constituting block diagrams, tables, etc. along with figure/table data as keywords, and keywords extracted by this keyword extraction section or input from the input section. A keyword extraction method characterized by comprising an output section for displaying and outputting a document. (2) The keyword extraction method according to claim 1, wherein the keyword extraction section includes a dictionary containing words and phrases that are not extracted as keywords. (3) The keyword extraction method according to claim 1, wherein the keyword extraction section performs word segmentation based on different character types such as katakana, hiragana, kanji, and alphabets.