JPS63217418A - System for extracting japanese key word text - Google Patents

Abstract

PURPOSE:To extract a key word in a consistent extraction method with no intervention of man power by deleting such character strings that cannot give features of a text with reference to a non-key word file and extracting the character strings having high emerging frequency table. CONSTITUTION:When a Japanese word text data is converted into codes, a character extracting means 15 extracts a character string containing at least KANJI (Chinese character) or KATAKANA (square form of Japanese syllabary) out of a text based on a range of KANJI codes or KATAKANA codes. Then a key word extracting means 18 decides from the information on a non-key word file 16 whether said extracted character string is generally improper or not to give features to a text. Then the emerging frequency of a character string which is not decided improper is obtained from a character string emerging frequency table 17. Then the character string having a high emerging frequency is decided as a key word of the text. Thus it is possible to extract a key word of the Japanese word text data with no intervention of man power.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a Japanese text keyword extraction method.

[Conventional technology]

Conventionally, Japanese text keywords have been extracted by manually reading through the Japanese text and manually extracting keywords that characterize the text.

[Problem that the invention seeks to solve]

In the conventional manual Japanese text keyword extraction method described above, the text was read through manually and the keywords were extracted manually, so it took a lot of time to extract one keyword, and the extraction of keywords was difficult for each individual. The disadvantage is that it is difficult to set homogeneous keywords when multiple people are asked.

[Means for solving problems]

The Japanese text keyword extraction method of the present invention includes a character/code conversion means for converting Japanese characters into codes, and a coded data of the Japanese text generated by the character/code conversion means. a character string extracting means for extracting a character string containing at least one of the characters; and a non-keyword storing information for excluding character strings that do not characterize the text from the character strings extracted by the character string extracting means. From the file, a character string appearance number table representing each character string and its number of occurrences in the Japanese text as an image, and the character strings extracted by the character string extraction means, the character strings in the non-keyword file are While referring to the information, exclude strings that do not characterize the text, and for strings that are not excluded,
The character extraction means is created for each Japanese text, and keyword extraction means is provided for extracting only characters that appear frequently from the created table as keywords.

[Operation] When encoding Japanese dex data, the character extraction means extracts character strings containing at least one of kanji and katakana from the text based on the kanji code range and the katakana code range, and the keyword extraction means However, based on the information in the non-keyword file, it is determined whether the string is inappropriate for characterizing the text in general, and the number of occurrences of strings that are not judged to be inappropriate is calculated using a string appearance number table. The keywords that appear most often are used as keywords for the text.

Therefore, keywords from Japanese text data can be extracted without manual intervention.

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the Japanese text keyword extraction method of the present invention, FIG. 2 is a diagram showing an example of the information format of the non-keyword file 16 in FIG. 1, and FIG.
4 and 5 are flowcharts showing the processing of the character string extracting means 15, keyword extracting means 18, and database updating means 19 in FIG. 1, respectively.

As shown in FIG. 1, this embodiment includes an input means 11, a display device 12, a Japanese text creation means 13, a Japanese text file 14, a character string extraction means 15, a non-keyword file 16, and a character string appearance number table. 17, keyword extracting means 18, database updating means 19, and keyword base 20.

The input means 11 is an input means for the Japanese text creation means 13 and the database updating means 19.

The display device 12 includes an input means 11 and a keyword extraction means 1.
This is a device that displays information from 8. The Japanese text creation means 13 is means for generating encoded data from the character string information of the text input from the input means 11.

The Japanese text file 14 is a file in which encoded text information generated by the Japanese text creation means 13 is registered. The character string extraction means 15 is a means for extracting character strings including kanji and katakana from text data, based on the kanji code range and katakana code range that it owns. Even if the non-keyword file 16 contains character strings including kanji and katakana,
This is a file that lists kanji and katakana character strings that do not characterize the text, in order to determine keywords that generally do not characterize the text, such as ``thought'', and this information is shown in Figure 2. is stored. As shown in Table 1, the character string appearance number table 17 is a table that expresses character strings including kanji and katakana in the target Japanese text and the number of times they appear.

The keyword extracting means 18 extracts, based on the non-keyword file 16, out of all the character strings consisting of kanji and katakana in the text extracted by the character string extracting means 15, those that do not generally characterize the text. The number of occurrences of the same keyword, which has been reduced to 15 by excluding, is determined using a character string appearance number display 17, and the keywords with the highest number of occurrences are displayed on the display device 12. Database update means 1
9, from among the keywords displayed on the display Ti 12 by the keyword extracting means 18, the user determines that the keywords do not generally characterize the text, and adds the manually entered keywords from the input device 11 to the non-keyword file 16 as non-keywords. However, the keywords that are not set as non-keywords above are registered in the keyword database 20. As shown in Table 2, keywords are registered in the keyword database 20 along with the text number of each text data in the Japanese text file 14.

Table 2 Next, FIG.
This will be explained with reference to FIGS. 4 and 5.

(1) Character string extraction means 15 (Figure 3) First, one character of coded text is inputted by the Japanese text creation means 13 (step 31),
Determine whether the text data is finished (Step 3)
2). If the text data is not finished, it is determined whether the character currently being examined is a kanji or katakana (step 33). If either kanji or katakana is in progress, if you are creating a string of keyword candidates, add the character you are currently looking for to the end of the string you are creating, and if you are currently creating a string of keyword candidates. If not, one new character is set as a keyword candidate in the character string being created, and the process jumps to step 31 (step 34). If it is neither kanji nor katakana, it is determined whether there is a character string that is currently being created that includes kanji or katakana (step 35). If there is a character string that is currently being created that includes kanji or katakana, use the created character string as one of the character strings that will be input to the keyword extraction means 18, leave the character string that is currently being created empty,
Jump to step 31 (step 36).

When it is determined in step 32 that the text data has ended,
The process ends (step 37).

(2) Keyword extraction means 18 (Figure 4) First, after all character strings including kanji and katakana in the text are extracted by the character string extraction means 15, the non-keyword file 16
and all kanji extracted from this text,
Step 41) To extract only those katakana character strings that are not in the non-keyword file 16, create a string appearance count 17 for the extracted character strings, and update the appearance count value Step 42 ), a character string whose number of appearances is greater than a preset specified number of times is extracted as a keyword (step 43), the extracted keyword is displayed on the display device 12 (step 44), and the process ends.

(3) Database update means 19 (FIG. 5).

First, the keyword extracting means 18 refers to the non-keyword file 16 and the number of character appearances 17, and after completing the extraction of keywords, the keywords extracted by the keyword extracting means 18 are inputted from the user through the input means 11 into text. In step 51), the specified keyword is deleted as a keyword that does not have a characteristic, and in step 52), the keyword that has not been deleted is registered in the keyword database 2o.In step 51, the new text specified from the input means 11 is Add character strings that cannot be characterized to the non-keyword file 16 (step 53)
, ends the process.

Next, the overall operation of this embodiment will be specifically explained.

For example, input means 11, display device 12, Japanese text creation means 13, Japanese text file 14, character string extraction means 15, non-keyword file 16, character string appearance count 17, keyword extraction means 18, database updating means 19 and keyword database 20, respectively, in a CRT monitor, a Japanese text creation system that is a computer program, a character string extraction system that is a computer program, a non-keyword file using a magnetic disk as a recording medium, and a main memory of a computer. A keyword extraction system that is a computer program, a database update system that is a computer program, and a keyword database that uses a magnetic disk as a storage medium.
do.

When registering Japanese text data together with a text data number into a Japanese text file using a keyboard, the Japanese text creation system converts the input characters into codes. At this time, the string extraction system
The converted code is manually input and character strings containing kanji and katakana are extracted using the process shown in FIG. 3 described in F. From the extracted character strings containing kanji and katakana, the keyword extraction system uses the process shown in Figure 4 above to exclude those that do not generally characterize the text, and calculates the number of occurrences of the same character string.
The number of occurrences of character strings is counted by creating a record, and those with the highest number of occurrences are displayed on a CRT monitor. Based on the input from the keyboard, the database update system deletes the character strings displayed on the CRT monitor that are determined not to be characterized by the text according to the process shown in FIG. 5 described above, and updates the keyword database. Character strings that are not judged to be inappropriate are registered together with the text number of the Japanese text, and character strings that are judged not to generally characterize the text are registered in a non-keyword file.

〔Effect of the invention〕

As explained above, the present invention extracts character strings containing kanji and katakana from Japanese text according to the range of kanji codes and the range of katakana codes, and extracts character strings that do not characterize the text from the extracted character strings. By referring to the non-keyword file and deleting it, and extracting the frequently occurring keywords as keywords using the character string appearance count, it is possible to create keywords for Japanese text data without the need for human intervention. This has the effect that keywords can be extracted using iL and other extraction methods.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an embodiment of the Japanese text keyword extraction method of the present invention, FIG. 2 is a diagram showing an example of the expression format of information in the non-keyword file 16 in FIG. 1, and FIG.
4 and 5 are flowcharts showing the processing of the character string extraction means 15, the keyword extraction means 18, and the database updating means 19 in FIG. 1, respectively. 11... Input means, 12... Display device, 13.
...Japanese text creation means, 14...Japanese text file, 15...Character string extraction means, 16...Non-keyword file, 17...Character string appearance count, 18...Keyword extraction Means, 19...Database update means, 20...Keyword database. Figure 4 Figure 5

Claims (1)

[Claims] Character/code converting means for converting Japanese characters into codes; and characters containing at least one of kanji and katakana from coded data of Japanese text generated by the character/code converting means. A character string extraction means for extracting a string, and a non-keyword file storing information for excluding character strings that do not characterize the text from the character strings extracted by the character string extraction means. The text is extracted from the character string appearance count table representing each character string and its number of occurrences in the Japanese text, and the character strings extracted by the character string extraction means, while referring to the information in the non-keyword file. Exclude strings that cannot be characterized, and for strings that are not excluded,
A Japanese text keyword extraction method, comprising: a keyword extraction means for creating the character string appearance table for each Japanese text, and extracting only keywords that appear frequently from the created table as keywords.