JPH05274351A - Keyboard extracting system - Google Patents

Abstract

(57) [Summary] [Purpose] To realize keyword extraction that does not depend on the character type and does not require an unnecessary word dictionary in the keyword dictionary. [Structure] The Japanese document input by the input means 1 is
The morpheme analysis unit 2 divides into words and gives a part of speech to each word. It is judged whether a keyword feature of a compound word base, proper noun constituent word, or prefix modification is given to the target part-of-speech such as a general noun, proper noun, or prefix, and keyword extraction is performed using information based on this keyword feature. I do.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a keyword extraction method, and more particularly to a keyword extraction method for automatically extracting a keyword from a document.

[0002]

2. Description of the Related Art As a publicly known document describing the prior art of the present invention, there is a "keyword extraction method" in Japanese Patent Laid-Open No. 3-8070, for example. In this publication, when extracting keywords from a document, only emphasized character strings are targeted, and it is determined that the targeted emphasized character strings are independent words by using a keyword dictionary. Is determined as a keyword. However, there are disadvantages that only the emphasized character strings in the document are targeted for keywords, and a keyword dictionary is required separately.

Also, "Automatic keyword extraction in newspaper article database" (Information Management vol.32 No.4, July 1
989) refers to the difference in the type of character, pinches the word, extracts the noun using a word table, etc., and deletes the terms recorded from the extracted noun using the unnecessary word table. Then, the keyword is decided. However, since the words are separated by the difference in the character type, it is not possible to accurately extract the keyword candidates, and there is a disadvantage that an unnecessary word dictionary is needed.

Further, the "automatic indexing system" proposed in Japanese Patent Laid-Open No. 2-148265 discloses that when extracting an index word from a document, first a morphological analysis is performed to extract a noun accompanied by a case particle. After deleting unnecessary words from the data using the unnecessary word dictionary, the remaining words are determined as index words. However, there is a drawback that the unnecessary word dictionary is still required.

[0005]

[Purpose] The present invention has been made in view of the above circumstances, and has as its object to provide a keyword extraction method that is not influenced by the character type and does not require a keyword dictionary or an unnecessary word dictionary. ..

[0006]

In order to achieve the above object, the present invention provides (1)
An input unit for inputting a Japanese document and a document input by the input unit are divided into word units, and a morphological analysis unit for giving a part of speech to the word is used to extract a keyword using the part of speech given by the morphological analysis unit. Extracting keyword from the document by using the part-of-speech information obtained as a result of the analysis by the morpheme analysis means and the keyword feature information, and (2) above. By using a compound word base that is one of the keyword features, unnecessary keyword candidates are reduced, and further, (3) by using a proper noun constituent word that is one of the keyword features, unnecessary By reducing the number of possible keyword candidates, and (4) using prefix modification, which is one of the keyword features,
The required prefix is characterized by being extracted as part of the keyword. Hereinafter, description will be given based on examples of the present invention.

FIG. 1 is a block diagram for explaining an embodiment of a keyword extraction system according to the present invention. In the figure, 1 is an input means, 2 is a morpheme analysis means, and 3 is a keyword extraction means. A Japanese document is input by the input unit 1, and the document input by the input unit 1 is divided into word units by the morphological analysis unit 2 and a part of speech is given to the word. The keyword extracting means 3 extracts keywords using the part of speech.

FIG. 2 is a flow chart for explaining the operation of the keyword extraction method according to the present invention. Less than,
The steps will be described in order. It is assumed that the sentence is divided into words in advance by morphological analysis.
First, the first word is input (step 1). Here, the counter is set to 0 (step 2). Further, it is judged whether the word is an unregistered word or proper noun (step 4), and if it is applicable, 1 is added to the counter (step 5). If it does not correspond, it returns to the point where one word is input again (step 1). If it is a general noun in step 3, then it is checked whether or not the general noun is given a keyword feature of either "compound word base" or "proper noun constituent word" (step 6). If not given, 1 is added to the counter as in the case of the previously unregistered word / proper noun (step 5). In short, if the input first word corresponds to any of the general noun or proper noun / unregistered word to which no keyword feature is given, 1 is added to the counter.

For all the remaining items, 1 is added to the counter (step 7), the stack is added (step 8), and the next word is input (step 9). Then, it is determined whether the input word is any of a proper noun, a general noun, a sahen noun, a prefix (of keyword feature “with prefix modification”), a suffix, and an unregistered word (step 10). .. If it is, the process of adding 1 to the counter again and stacking it on the stack is continued (steps 7, 8 and 9), but if it is not, the judgment based on the part of speech is terminated here.

The word (s) saved in the stack here
Is determined by whether or not the counter is greater than 1 (step 1
1). If it is larger than 1, the word stored in the stack until then is extracted as a compound word (step 12),
If not, the contents of the stack are cleared (step 13) and the process returns to the beginning. In short, a general noun to which a keyword feature is not assigned, a proper noun, or an unregistered word (those subjected to the process of adding 1 in step 5) alone are extracted as a compound word, but The assigned general noun / sahen noun / prefix / suffix will not be extracted as a compound word unless two or more words having those parts of speech are connected.

Now, the keyword features used in the above description will be described. Keyword features include compound word bases
There are three types of proper noun constituent words and prefix modification. Table 1 below summarizes the parts of speech, characteristics, and roles that can be given to each feature.

[0012]

[Table 1]

Next, a specific embodiment of the present invention will be described using an example. Example 1 Equipment is planned to be developed at this factory. When this sentence is morphologically analyzed, this / plant / in / device / development / development / plan / Part of the above-mentioned part of speech (that is, proper noun, general noun, sahen noun, prefix (with keyword feature "prefix modification"), suffix, or unregistered word. These words will be referred to as keyword candidates hereinafter) The following applies to. Factory (general noun, with feature "compound word base") Device (general noun, feature "compound word base") Development (sahenon) Schedule (sahenon) According to the previous flowchart, sahenon is also a common noun ( The counter is n =
Although it is 1, all of them have appeared independently. After all, since n <1, no keyword is extracted from this sentence.

Example 2 Here, a laser surveying instrument is developed.
I plan to follow. If you morphologically analyze this sentence, you can find here /// laser / survey / device / of / development / do / work / plan / da /. The following are candidates for keyword. Laser (no general noun / keyword feature) Surveying (sa noun) Device (general noun, with feature “compound word base”) Development (sa noun) Schedule (sa noun) According to the previous flowchart, “laser” is the keyword Since no feature is added, it can be a keyword even if it appears alone. However, in this example, “survey” and “device” follow. These two words cannot be keywords by themselves. However, because it is continuous from "laser", words are stacked on the stack one after another, and eventually the counter becomes n = 4, that is, n> 1, and from this sentence "laser surveying device" is extracted as a keyword. Become. Note that "development" and "scheduled" have not appeared as keywords since both have appeared independently.

Example 3 Development of a high-quality large-screen television here
Has begun. If you morphologically analyze this sentence, you can see here / / / high / image quality / na / large / screen / TV / of / development / started / started /. The following are candidates for keyword. High (with prefix, feature "prefix modifier") Image quality (with general noun, feature "compound base") Large (with prefix, feature "prefix modifier") Screen (generic noun, feature "compound base") Tsuki) Television (general noun, with a feature "compound base") According to the previous flowchart, "High" and "Large" are prefixes, but since the prefix modification of keyword features is added,
If a keyword candidate follows, both are extracted as a keyword. After all, "high quality" is n = 2, "large screen TV" is n
= 3, both of which are n> 1, "high image quality" and "large screen television" are extracted as keywords from this sentence. Note that "development" has not appeared as a keyword because it appears only by itself.

[0016]

As is apparent from the above description, the present invention has the following effects. (1) Effect corresponding to claim 1: A Japanese document is subjected to morpheme analysis, and keyword extraction is performed using the part-of-speech information obtained as a result, so it is not possible to extract only special characters (for example, emphasized characters). Instead, all character strings can be regarded as the target of the keyword. Moreover, since the keyword features are used in addition to the part-of-speech information, it is possible to accurately extract keywords with few unnecessary words and less necessary words. (2) Effect corresponding to claim 2: By using a compound word base that is one of the keyword features, unnecessary words that are not considered suitable as keywords can be extracted as little as possible. (3) Effect corresponding to claim 3: By using proper noun constituent words, which is one of the keyword features, unnecessary words that are not suitable as keywords can be extracted as little as possible. (4) Effect corresponding to claim 4: By using the prefix modification which is one of the keyword features, it is possible to extract the prefix necessary as a part of the keyword.

[Brief description of drawings]

FIG. 1 is a configuration diagram for explaining an embodiment of a keyword extraction system according to the present invention.

FIG. 2 is a flow chart for explaining the operation of the keyword extraction method according to the present invention.

[Explanation of symbols]

1 ... Input means, 2 ... Morphological analysis means, 3 ... Keyword extraction means.

Claims (4)

[Claims] 1. An input means for inputting a Japanese document, a morpheme analysis means for dividing a document input by the input means into word units, and giving a part of speech to the word, and a part of speech given by the morpheme analysis means. It is characterized by comprising a keyword extracting means for extracting a keyword by using the part-of-speech information obtained as a result of the analysis by the morpheme analyzing means, and the keyword feature information to extract the keyword from the document. Keyword extraction method to do. 2. The keyword extraction method according to claim 1, wherein unnecessary keyword candidates are reduced by also using a compound word base that is one of the keyword features. 3. The keyword extracting method according to claim 1, wherein unnecessary keyword candidates are reduced by using proper noun constituent words which are one of the keyword features. 4. The keyword extraction method according to claim 1, wherein the required prefix is extracted as a part of the keyword by using a prefix modification which is one of the keyword features.