JPH0391077A - Keyword index generating system - Google Patents

Abstract

PURPOSE:To improve the retrieving rate of a desired sentence by reading out a word showing the abstract conception of a selected word and a word showing the abstract conception of a selected word and a word showing a conception, to which a name belongs, and outputting those words to a keyword index update part together with the selected words. CONSTITUTION:A word development part 16 reads out the word showing the abstract conception of the word selected by a word selection part 12, and the word showing the conception, to which the name belongs, from a storing means 15 and outputs those words to a keyword index update part 13 together with the selected words. Thus, the extremely concrete conception is developed from the selected word to the word showing the more abstract conception or developed from the name to the word showing the conception and not only the original word but also these developed word groups can be attached as the keyword of a record unit text and registered to a keyword index 14. Thus, even the word showing the conception not existent in the record unit text of the sentence can be registered to the keyword index as well and the retrieving rate of the desired sentence can be more improved.

Description

[Detailed Description of the Invention] (Summary) The purpose of this invention is to further improve the retrieval rate of a desired text with respect to a keyword index generation method used to search for a desired text from a text base. In the keyword index generation method, a word division means divides the word into words, a word sorting unit selects only words that can become keywords from the divided words, and the keyword index updating unit registers the words in the keyword arch. a storage means that stores in advance a group of words representing a more abstract concept of each possible word and a concept to which the name belongs; and a selection method from the storage means based on the words selected by the word selection section. The present invention is configured to include a word expansion section that reads out words indicating the abstract concept of the selected word and words indicating the concept to which the name belongs, and outputs them together with the selected words to the keyword index updating section.

(Industrial Application Field) The present invention relates to a keyword index generation method, and more particularly to a keyword index generation method used to search for a desired sentence from a text base.

(Prior Art) A keyword search method is known as a method for searching for a desired sentence from a text base in which a large amount of sentences are recorded in units of text (for example, one sentence or paragraph). This keyword search method adds a keyword group representing the content to each recording unit text of a sentence, and constructs an index that provides mapping from such keywords to each recording unit text. A desired sentence is searched using a corresponding keyword.

However, manually adding such keywords to a large amount of text requires a huge amount of man-hours, so traditionally, independent words were extracted from word groups obtained by word segmentation using a computer and dictionary. Method of extracting keywords (
A free-term system is in place.

[Problem to be solved by the invention]

In the above-mentioned free-term method, words that appear in a sentence are used as keywords, so there is a strong tendency for keywords to be the names of things (proper nouns) or words that represent very specific concepts. On the other hand, from the standpoint of text-based searchers, the names of objects and words expressing concrete concepts are often unknown, so words expressing abstract concepts are input as keywords into text-based search devices. There is a tendency to

Therefore, when using a keyword index conventionally generated by the free-term method, it is necessary for the searcher to be familiar with the field of text that is the target of the search, or to search for more specific keywords using a concept system (thesaurus).
The disadvantage is that there is a high possibility that the desired text will not be retrieved unless it is searched by .

The present invention has been made in view of the above points, and an object of the present invention is to provide a keyword index generation method that can further improve the search rate for desired sentences.

[Means to solve the problem]

FIG. 1 shows a block diagram of the principle of the present invention. In the present invention, a recorded unit text of a sentence is divided into words by a word division means 11, and from the divided words, a word selection unit 12 selects only words that can become keywords, and a keyword index update unit 13 creates a keyword index 14. This is a keyword index generation method for registering a keyword index, which is equipped with a storage means 15 and a word expansion section 16.

Here, the storage means 15 stores in advance, for each word that can be a keyword, a group of words representing a more abstract concept of the word and a concept to which the name belongs.

In addition, the one-word expansion unit 16 reads out from the storage unit 15 words indicating the abstract concept of the word selected by the word selection unit 12 and words indicating the concept to which the name belongs, and uses them together with the selected words in the keyword index. It is output to the update unit 13.

[Effect]

In the present invention, the word groups divided in the same manner as the conventional 7-term keyword index generation method are used in a concept classification system prepared in advance in the storage means 15, as well as concepts to which objects with various names belong. Knowledge is used by the word processor 16 to expand from selected words representing very specific concepts to words representing more abstract concepts, or from names to words representing concepts. Not only the words , but also these expanded word groups are added as keywords of the recording unit text and registered in the keyword index 14 .

Therefore, according to the present invention, words representing concepts that do not exist in the recording unit text of a sentence can also be registered in the keyword index 14.

[Embodiment] FIG. 2 shows a block diagram of a text retrieval device to which an embodiment of the system of the present invention is applied. In the figure, the same components as in FIG. 1 are given the same reference numerals. In FIG. 2, 17 is a word division unit, 18 is a word dictionary, and these are connected to the word division means 11.
It consists of Reference numeral 19 is conceptual classification system knowledge, in which words representing more abstract concepts than the input words are stored in advance.

Reference numeral 20 denotes name-concept related knowledge, in which words representing the concept to which the name of the input word belongs are stored in advance. These concept classification system knowledge 19 and name-concept related knowledge 120 constitute the storage means 15.

Further, 21 is a text input section, 22 is a text registration section, 23 is a text base, and 24 is a keyword search device.

Next, the operation of this embodiment will be explained. The text to be searched is divided into appropriate recording unit texts by the text input section 21. The divided recording unit text is registered in the text base by the text registration section 22, and is converted into a word string by the word division section 17 using the word dictionary 18.

This word string is input to the word selection section 16 after the word selection section 12 selects only words that can be used as keywords. The word development unit 16 uses concept classification system knowledge 19 and name-concept related knowledge 20 to find abstract words or words representing the concept to which the name belongs from the words (keywords) selected and input by the word selection unit 12. , these obtained word groups (R-opened word groups) are sent to the keyword index updating unit 13 together with the original words (keywords).

The keyword index update unit 13 updates and registers the input word group in the keyword index 14. Keyword search device 24
is operated by a text-based searcher to search the keyword index 14 for a keyword specified by the searcher, and read out the recording unit text corresponding to the keyword from the text base 23. Here, keyword index 1
4, as mentioned above, not only words in the recording unit text but also words representing concepts that do not exist in the recording unit text are generated and registered as keywords.
Even if the keyword entered by the searcher does not exist in the desired recording unit text, the desired recording unit text can be searched if it is a word that represents an abstract concept or a concept to which the name belongs in the word in the recording unit aquist. can do.

Therefore, the present embodiment can improve the retrieval rate of a desired recording unit text more than the conventional method.

Next, the text search operation according to this embodiment will be explained in more detail. For example, write a newspaper article headline using Acistbase 2.
3, it is assumed that the recording unit texts divided by the text input section 21 are as shown in FIG.

This recording unit text is processed by the word dividing unit 17. Vocabulary 118
When the word expansion unit 16 performs word division and word selection in the same manner as in the prior art, the selection results are as shown in FIG.

The present embodiment further uses hypernyms as shown in FIG.

The concept classification system knowledge 19 in which related words are classified by word representing the concept, and the name concept related knowledge 20 in which names and 41 categories representing the concepts to which the names belong, as shown in FIG. 6, are stored. A word expansion section 16 performs word expansion. As a result, from "Fujitsu" in Figure 4 to "Computer Manufacturer" as shown in Figure 6, (7) rFM-
TOWNS, from 1 to the name of Figure 6 - concept related knowledge 20
"PC" in Figure 4 is "MPU" in Figure 5.
From the concept classification system knowledge 19 in the figure, rLsIJ and "computer" are respectively expanded. Therefore, the word expansion section 16 obtains a word group as shown in FIG. 7 as a keyword.

These word groups are registered in the keyword index 14 by the keyword index updating unit 13 as shown in FIG. 8(B). The keyword index 14 has pointer information to the recording unit text stored in the text base 23. Therefore, the recording unit text shown in Figure 3 is changed to Figure 8 (
When stored in the position of number 10206 in the text base 23 as shown in A), the word group (keyword group) in Fig. 7 is stored in the same text number 1 as shown in Fig. 8 (B).
It is registered in the keyword index 14 along with 0206.

The keyword search device 24 uses the keyword index 14 to obtain a recording unit text corresponding to the keyword given by the searcher. As a result, according to this embodiment, when searching the recording unit text shown in FIG.
``Computer manufacturers'' that could not be searched using conventional methods.

You can also search by entering any of "PC J, rLsIJ," and "Computer" as keywords.

(Effects of the Invention) As described above, according to the present invention, words representing concepts that do not exist in the recording unit text are also registered in the keyword index, so the search rate is improved compared to keyword index generation using the free term method. This feature has the advantage of being able to reduce search omissions.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the principle of the present invention. FIG. 2 is a block diagram of a text retrieval device to which an embodiment of the present invention is applied. FIG. 3 is a diagram showing an example of recording unit text cut out by the text input section. , Figure 4 is a diagram showing an example of the results of word segmentation and word selection, Figure 5 is a diagram showing an example of the content of concept classification system knowledge, Figure 6 is a diagram showing an example of the content of name-concept related knowledge, FIG. 7 is a diagram showing an example of a keyword group obtained by the word expansion section, and FIG. 8 is a diagram showing an example of the contents of the text base and keyword index in FIG. 2. In the figure, 11 is word division means, 12 is word selection section, 13 is keyword index update section, 14 is keyword index, 15 is storage means, 16 is word expansion section, 19 is concept classification system knowledge, 20 is name-concept Demonstrate relevant knowledge.

Claims (1)

[Claims] A recording unit text of a sentence is divided into words by a word division means (11), and a word selection unit (11) selects words from among the divided words.
12) selects only words that can be used as keywords, and the keyword index update unit (13) creates a keyword index (
14) In the keyword index generation method registered in 14), for each word that can be a keyword, a storage means (15) that stores in advance a group of words representing a more abstract concept of that word and a concept to which the name belongs; Based on the words selected by the word selection section (12), words representing the abstract concept of the selected words and words representing the concept to which the name belongs are read out from the storage means (15), and these are read out from the storage means (15), and words representing the concept to which the name belongs are read out from the storage means (15). A keyword index generation method, comprising: a word expansion unit (16) that outputs the words together with the keyword index update unit (13).