JPH056398A - Document register and document retrieving device - Google Patents

Abstract

PURPOSE:To automatically extract a keyword without using excess data such as a dictionary by providing this document registering/retrieving device with a document coding means, a document registering means and a document index storing means for storing corresponding relation between a superimposed code and a document. CONSTITUTION:The document coding means 1 converts a document to be registered or a keyword applied to the document into a superimposed code. When the document is an object to be registered, the keyword is automatically allocated and manual keyword application is unnecessary. The document registering means 2 allows a bit string obtained by the means 1 to correspond to the document and stores the correspondence in the document index storing means 3. The means 3 correspondingly stores the bit string and the document. Since the keyword is automatically allocated, labor for registering the document can be reduced, and since large data such as a word dictionary are not used, the number of memories or disks to be driven can be reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document registration device and a document search device which automatically save a document with a keyword attached thereto and retrieve the corresponding document by inputting the keyword.

[0002]

2. Description of the Related Art Conventionally, in the case of filing an electronic document, it is generally performed that a keyword is added and registered, and then the keyword is designated to retrieve the corresponding document. In this case, how to "grant" the keyword and how to "search" using the keyword
There is the question of whether to do it.

[0003] First of all, a human is generally given a keyword, but there is also a system in which a keyword is automatically given from the contents of a document. That is, a method is used in which all nouns are extracted from a document using a word dictionary, and words that have been previously determined not to be keywords called unnecessary words are removed from the nouns. Further, a search by a keyword is generally performed by using a file called a transposed file in which a document is associated with a keyword.

As another method, for example, Japanese Unexamined Patent Publication (Kokai) 2
There is also a method of using a superposed code, as disclosed as a “dictionary lookup device” in Japanese Patent Publication No. 297193. That is, this is a method in which a keyword is hashed into a bit string of a specific length, and the bit string obtained by logically adding the bit strings of all the keywords corresponding to one document is stored in association with the document. Also,
At the time of a search, a search keyword is hashed into a bit string of a specific length, the bit string is used as a search key, and a document whose logical product with the bit string corresponding to the document is equal to the search key is searched. Even when a plurality of search keywords are specified, if the logical sum of each bit string is used as the search key, the search can be performed easily.

[0005]

In the above-described conventional keyword assignment and retrieval, the following problems occur. First, the first problem is that large-scale dictionary data is required when automatically adding keywords. That is, in the conventional method, large-scale data such as a word dictionary and a grammar dictionary are required for automatic keyword extraction. This increases the amount of memory or disk required to operate the system and slows down the processing speed. In addition, there is a problem that enormous effort is required to create and maintain dictionary data, and the cost for creating the system is high.

The second problem is that the keyword cannot be searched unless the same keyword as the given keyword is input. That is, the conventional method has a problem in that the added keyword is used as it is as an index, so that it is impossible to search with a slightly different keyword. As an example, a document registered by adding the keyword “information device” cannot be searched for by a keyword such as “information search” or “search device”.

[0007]

According to a first aspect of the present invention, a document or an assigned keyword is delimited by a change point of a character type, and all consecutive two character groups in a character string of each section are superposed. A document encoding means for converting into a paused code is provided, a document registration means for registering a correspondence relationship between the superposed code and the document is provided, and a correspondence relationship between the superposed code and the document is held. A document index holding means is provided.

According to a second aspect of the invention, in the first aspect of the invention, the document encoding means does not encode the hiragana section.

According to the third aspect of the present invention, the keyword encoding means is provided for delimiting the keywords at the changing points of the character type and converting all two consecutive character sets in the character string of each section into a superposed code. And a document search means for retrieving a document corresponding to the superposed code, and a document index holding means for holding a correspondence relationship between the superposed code and the document.

According to a fourth aspect of the invention, in the third aspect of the invention, the keyword encoding means does not encode the hiragana section.

[0011]

According to the first aspect of the present invention, since the keywords are dynamically added, the labor of document registration can be reduced, and since a large-scale data such as a word dictionary is not used, a memory that requires an operation is required. And less disks are required.

According to the second aspect of the present invention, since the hiragana character string that occupies a large part of the sentence is not processed, the processing speed is further increased, and it is not necessary to use frequently used characters for hashing. Therefore, the design of the hash function becomes easy.

According to the third aspect of the present invention, the keyword is decomposed and the search is performed, and the search is possible even if the keyword is not the same as the added keyword.

According to the fourth aspect of the present invention, since the hiragana character string that occupies a large part of the sentence is not targeted for processing, the processing speed is further increased, and the frequently used characters are not targeted for hashing. Therefore, the hash function can be easily designed.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows the overall configuration of this apparatus, which comprises a document encoding means 1, a document registration means 2, a document index holding means 3, a keyword encoding means 4, and a document search means 5. . 2 shows the flow of processing at the time of document registration, and FIG. 3 shows the flow of processing at the time of document search. The configuration of each unit will be sequentially described below with reference to these flows. .

First, the document encoding means 1 will be described. This is to convert a document to be registered or a keyword given to the document into a superimposed code. In this case, when a document is targeted, a keyword is automatically added, and it is not necessary for a human to add the keyword. Since the operation is the same for both of them, a method of encoding a document will be described here.
That is, as described below, each step is sequentially taken.

As [step 1], the document is divided at the character type change points. The change point of the character type is such a point that it becomes hiragana → kanji or alphabet → katakana. For example, the document “I developed an information retrieval device yesterday”
It can be divided into "/ yes / information retrieval device / developed / developed /". Subsequent operations are performed in units of divided sections. However, no processing is performed on the section consisting of hiragana.

As [step 2], all sets of two characters are extracted from the character string in the section. For example, from the section "information retrieval device", five types of two-character sets "information", "inspection", "search", "search" and "device" can be extracted.

As [step 3], the character code of the two-character set is used as a key and converted into a bit string of a predetermined length by a predetermined hash function. At this time, the hash function is set so that the number of 1s in the converted bit string is the same. For example, if the hash function is defined to convert 4 bits of a bit string of length 32,

[0020]

[Table 1]

(The actual bit string that can be obtained depends on the design of the hash function).

As [step 4], the logical sum of all the bit strings obtained in step 3 is calculated. In the example above,
The next bit string is obtained by taking the logical sum of the five bit strings.

[0023]

[Table 2]

As [step 5], the logical sum of all bit strings corresponding to each section is calculated. This is the superposed code corresponding to the document.

Next, the document registration means 2 will be described.
This is to store the bit string obtained by the document encoding means 1 and the document in the document index holding means 3 in association with each other.

Next, the document index holding means 3 will be described. This stores a bit string and a document in association with each other. For example:

[0027]

[Table 3]

Next, the keyword encoding means 4 will be described. This converts a keyword specified as a search key into a superposed code.
In this case, the operation is exactly the same as the document encoding means 1. If a plurality of keywords are specified for the AND search, the logical sum of the superimposed code for each keyword becomes the code corresponding to the search key. One example is “information retrieval”
The search keyword is encoded as follows.

[0029]

[Table 4]

Finally, the document retrieval means 5 will be described. This is to search the document index holding unit 3 for a document that matches the search key obtained by the keyword encoding unit 4. That is, the following steps are taken.

As [step 1], the logical product of the superposed code and the search key corresponding to each document is calculated.

As [step 2], it is determined that the document matches the search key if the logical product obtained in step 1 is equal to the search key, and does not match if they are not equal.

For example, if the search key is

[0034]

[Table 5]

, The logical product 1 of the "information search" and the document 1 is

[0036]

[Table 6]

And is equal to the search key. Therefore, it is determined that the document 1 matches the search key.

Further, the logical product 2 of "information retrieval" and document 2
Is

[0039]

[Table 7]

And is different from the search key. Therefore, it is determined that the document 2 does not match the search key.

[0041]

According to the first aspect of the present invention, a document or a given keyword is separated by a change point of a character type and all consecutive two character sets in a character string of each section are made into a superposed code. A document index holding means for holding a correspondence relationship between the superposed code and the document is provided with a document encoding means for converting, and a document registration means for registering the correspondence relationship between the superposed code and the document. Since a means is provided, keywords can be dynamically added to reduce the trouble of document registration, and since large-scale data such as word dictionaries are not used, less memory and disk are required for operation. ,further,
Since it is a simple operation, the processing can be performed at high speed.

According to a second aspect of the present invention, in the first aspect of the present invention, the document encoding means does not encode the hiragana section, so that the hiragana character string that occupies a large part of the sentence is processed. Since it does not, the processing speed is further increased, and since the frequently used characters do not have to be the object of hashing, the hash function can be easily designed.

According to a third aspect of the present invention, the keyword encoding means divides a keyword at a character type change point, and converts all consecutive two character sets in a character string of each section into a superimposed code. Is provided, and a document search means for retrieving a document corresponding to the superposed code is provided, and a document index holding means for holding a correspondence relationship between the superposed code and the document is provided. Will be searched
It is possible to search even if it does not have the same shape as the given keyword.

In the invention according to claim 4, in the invention according to claim 3, since the keyword encoding means does not encode the hiragana section, the hiragana character string occupying a large part of the sentence is targeted for processing. The processing speed is further increased because it is not necessary, and the hash function can be easily designed because the frequently used characters do not have to be the object of hashing. Furthermore, since the particle and the verb ending are hiragana, the “information retrieval device A keyword entered in the form of a phrase such as "can be searched without special operation by removing the hiragana.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a flow of processing at the time of document registration.

FIG. 3 is a flowchart showing the flow of processing at the time of document search.

[Explanation of symbols]

1 Document encoding means 2 Document registration means 3 Document index holding means 4 Keyword Encoding Means 5 Document search means

Claims (4)

[Claims] 1. A document encoding means for separating a document or an attached keyword at a change point of a character type and converting all consecutive two character sets in a character string of each section into a superimposed code. Document registration means for registering a correspondence relationship between the superposed code and the document;
A document registration apparatus comprising: a document index holding unit that holds a correspondence relationship between the superposed code and the document. 2. The document registration device according to claim 1, wherein the document encoding means does not encode the hiragana section. 3. A keyword encoding means for dividing a keyword at a character type change point and converting all consecutive two character sets in a character string of each section into a superimposed code, and the superimposed code. A document retrieval device comprising document retrieval means for retrieving a document corresponding to a code, and document index retaining means for retaining a correspondence relationship between the superposed code and the document. 4. The document retrieval device according to claim 3, wherein the keyword encoding means does not encode the hiragana section.