JPH06162092A - Information retrieval device - Google Patents

Abstract

PURPOSE:To reduce the labor of maintenance of dictionary, to reduce index capacity and to retrieve text data by retrieving the corresponding character string from a text by means of n-character index and word index. CONSTITUTION:A word division processing section 6 divides the text into words and a word index preparation section 8 prepares the word index 3 linking the text while taking the word as a word index. An n-character index preparation section 9 prepares an n-character index 4 to be linked from an n-character index taken out from the start on the word index. In this case, an unregistered word processing section 7 makes a word index on the character string unregistered in a word dictionary 12 in a batch based on the character information. According to the retrieval direction with a keyword specified, the corresponding word index of a word index 3 to be linked is found from the n-character from the corresponding n-character index of an n-character index 4. Then the corresponding character string in the text 2 is retrieved and outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval device for retrieving documents. In recent years, as a large amount of electronic documents have become available, a search technique for extracting necessary information from these documents is required.

Searches can be sped up by having an index file. Minimizing this index file is required to build a search system. On the other hand, there is a demand to search for any character string regardless of whether or not the character string is contained in the index file. In these cases, it is also required that there is little maintenance work on word dictionaries.

[0003]

2. Description of the Related Art Conventionally, the following has been used as a method for searching a word in a document. (1) Keyword index method using morphological analysis: This is because a keyword extracted by performing morphological analysis on a document is indexed, so that an index file can be small. Moreover, since each keyword has a dictionary attribute or the like, it is possible to perform processing using knowledge of words. However, it is necessary to maintain a word dictionary. Further, if the unregistered word extraction fails, the character string of this unregistered word cannot be searched.

(2) n-character index method: Since all characters in a document are indexed, knowledge of words in a dictionary or the like is not required, and as a result, dictionary maintenance costs can be reduced. Further, since all the appearing character strings are the search targets, there is no omission in the search. However, the amount of indexes becomes huge.

[0005]

The keyword index method based on the morphological analysis using the word dictionary of (1) described above requires maintenance of the word dictionary in order to maintain accuracy, and the labor thereof is great. There is a problem that it is difficult. In addition, if a transposed file in units of n characters (for example, in units of 1 character) is created by the n-character index method of (2) described above, the amount of index increases and becomes larger than the original text. There is a huge problem.

The present invention solves these problems.
The purpose is to reduce the maintenance work of the dictionary, reduce the index capacity, and enable text data search.

[0007]

Means for solving the problems will be described with reference to FIGS. 1 and 2. FIG. 1 and 2, the word division processing unit 6 divides the text into words to form words.

The unregistered word processing unit 7 collects character strings not registered in the word dictionary 12 by character type information or the like. The word index creation unit 8 creates a word index 3 that links a corresponding position in the text with the word obtained by dividing the text as a word heading.

The n-character index creation unit 9 links the word index of the word index 3 with the n-character index 4 that is linked from the n-character index extracted from the beginning or the like.
Is to create.

[0010]

According to the present invention, as shown in FIGS. 1 and 2, the word division processing unit 6 divides the text into words, and the word index creation unit 8 links the texts using the words as word headings. 3 is created, and the n-character index creating unit 9 creates an n-character index 4 that links these word headings from the n-character headings extracted from the beginning or the like. On this occasion,
The unregistered word processing unit 7 collects the character strings that are not registered in the word dictionary 12 based on the character type information to form a word heading.

Further, in response to the search instruction specifying the keyword, the corresponding word heading of the word index 3 to be linked is found from the corresponding n character heading of the n character index 4, and from this link the corresponding character in the text 2 is searched. I am trying to search for a column and output it.

In these cases, the n-character index 4 is a 1-character index. Therefore, by searching the corresponding character string from the text 2 using the n-character index 4 and the word index 3, the maintenance work of the word dictionary 12 can be saved and the index capacity can be reduced to search the text data. Is possible. In particular, by creating the word index 3, it is possible to add an index to a word (2 characters, 3 characters, 4 characters, etc.) registered in the word dictionary 12 and an unregistered word collected by character type information, and reduce the index amount. it can. Also, by linking the word heading of this word index 3 from the n-character index 4, especially the 1-character index, it becomes possible to eliminate the omission of the character string in the text.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the construction and operation of an embodiment of the present invention will be described in detail with reference to FIGS.

FIG. 1 shows a block diagram of an embodiment of the present invention.
In FIG. 1, text data 1 is text data to be searched and stores a plurality of texts 2.

The word index 3 is composed of a plurality of sets of word headings and links, and a word heading is an unregistered word obtained by dividing the text 2 into words and character type information. It is a linked string of characters. For example, as shown in the figure, the word heading link “information” is a link to the corresponding character string “information” in the text 2, and the word heading “information” is a link to the corresponding character string “information” in the text 2 ( Pointer).

The n-character index 4 is an n-character index obtained by extracting n characters from the beginning (for example, the beginning or the whole) of the word index of the word index 3.
The corresponding word is linked from the character heading. For example, as shown in the figure, n-letter heading link information link to "information", link to "condition" link to "information search" Links to the corresponding character strings “information”, “condition”, “information search”, etc. in the index 3 are set. As a result, the search for the first n characters of the keyword, especially one character, can be performed at high speed. When it is determined by the one-character index 4 that the first character of the keyword is present, a character string including the first one character can be extracted from the heading of the linked word index 3 and the matching with the keyword can be determined at high speed. When they match, the corresponding character string in the text 2 can be searched by further following the link of the word heading of the word index 3.

The keyword is a character string to be searched.
When searching, the first character of the keyword is "JO"
Is found from the 1-character index 4 which is the n-character index 4, and the word of the word index 3 containing the keyword is found from the found 1-character link. When found, the relevant character string is extracted from the text by the link of this word and displayed.

Next, referring to FIG. 2, an index creating system 11 for creating the word index 3 and the n-character index 4 in FIG. 1 and the created n-character index 4 will be described.
The operation of the search system 21 when searching for a character string using the word index 3 will be sequentially described in detail.

(1) The index creating system 11 will be described. In FIG. 2, S1 takes in the text data 1. In S2, the preprocessing unit 5 performs preprocessing. For text data 1, take line breaks and make one sentence per line.

In S3, the word division processing unit 6 divides the text data 1 into words. This refers to the word dictionary 12,
For example, as shown in a word division result of FIG. 3 described later, the word is divided into ////////// collect / ... by the / delimiter.

In step S4, the unregistered word processing unit 7 processes the unregistered word. For text data 1 that is not in the word dictionary 12, character type information, such as a series of katakana character strings, is an unregistered word, or a series of kanji characters is an unregistered word.

In S5, the words divided by the word division processing unit 6 and the unregistered words divided by the unregistered word processing unit 7 are
The word index creation unit 8 uses these words and unregistered words as headings and links the text data 1
Link the corresponding position in. As a result, the word index 3 in S6 has been created.

In step S7, the n-character index creating unit 9 extracts the first n characters (for example, the first character) of the word heading of the word index 3, sets the extracted n characters as an n-character heading, and sets the word heading. Link n
Create character index 4. As a result, a series of word index 3 and n-character index 4 can be created, and the text data 1 is ready to be searched.

In order to improve the efficiency of the word index 3, the dictionary editor 13 tunes the word dictionary 12 (registers a new word, corrects it, or registers an unregistered word as a new word). And so on).

As described above, the n-character index 4 and the word index 3 for searching an arbitrary word and unregistered word from the text data 1 can be created.

(2) The search system 21 will be described. Here, the n-character index 4 is a 1-character index. In FIG. 2, in S11, a keyword is input. For this, the operator inputs a keyword to be searched from the screen.

In step S12, the first character of the keyword is compared with the one-character headline. This compares, for example, the leading one character “information” with respect to the keyword “information” and the one character index of the one character index 4, and finds a match. If there is not, a message indicating that there is no is displayed on the screen. If there is, go to S13.

In step S13, the word index of word index 3 is compared with the keyword. This is because it is found in S12 that one character at the beginning of the keyword, for example, "information", exists in the one-character index of the one-character index 4, so the word of the word index 3 of the link destination of this one-character index and the keyword To compare. If there is no match, that is, if there is no match with the keyword or no matching part is included, a message indicating that there is no match is displayed on the screen. If there is, go to S14.

In step S14, the text is compared with the keyword. This is because in S13, the keyword matches the word heading or the keyword is partially included in the word heading, so the text of the link destination and the keyword are compared. When there is no keyword, that is, when the keyword does not match the text, a message indicating that there is no keyword is displayed on the screen. If yes, the process proceeds to S15.

In S15, since it is found in S14 that the keyword matches the character string of the text, the contents of the position of this text (the position of the link destination of the word heading) is displayed on the screen. As a result, the text in which the keyword instructed to search is present (for example, text unit, paragraph unit, page unit text) is displayed on the screen.

As described above, in response to the input of a keyword on the screen, a one-character index having a one-character index 4 matching the leading one character of the keyword is found, and the word index 3 of the link destination of this one-character index is searched. When a matching one of the word headings is found, and when the linked text of the matching word heading and the keyword match, the text in this range is displayed as a search result on the screen. As a result, the character string on the text in which the keyword exists is displayed.

FIG. 3 is a diagram for explaining the creation of the word index according to the present invention. The text data 1 is a document, and in the case of T2 shown in the figure (sentence unit, paragraph unit, page unit, etc.), "collects information about a certain matter and stores it in a file. ... "
Is.

Here, the preprocessing results are summarized in sentence units. The word division result is obtained by dividing the preprocessing result into words by referring to the word dictionary 12 and further dividing the unregistered words into words based on the character type information.
Here, the words are divided into the following words with the delimiters shown in / as shown below.

T2 / on / on / on / information / collecting / ... Further, the divided words and links are arranged in an easy-to-understand manner as follows.

Regarding the word heading link T2 information of T2 T2 is T2 The word index 3 is, as to the word division result, as the word heading and the link, as shown below,
Store in word index 3.

Word heading link information T2, T3, T10 ... Situation T8, T9, T23 ... And the first character of the word heading of these word index 3 is taken out and one character heading of 1 character index 3 is taken out. And the link points to the position of the word heading.

As described above, the preprocessing result is obtained corresponding to the designation of the text data, the word division result is obtained next, and the heading and the link of the word index 3 are registered based on the word division result. To do. Then, the first one character of the word heading is set as the one-character heading of the one-character index 4, and the word heading is linked. As a result, it becomes possible to automatically create the word index 3 and the one-character index 4 from the text data.

[0038]

As described above, according to the present invention,
A configuration in which the word index 3 is created by dividing the word from the text data 1 and the n characters of the word heading of this word index 3 are set to the n character index 4 and a link is set to create the word index 3 and the n character index 4. Therefore, the unregistered word that is collected by the word of the arbitrary character string and the character type information is extracted from the text data 1 as the word index of the word index 3,
The number of entries can be reduced to reduce the memory capacity, and the first n characters (for example, the first character) of the word heading can be changed to n.
By setting the character index to 4, it is possible to eliminate word leakage. As a result, the labor of maintaining the word dictionary can be saved, the index capacity can be reduced, and the text data can be searched. In particular, by creating the word index 3, an index is given by setting one word (two characters, three characters, four characters, etc.) registered in the word dictionary and one unregistered word collected by the character type information. Yes, the amount of indexes can be reduced. Also, by linking the word heading of the word index 3 from the n-character index 4, especially the 1-character index, it becomes possible to eliminate the omission of the character string in the text.

[Brief description of drawings]

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

FIG. 2 is an operation explanatory diagram of the present invention.

FIG. 3 is an explanatory diagram of creating a word index according to the present invention.

[Explanation of symbols]

 1: Text data 2: Text 3: Word index 4: N character index 5: Pre-processing unit 6: Word division processing unit 7: Unregistered word processing unit 8: Word index creation unit 9: N character index creation unit 12: Word Dictionary 13: Dictionary editor

Claims (4)

[Claims] 1. An information retrieval apparatus for retrieving a document, wherein a text is word-divided into words, and a word index (3) for setting a link to the corresponding text as a word heading, and these word headings. , An n-character index (4) for setting a link to the corresponding word heading from an n-character heading taken out from the beginning or the like, and an information retrieving apparatus. 2. When the above-mentioned text is word-divided into words, unregistered words that are not registered in the word dictionary (12) are grouped by character type information into word headings. The information search device according to claim 1. 3. A corresponding word heading of the word index (3) linked from the corresponding n character heading of the n character index (4) is found in response to a search instruction specifying a keyword, and this word heading The information retrieval apparatus according to claim 1, wherein the information retrieval apparatus is configured to retrieve and output a character string of the corresponding text from the link. 4. The n character index (4) is 1
The information retrieval device according to claim 1, wherein the information retrieval device is a character index.