KR101331946B1 - Search method using wildcard matching - Google Patents

Abstract

The present invention relates to a search method using wildcard matching. According to the method of the present invention, when a wildcard search query is input to a database by a user terminal, a search engine corresponds to the wildcard search query using an index file which has undergone an index filtering process by the following software algorithm. It is characterized by extracting the result.
(a) extracting index word data from a set of documents in the database;
(b) categorizing index words having the same nth character (n is an integer greater than 0) in the extracted index word data to generate a category set; And
(c) pre-storing the set of categories as location information of categorized index words;
Through such a configuration, the present invention can extract search results more quickly and economically in searching for various wildcard search queries, such as a forward truncation search query, a backward truncation search query, a central truncation search query and a bidirectional truncation search query. have.

Description

Search method using wildcard matching {SEARCH METHOD USING WILDCARD MATCHING}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a retrieval method for retrieving information scattered over a wide range of computer networks. In particular, the present invention relates to a retrieval method for easily searching a query word including a wildcard for forward and backward truncation search.

Search engines have long been an indispensable function for using the Internet.In addition to the leading portal services at home and abroad, companies, organizations and organizations that have their own data also offer high quality services and efficient work. It provides its own search function.

Generally, such a search is performed on a large set of documents, so a simple search method requires a lot of time and money. Therefore, in the search preparation phase, the words that are expected to be used to express the information needs of the user are extracted from the document set to be searched, and the index is constructed in the reverse direction. The words that make up the index are called "indexes."

In general, since a user's information request is formed in the form of general nouns or proper nouns, it is necessary to selectively select only general nouns or proper nouns from the document set. The process of analyzing the parts of speech structure of sentences according to the grammar is called "morphological analysis." Most commercial search services have their own stemmers.

The process of analyzing and selecting index words according to the purpose of the search service among the parts of speech information analyzed primarily is called index word extraction process. Ordinary nouns or proper nouns may be used, and verbs, adjectives, etc., which are considered to be high in foreign words, rhetoric, and utility, may be additionally extracted.

Referring to FIG. 1, assume that there are three documents in a database D of documents to be searched, each having a title of "Information Search Theory", "Latest Information Search Theory", "Information Communication Article", and the like. Document 1 titled "Retrieval theory" contains the sentence "This book covers the basic theory of information retrieval…", and Document 2 titled "Modern Information Retrieval Theory" contains "the latest information widely used in industry." Suppose that there is a search system… ”, and document 3, titled“ Information and Communication Engineer, ”contains the sentence“ Information and Communications Management Capability Certificate ”. When these documents are decomposed by the morphological analyzer and the index extractor 10 by the morphological analysis process, as shown in the extracted index word list of 1a, 1b, 1c, the information, search, theory, theory, It can be seen that it consists of words such as "latest" and "communication". Among these, words that are determined to be used to describe the information request of the users may be separately recorded in which document the information appeared in the document, and the index structure 20 may be constructed.

By applying the above process to all documents to be searched, the whole process of extracting the appearance document information for all index words appearing in the document set is called index construction.

The user search may be performed using the index thus constructed. Now, let's look at the conventional keyword search method using index building. Suppose that a search query of “basic search” is input through the web site input window 50 as shown in FIG. 2. Similarly, morphological analysis reveals that meaningful query terms are index terms of "primary" and "search", and document appearance information of "primary" index word and document occurrence information of "search" index word are searched through the index structure 20. The appearance information that is commonly displayed by fusion is returned to the user as a final result on the screen of the web page 100.

When building an index, this decomposition takes a lot of time because it is applied to all documents in batches, but once the index is built, the search can be performed very quickly. In general, since the index is built once during the search service cycle and reused until the end of the search service cycle, the search speed is much faster than the sequential search without the index from the point of view of the user using the search.

A lot of information can be easily accessed through the medium of the Internet, and the quality of search work has been greatly improved through long research. However, as the conventional search technique is enjoyed not only for technical experts but also for general non-professional users, new problems have begun to emerge. It is a problem that it is difficult for a user to describe exactly the information he wants.

The information request of the user expresses the information that the user wants to find out through the search in the form of sentences or words, and the search grasps the intention of the user and the "documents containing the keywords constituting the query presented by the user in order of importance. Will be listed. " In other words, the search query corresponds to the index word in the inverted index constructed above, and a series of operations that show the document subset including the word in the proper order form the core of the search service. For example, if a user enters the query "Year of King Sejong was born?", The search is to present the results in the most appropriate order among the documents containing the query.

Statistically, the user's information request is known to be represented using 1.6 search query words. In most cases, one word is used to search, and if only one word is difficult to express an information request, it means that two or more search query words are combined to search.

However, as there is a lack of prior knowledge of this process and foreign users frequently use Internet search in the field of their search, it has become increasingly difficult to select these search queries. It is also difficult to think that the query input by the user is not formalized and that the query input by individual users is systematically managed.

Similarly, when a search query word that can accurately represent the content that the user wants to find is not found, the word may not have a complete and clear form and may consist of only partial syllables. For example, if you want to search for books related to IR, you may want to construct a query word using special characters such as "* IR *" rather than simply using the word "IT" as your search query word. have. In this case, it means that a certain syllable may correspond to zero or more characters in place of the "*" character. These special characters are commonly called "wildcards". If a book title of "latest information retrieval theory" is included among the index words possessed by the search engine, the corresponding book information is output to the user corresponding to the above expression.

Such a search using a wildcard-inserted query is commonly referred to as a "cutting search".

Truncation search can be further subdivided according to the position of the wildcard. If a wildcard symbol follows the query, such as "OOO *", it is called "rear truncation search" and a wildcard symbol such as "* OOO" In the case of the query term, the case of "front truncation search" and "OO * OO" such that the query character is wrapped before and after is called "central truncation search".

The problem of comparing one string and a string including wildcards with each other and determining whether they correspond is called "one-to-one pattern search". Already through conventional research, algorithms for determining whether a single string corresponds to a string including a wildcard as described above have been proposed through many studies.

However, the conventional search service has a very large number of index words, and using the above-described method, a process of checking whether or not it corresponds to a query word including a wildcard sequentially takes a very long time. Thus, an improved search service uses a special data layout structure inside the engine to quickly find the index words needed, and this search is called "one-to-many pattern search".

While the conventional tree or trie forms are easy to search for query words that do not include wildcards, additional search costs are consumed to search for query words that include wildcards. So new data layout structures, such as suffix trees and bidirectional tries, were studied. However, these data structures require additional storage space and are used through complex processing, which creates a problem of increasing maintenance and management burden.

The inventors of the present invention have completed the present invention after research efforts to solve these problems.

The present invention relates to an easy algorithm to search for query words including wildcards for forward and backward truncation search. The purpose is.

On the other hand, other unspecified purposes of the present invention will be further considered within the scope of the following detailed description and easily deduced from the effects thereof.

In order to achieve the above object, in the present invention, when the wildcard search query word is input to the database by the user terminal, the wildcard using the index file that the search engine has undergone the index filtering process by the following software algorithm A search method using wildcard matching, which extracts a search result corresponding to a search query.

(a) extracting index word data from a set of documents in the database;

(b) categorizing index words having the same nth character (n is an integer greater than 0) in the extracted index word data to generate a category set; And

(c) pre-storing the set of categories as location information of categorized index words;

In the search method using wildcard matching according to an embodiment of the present invention, when the wildcard search query is a backward truncated search word,

An intersection of category sets corresponding to syllables from the first letter of the wildcard search query to the wildcard symbol may be generated and output as a search result.

Further, in the search method using wildcard matching according to an embodiment of the present invention, when the wildcard search query is a forward truncated search word,

The intersection of category sets corresponding to the last syllable after the wildcard symbol of the wildcard search query word may be generated for all candidate groups up to the maximum length of the index word, and the union of the generated intersections may be combined again and output as a search result. have.

In the search method using wildcard matching according to an embodiment of the present invention, when the wildcard search query is a central truncated search word,

Based on the wildcard symbol, split into a first search query query and a second search query query,

The search result for the first search query is obtained by intersecting the set of categories corresponding to the syllables from the first letter of the first search query until the wildcard symbol appears.

The search result for the second search query word generates a set of categories of categories corresponding to the last syllable after the wildcard symbol for all candidate groups up to the maximum length of the index word, and combines the generated sets again. After getting

The result of the intersection of the search result for the first search query term and the search result for the second search query term may be output as a final search result for the central truncated search term.

Further, in the search method using wildcard matching according to an embodiment of the present invention, when the wildcard search query is a bidirectional truncated search word,

Split into a front wildcard symbol and search query pair and a search query and rear wildcard symbol pair,

The search result of the forward wildcard symbol and the search query pair is generated for all candidate groups from the syllables following the front wildcard symbol to the maximum syllable length of the set of categories corresponding to the last syllable. Get the intersection again,

The search results for the search query and backward wildcard symbol pairs are obtained by intersecting sets of categories corresponding to syllables before the trailing wildcard symbol.

The result of intersection of the search result for the front wildcard symbol and the search query pair and the search result for the search query word and the rear wildcard symbol pair may be output as a final search result for the bidirectional truncated search word.

Another aspect of the invention is characterized by a computer-readable recording medium containing a computer program for executing the above search method algorithm using wildcard matching.

Through the above-described problem solving means, the present invention, even if the user who tries to search first can not accurately grasp the information request that can have an incomplete search query can be used to mirror the effect of improving the effectiveness of the search.

In addition, according to the search method of the present invention, the same effect as inputting various kinds of search query words in a short time can save the time required for searching, and can review a lot of data in a short time It can provide an edge.

Even if effects not specifically mentioned in the specification of the present invention are incorporated, the provisional effects expected by the technical features of the present invention are treated as described in the specification of the present invention.

1 and 2 are diagrams illustrating an index structure according to a conventional search method.
3 is a diagram conceptually illustrating an index word filtering process before a search according to the present invention.
4 is a diagram illustrating a categorized index structure according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating a search result extraction method when a backward truncation wildcard search term is used using the index structure of FIG. 4.
FIG. 6 is a diagram illustrating a search result extraction method when a forward truncation wildcard search term is used by using the index structure of FIG. 4.
FIG. 7 is a diagram illustrating a search result extraction method when a central truncated wildcard search term is used using the index structure of FIG. 4.
* The accompanying drawings illustrate examples of the present invention in order to facilitate understanding of the technical idea of the present invention, and thus the scope of the present invention is not limited thereto.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may obscure the subject matter of the present invention.

3 conceptually illustrates an index file forming process through an index word filtering process according to the present invention. This process is done by software algorithms. The index word is extracted by the stemmer and the index word extractor from the document set (D) in the database existing in the computing network (S10), and the nth letter from the index word from which the categorization module is extracted (n is an integer greater than 0, is less than or equal to) The same index words are categorized (S20), the location information of each categorized index word is stored separately (S30), and through this process, the final index file (F) is formed. The index word formed into the index file F is the result of filtering the index word through the categorization process, which creates a new search query record structure.

<Document set>

Commercially available search engines are premised on the construction phase of the conceptual index word archiving that they have.

The document set D to be searched is composed of documents each assigned an identifier (a unique number for a document for speeding up a search result operation), and the document can be distinguished through this identifier in the future. In some cases, it may consist of several document sets. At this time, each document set also has an identifier so that only one document can be distinguished.

<Location information of index>

Each document goes through the process of stemming and keyword selection at indexing stage, and extracts important words used for search. The criteria for selecting these words may differ slightly depending on the type of material and the purpose of the search.

The words thus extracted are registered in the index structure. In the present invention, the identifier and location information of the document in which the word itself appears are connected and recorded like a tag. If a word occurs in multiple documents, the identifier of the document is linked and stored as a tag for that word. For example, referring to FIG. 1, it can be seen that the word “information” appears at document identifiers 1, 2, and 3. This process is performed for all documents in the document set to be searched and indexed.

However, the present invention does not use the tree-shaped or tri-shaped index structure as shown in FIG. 2. The location information 131 of the categorized index word as shown in FIG. 4 is stored in advance in the index database in the document indexing step. That is, in the present invention, an index file having position information 131 of categorized index words stored in advance is formed, and a search engine performs a search corresponding to a keyword by using the index file.

<Index filtration method>

Let us explain the index word filtration method according to the present invention as shown above. An index can be built with an index extracted from the document set (D). As shown in Figure 4, the present invention undergoes a categorization process of the index words being used in the index. That is, the document set D is decomposed by the morphological analyzer and the index word extractor 100 by the morphological analysis process to generate the index word list 120, and then subjected to the categorization process.

In the categorization method according to the present invention, as shown in the index word filtration set 130 of FIG. 4, the index words of n (n is an integer greater than 0) are categorized. For example, shoes and Sinjeong-dong can be categorized together because the first letter is the same "god". Munjeong-dong and Sinjeong-dong also belong to the same category as the "jeong". As described above, each category has unique location information. This is summarized in Table 1 below. The sign indicating the location information is indicated by a reference numeral for convenience.

Location information Index words belonging to the category Category Rules 131a Alarm, information Index word with second syllable "Bo" 131b Advanced Information, Woo Rae Jung, Latest Information, Communication Policy Index word with third syllable "positive" 113c Munjeong-dong, Sinjeong-dong Index word with second syllable "positive" 131d Betrayal, Spirit, Latest Technology, Latest Information, Latest Politics, Communication Policy Index word with second syllable "God" 131e Shisebo Index word with third syllable "bo" 131f Shoes, Sinjeong-dong Index word with first syllable "God" 131 g Intelligence Index word with first syllable "positive" 131h Advanced Information, Chosun Daily, Latest Information Index word with fourth syllable "Bo" 131i Recent, Latest Technology, Latest Information, Latest Politics Index word with first syllable "Last"

When the categorization process is performed as described above, one word may have multiple location information.

<How to extract search results when wildcard search terms are used>

Now, let's look at a method for performing truncation searches through the categorized index filter set 130 as described above. Extraction of search results for wildcards.

First, in the case of backward truncation, the search result is the intersection of the set of categories corresponding to the syllables from the first letter of the search query to the wildcard symbol.

For example, in an embodiment where the search query is "latest *", query analysis compares the index filter set belonging to the category with the first letter "latest" and the category with the second letter "O shin" and corresponding to the intersection. Output index word elements as search results. Generates the following category candidates:

Candidate 1) Choi, O Shin

As shown in FIG. 5, the set of index words in the above category candidates are respectively:

Candidate "latest" = {latest, latest technology, latest information, latest political science},

Candidate "O God" = {Betrayal, Spirit, Latest Technology, Latest Information, Latest Politics, Communication Policy}.

The intersection of index terms in each category is now obtained, and the search results are "latest technology", "latest information", "latest politics", and the result is "latest *" backward truncation search.

Next, let's look at the case of forward truncation search. Since we can't predict the length of a wildcard match, we have to figure out all the possible matches. That is, the intersection of the set of categories corresponding to the last syllable after the wildcard symbol is generated for all candidate groups up to the maximum length of the index word, and the final index word is generated by re-joining the intersection sets of all the candidate groups thus generated. Thus, category candidates are generated from the shortest to the longest form of this index.

In the embodiment as shown in Figure 6, when the search query is "* information", since the second letter after the string is "jeong" and the last letter is "bo" information can be known as the category candidates are generated as follows.

Candidate 1) Chung, O

Candidate 2) O-cho, OO report

Candidate 3) OO Chung, OOO Bo

In the example of FIG. 6, since the longest candidate group is "OO tablets, OOO beams", up to candidate 3 may be generated. Now, we compare the index word filtering sets belonging to each candidate category, and generate the index word candidate group corresponding to the intersection. In this embodiment, each result is calculated as follows.

Candidate 1) "Jeong" ∩ "Obo" = {Informational}

Candidate 2) "O-jeong" ∩ "OO-Bo" = {}

Candidates 3) "OO-cho" ∩ "OOO-Bo" = {Advanced Information, Latest Information}

Now, when the union of each index word candidate group is found, the result is a forward truncation search. In the present embodiment, {information, advanced information, latest information} is the result for the given forward truncation search query "* information".

Next, in the embodiment related to the central truncated search query, for example, it may have a form such as "OO * XX", and two search queries, "OO *" and "* XX", that is, a first search query centered on wildcards. And performing the same process as the method of backward truncation searching for the first search query in the above-described method, and intersecting the results of applying the forward truncation search for the second search query. The result obtained is the result of this central truncation search.

If you attempt to perform a central truncation search for the "latest * information" for the above example, the result of intersecting the results obtained by each of the "latest truncation" search of "latest *" and the forward truncation search of "* information" is the final center truncation search. As a result, as shown in FIG. 7, a result of "latest information" can be obtained.

Also, an embodiment of a bidirectional truncated search query may be performed by the same rules as the central truncated search query. For example, in the case of "* information *", a forward truncation search may be performed for "* information", and a back truncation search may be performed for "information *", and the result of intersection of these may be output as a final search result. .

In other words, a forward truncation search is performed for the forward wildcard symbol and the search query pair "* information". The search result for the forward wildcard symbol and the search query pair is a category corresponding to the last syllable from the syllable after the forward wildcard symbol. An intersection of sets is generated for all candidate groups up to the maximum length of the index word, and each intersection is generated again by combining. In addition, the search result for the search query and the "wildcard" information pair "information *" is obtained by the intersection of the category sets corresponding to the syllables before the trailing wildcard symbol. Then, the search results obtained from each can be alternated again and output as the final search result.

In the present invention, the wildcard symbol is mainly described with respect to "*". However, the present invention can be extended to wildcard symbols such as the "?" Symbol (any syllable corresponds only to one). When the wildcard query word using the wildcard symbol "?" Is "information", the truncation search may be performed by intersecting the "OO" candidate group and the "OO bo" candidate group.

On the other hand, the scope of protection of the present invention is not limited by the embodiments explicitly described above. Further, it should be noted that the protection scope of the present invention may not be limited due to obvious changes or substitutions in the technical field to which the present invention belongs.

Claims (6)

When a wildcard search query is input to a database by a user terminal, a search engine extracts a search result corresponding to the wildcard search query by using an index file that has undergone the index filtering process by the following software algorithm. Search method using wildcard matching.
(a) extracting index word data from a set of documents in the database;
(b) categorizing index words having the same nth character (n is an integer greater than 0) in the extracted index word data to generate a category set; And
(c) creating an index file by uniquely assigning location information of index words belonging to the category set to each category and storing the unique location information of the index words in an index word database. The method of claim 1,
If the wildcard search query is a backward truncated query,
And generating an intersection result of a set of categories corresponding to a syllable from a first letter of the wildcard search query to a wildcard symbol and outputting the result as a search result. The method of claim 1,
If the wildcard search query is a forward truncated search term,
Generating a set of category sets corresponding to the last syllable from the next syllable and the last syllable of the wildcard search query for all candidate groups up to the maximum length of the index word, and combining each generated set again and outputting the result as a search result. A search method using wildcard matching. The method of claim 1,
If the wildcard search query is a central truncation query,
Based on the wildcard symbol, split into a first search query query and a second search query query,
The search result for the first search query is obtained by intersecting the set of categories corresponding to the syllables from the first letter of the first search query until the wildcard symbol appears.
The search result for the second search query word generates a set of categories of categories corresponding to the last syllable after the wildcard symbol for all candidate groups up to the maximum length of the index word, and combines the generated sets again. After getting
And a result of intersecting the search result of the first search query term and the search result of the second search query term as a final search result for the central truncated search term. The method of claim 1,
If the wildcard search query is a bidirectional truncated query,
Split into a front wildcard symbol and search query pair and a search query and rear wildcard symbol pair,
The search result of the forward wildcard symbol and the search query pair is generated for all candidate groups from the syllables following the front wildcard symbol to the maximum syllable length of the set of categories corresponding to the last syllable. Get the intersection again,
The search results for the search query and backward wildcard symbol pairs are obtained by intersecting sets of categories corresponding to syllables before the trailing wildcard symbol.
Wildcard, characterized in that the result of intersecting the search result for the front wildcard symbol and the search query pair and the search result for the search query word and the rear wildcard symbol pair is output as the final search result for the bidirectional truncated search word. Search method using matching. A computer-readable recording medium containing a computer program for executing a search method algorithm using wildcard matching according to any one of claims 1 to 5.

Images