KR100906810B1 - Keyword searching method - Google Patents

Abstract

A keyword search method is provided to show a web document highly related to keywords that a user sincerely wants. A set recommendation keyword set correspondingly to a search keyword that a user inputs is read(S206). A web document including the search keyword is searched(S207). The web document including the recommendation keyword is searched(S209). A ranking list is generated by scoring web documents including the search keyword(S211). According to the ranking list, summary information of each web document including the search keyword is generated(S213). The summary information is provided to the user(S214).

Description

How to search for keywords {KEYWORD SEARCHING METHOD}

The present invention relates to a keyword retrieval method, and more particularly, to a keyword retrieval method in which a search engine searches for a keyword that satisfies a user's search intention using the collective intelligence of the user.

In general, in order for a user to obtain desired information on the Internet (for example, a web site, a web document, etc.), the user accesses a search site that provides a search service through his terminal, and then provides the search site. By entering a search keyword in a search box and executing a search, a search engine of a corresponding search site searches a web site or a web document including the search keyword and provides the search result as a summary information.

In this case, if the user wants to obtain the desired information by the search keyword input in the search box provided by the search site, the search keyword that satisfies the user's search intention should be entered in the search box. By inputting an incorrect search keyword due to the neutrality and stemming error of a word, a mistake of a document author, and the like, there is a problem that the input search keyword does not sufficiently reflect the search intention of the user.

Looking at the problem caused by the first neutrality in more detail as follows.

When a user searches on the search engine with the keyword 'next', even though the actual user enters the keyword 'next' with the intention of obtaining information related to 'next communication', the existing search engine simply inputs the user. Because search results are provided only by whether or not they contain one search keyword, documents containing the keyword "next" as opposed to "previous" are mainly provided as search results, and the meaning of "corporate and internet services" 'Next' may only provide some.

In other words, other search terms that users who normally search for 'next' enter with the intention of obtaining information related to 'next communication' may include companies or 'next search', 'next media', 'next cafe', etc. Despite being 'next' as an internet service, the word 'next' may appear more in the Korean document as opposed to 'previous' when searched with the keyword 'next' in a search engine.

In another example, if a search engine performed a search with the keyword "LA", the actual user entered the keyword "LA" for the purpose of obtaining information about "US cities", The documents containing the back are mainly searched, and thus the accuracy is low.

Looking at the problem caused by the second stemming and documentation errors, for example, as follows.

If a search engine performed a search with the keyword 'Edo', the actual search engine stemmer would search for 'Edo' even though the user entered the keyword 'Edo' with the intention of obtaining information related to 'Japan'. It is possible to determine probabilisticly whether the word is used as a noun or an investigation, but because of the many errors in the writing of a web document, such as in a sentence such as 'I wasn't frustrated with polarization,' 'Edo' as an investigation More often, documents containing keywords are provided as search results, and only a few 'Edo' as a noun can prevent the stemmer from producing accurate search results.

In other words, other search terms entered by users searching for 'Edo' with the intention of obtaining information related to 'Japan' are 'Edo' as a noun, such as 'Edogawa Ranpo' and 'Edo'. Nevertheless, when the keyword "Edo" is searched in a search engine, the word "Edo" is more often used as a search in Korean documents, so it is difficult to find "Japan" related documents in the keyword search results. .

In another example, if a search engine performed a search with the keyword 'gaya', the actual user entered the keyword 'gaya' with the intention of obtaining information related to 'historic' or 'gaya cc'. It is often the case that documents containing 'gaya' as a verb are mainly searched for documents such as 'way to go' and 'how long should coma go'.

As described above, the conventional keyword search has a problem that the user's search intention is not sufficiently satisfied due to the neutrality of the Hangul word, the morphological analysis, and the error of document writing.

An object of the present invention is to solve the problems described above, and to provide a keyword search method that allows a search engine to search for a keyword that satisfies the user's search intention using the collective intelligence of the user. .

In addition, the present invention utilizes the collective intelligence of a person using a search service in a search engine to perform a keyword search that satisfies the search intention of the user of the search service even when the query is composed of two or more terms. In addition, the dynamic update of the keyword search database can be freely performed, so that a user can search a web document that is highly correlated with the keywords he / she really wants, thereby providing a high accuracy of the user's search results. Its purpose is to make it possible.

In order to solve this problem, according to the present invention, when searching for a specific keyword in the search engine to analyze the user pattern (ie, the collective intelligence of the user) for the search words entered by the user to use the user's search intention It extracts related keywords that can be well reflected and organizes them into a database of suggested keywords for each search word. It analyzes the web documents to be searched in advance and analyzes the web documents including the search terms and the web documents including the suggested keywords. After forming an inverted index structure (i.e., data), we search for suggestions with two or more terms by searching for relevant keywords for the search term in real keyword search and using the inverted index structure to expose only relevant web documents. Even if you do, it will fully satisfy your search intent. Emitter and allows dynamic updating of the database is made more freely than, and thus allows the user to give a truly shows the high correlation between web pages with the keywords you want.

According to an aspect of the present invention, in a method of searching for a keyword input by a user, when a user searches for a keyword, a recommended keyword corresponding to the keyword to be searched is read, and each web including the keyword to be searched is included. Searching each web document including a document and a corresponding keyword, and applying a weight to a score of each web document including the recommendation keyword among the web documents including the keyword to be searched to include the keyword to be searched. Adjusting a ranking list of each web document, and generating summary information of each web document including the keyword to be searched according to the adjusted ranking list and providing the same to a user. .

The keyword search method may include setting information about a ID of each web document including the keyword to be searched and making a database, and setting recommended keywords corresponding to the keyword to be searched by using the collective intelligence of the user. Database and setting information about the ID of each web document including the recommendation keyword to make a database; and setting and databaseting information on the body of the web document corresponding to the web document ID. It further comprises a step. The keyword search method may correspond to each web document ID including the keyword to be searched, recommended keywords corresponding to the keyword to be searched, each web document ID including the recommendation keyword, and the web document ID. And periodically resetting information on the body of the web document.

The searching of each web document may include: searching an ID of each web document including a keyword input by the user from the database, reading a recommendation keyword corresponding to the keyword input by the user from the database; And retrieving the ID of each web document including the recommendation keyword corresponding to the keyword input by the user from the database.

The adjusting of the ranking list of each web document may include: scoring each web document including a keyword input by the user according to a matching degree or page ranking with the keyword input by the user; Determining whether there are web document IDs including the recommended keyword among web document IDs including the entered keyword, and if there are web document IDs including the recommended keyword, Adding weights to scores, re-adjusting scores for each web document including the keywords entered by the user, and generating a predetermined number of web document IDs in a list in order of increasing the re-adjusted scores; do. Alternatively, the adjusting of the rank list of each web document may include determining whether there are web document IDs including the recommendation keyword among web document IDs including the keyword input by the user, and the recommendation. If there is a web document ID including a keyword, applying a weight to each web document corresponding to the keyword, whether the weight is applied to each web document including the keyword input by the user, and Scoring according to a matching degree or page ranking with the input keyword, and generating a predetermined number of web document IDs in a list in order of increasing score.

Generating and providing the summary information to the user may include reading information from the database about the body of the web document corresponding to each web document ID in the web document ID list, according to the read web document body. Generating summary information of the web document corresponding to each web document ID of the web document ID list, and transmitting the summary information of the generated web document to the user terminal according to the web document ID list to the user as a search result; Providing a step.

According to another aspect of the present invention, in a method for searching a keyword input by a user, setting and databaseting information on the ID of each web document including a keyword to be searched by the user, and grouping the user's collective intelligence Setting and databaseting the recommended keywords corresponding to the keyword to be searched by the user using the database; setting and databaseting information on the ID of each web document including the recommended keyword; Setting database information on the body of the web document and retrieving the ID of each web document including the keyword input by the user when the user searches for the keyword from the database; The database of recommended keywords corresponding to keywords Reading from the database the ID of each web document including the recommended keyword corresponding to the keyword input by the user from the database, and the recommendation among the web document IDs including the keyword input by the user. Determining whether there is a web document ID including a keyword, applying a weight to each web document corresponding to the web document ID including the recommended keyword, and inputting the keyword entered by the user. Scoring according to whether the weight is applied to each web document, whether the weight is applied, the degree of matching with the keyword input by the user, or the page ranking, and a predetermined number of web document IDs are listed in order of the high score. Generating, and the generated web document child Depending on the list to generate summary information for each Web document that contains the keywords that the user input comprises the step of providing to the user.

As described above, according to the present invention, after searching a user's search word for a search word that reflects the user's search intention and forming an inverted index structure for the web document including the search word and the web document including the suggestion keyword, the user searches for a specific keyword. In order to use the database and the inverse index structure, score weights are given to the web documents containing the suggested words among the web documents containing the corresponding keywords, and the web document ranking list is highly correlated with the desired keywords. By adjusting and providing it to the user, it is possible to present a web document that is highly correlated with the keywords that the user really wants. In addition, according to the present invention, even when a search is performed using a query consisting of two or more terms, the search intention of the user can be sufficiently satisfied, and the dynamic update of the database and the inverted index structure can be made more freely.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding other components unless specifically stated otherwise. In addition, the terms “… unit”, “… unit”, “… module” described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. Can be.

Now, a keyword search method according to an embodiment of the present invention will be described in detail with reference to the drawings.

1 is a schematic conceptual diagram of a system for keyword search according to an embodiment of the present invention.

As shown in FIG. 1, a system for keyword search according to an embodiment of the present invention includes a user terminal 110, a wired / wireless communication network 120, a search engine 130, and a database 140.

When the user inputs a search word for searching for a specific keyword, the user terminal 110 transmits the input search word to the search engine 130 through the wired / wireless communication network 120, and the search received from the search engine 130. Provide the results to the user.

The wired / wireless communication network 120 connects the user terminal 110 and the search engine 130 to each other to perform communication.

The search engine 130 searches a web document including a search word received from the user terminal 110, reads a recommendation keyword corresponding to the search word received from the user terminal 110 from the database 140, and reads the read recommendation. Search the web document containing the keyword, give the score weight to the web document containing the suggested keyword among the web document containing the search term, and then score the web document including the search term. It generates a list of IDs (IDs) of documents, and generates summary information on a corresponding search result (ie, a web document including a search word) according to the ID list and as a search result, the user terminal (eg, a user terminal through the wired / wireless communication network 120). 110).

The search engine 130 includes a query broker 131, a searcher 132, and a summary 133.

The query broker 131 receives the search word input by the user from the user terminal 110 through the wired / wireless communication network 120 and provides the database 140 with information on recommended keywords for reflecting the user search intention corresponding to the search word input by the user. After reading at, the information about the read recommendation keyword is applied to the searcher 132 together with the received search word. The query broker 131 authorizes the summary 133 with information about the list of web document IDs authorized from the searcher 132. The query broker 131 transmits summary information of each web document authorized from the summary 133 to the user terminal 110 through the wired / wireless communication network 120 as a search result.

The searcher 132 searches the database 140 for a web document that includes both a search term and a recommended keyword authorized by the query broker 131, weights the score of the searched web document, and then includes the search term. Scoring is performed on the web document, and a list of IDs of the web document according to the scoring is generated to authorize the information on the generated ID list of the corresponding web document to the query broker 131.

The summary 133 receives information about the list of web document IDs from the query broker 131 and reads the body content of the web document corresponding to each web document ID of the authorized web document ID list from the database 140. In response to the list of authorized web document IDs, summary information of each web document is generated and authorized to the query broker 131.

The database 140 may include information on recommended keywords for reflecting a user's search intention corresponding to each search word of users, information about an ID of a web document corresponding to each search word of users, and an ID of a web document corresponding to each recommendation keyword. Information and the information about the web document body corresponding to each web document ID are stored in a database. At this time, it should be understood that the database 140 may be configured as a separate memory according to its function or role.

Hereinafter, a keyword search method according to an embodiment of the present invention will be described in detail with reference to the flowchart of FIG. 2.

First, you'll need to prepare before a user searches for a particular keyword. When a user inputs a search word, a related search word set for more accurately identifying the search intention of the user is generated and stored in the database 140.

In other words, for a keyword that a user can enter for a search (hereinafter referred to as a query), the user's collective intelligence can be used to better reflect the user's search intent (ie, recommend a keyword). Keyword), and sets information on a recommendation keyword for reflecting a user's search intention corresponding to each query, and stores the set information in a database 140 (or a memory provided separately) (S201). . Here, the query may be composed of a plurality of terms. For example, 'next communication' consists of terms 'next' and 'communication'.

In addition, all the web documents that are to be searched for the query that the user can input for the search are analyzed in advance, and the corresponding analyzed data values are stored in the database 140 so that they can be used for later indexing.

In other words, in order to index the web document including the query through stemming the query that the user can enter for the search, the keyword is extracted from all the web documents to be searched. If given, to set the information about the web document ID corresponding to each keyword to find the ID of the respective web document and database it and store it in the database 140 (or a memory provided separately) (S202) . At this time, the structure of the database 140 for the web document ID corresponding to each keyword is referred to as a 'general inverted index structure'.

In addition, in order to index a web document including a keyword corresponding to the query without stemming the query that the user can enter for the search, the keyword is extracted from all the web documents to be searched. In order to find the ID of each web document when a corresponding keyword is given, the database 140 sets the information about the web document ID corresponding to each keyword and then converts the database 140 into a database 140 (or a memory provided separately). Stored in (S203). At this time, the structure of the database 140 for the web document ID corresponding to each recommendation keyword is referred to as a 'referred word index structure'.

In addition, when the web document ID is given, the database 140 (or separately provided) is provided after setting and databaseting information on the web document body corresponding to each web document ID so that the body of each web document can be found. Stored memory) (S204).

In the future, information about a suggestion keyword for reflecting a user's search intention corresponding to each search term of users, information about an ID of a web document corresponding to each search term of users, information about an ID of a web document corresponding to each recommendation keyword, When the information on the web document body corresponding to each web document ID is deleted, changed, or modified, the deleted, changed, or modified information is upgraded in the database 140. Alternatively, by periodically repeating the above-described steps S201 to S204, information about the recommended keyword for reflecting the user search intention corresponding to each search word of the user, information about the ID of the web document corresponding to each search word of the user, and each After resetting the information on the ID of the web document corresponding to the recommended keyword and the information on the body of the web document corresponding to each web document ID, the reset information is upgraded in the database 140, thereby improving the user's search intention. You can keep it in a state where you can better reflect it.

By configuring the database 140 as described above, the dynamic update of the database 140 has a more free advantage in order to grasp the user's search intent in real time, and also easily search for queries consisting of two or more terms The advantage is that you can.

Then, in order to actually perform a search for a specific keyword, the user accesses the search site through the wired / wireless communication network 120 using the user terminal 110, and then inputs the specific keyword in the search window provided by the search site. You enter a query to search for and select the search command.

Accordingly, the user terminal 110 generates a search command message including a query input by the user according to the search command of the user and transmits the search command message to the search site through the wired / wireless communication network 120.

Accordingly, the search engine 130 of the search site receives a search command message transmitted from the user terminal 110 through the wired / wireless communication network 120. In this case, the query broker 131 provided in the search engine 130 receives a query input by the user through the wired / wireless communication network 120 from the user terminal 110 (S205) and searches for a user corresponding to the query input by the user. After reading (i.e., searching) information on the suggestion keyword for reflecting intention from the database 140 (S206), the search engine 130 includes information about the query input by the user and the corresponding recommendation keyword. 132 is applied.

Then, the searcher 132 receives information about the query input by the user from the query broker 131 and the recommended keyword corresponding thereto, and performs a general index to search the web documents including the query input by the user. After performing a scoring and performing a referral index that searches for web documents containing keyword suggestions corresponding to the query entered by the user, the web documents that satisfy both the regular index and the referral index (ie, the user). The web documents that contain both the query and the suggested keyword) are added to the weights to rescore the searched web documents, and the ID list of each web document is generated according to the scoring. To the query broker 131.

In more detail, the searcher 132 receives the query input by the user from the query broker 131 and corresponds to the ID (ie, the query input by the user) of each web document including the query input by the user. The web document ID) is read (that is, retrieved) from the general inverted index structure of the database 140 (S207).

Accordingly, the searcher 132 scores the web document IDs of the web documents including the query input by the user through the matching degree, the page ranking, and the like of the query input by the user. Alternatively, the searcher 132 ranks the read respective web document IDs (S208).

Thereafter, the searcher 132 receives the recommendation keyword corresponding to the query input by the user from the query broker 131, and the ID of each web document including the recommendation keyword (that is, the web document ID corresponding to the recommendation keyword). Is read (ie, retrieved) from the recommendation inverse index structure of the database 140 (S209).

In this case, the searcher 132 may read the web document IDs read from the general inverted index structure of the database 140 in step S207 and the web document IDs read from the recommended index index structure of the database 140 in step S209. Scoring is performed again by assigning more weights to the same web document IDs.

In other words, from among the web document IDs read from the general inverted index structure of the database 140 in step S207 described above, it is determined whether there are web document IDs read from the recommended index index structure of the database 140 in step S209 described above. Then, the web document IDs read from the general inverted index structure are further weighted according to the presence or absence of web document IDs read from the recommended index index structure to readjust the score given in step S208. Alternatively, the searcher 132 reranks the corresponding web document IDs (S210).

Alternatively, the searcher 132 receives information about the query input by the user and the recommended keyword corresponding thereto from the query broker 131, and includes the query input by the user through the general index similar to the above-described step S207. At the same time as searching the web documents, the web documents including the recommendation keyword corresponding to the query input by the user are searched through the recommendation index similarly to the above-described step S209.

Accordingly, the searcher 132 gives more weight to the ID of each web document including the recommended keyword corresponding to the query input by the user among the web documents including the query input by the user, and then inputs the user. For each web document that contains a query, scores are applied to each web document ID that contains the query entered by the user by applying weights, matching with the query entered by the user, and page ranking. .

Then, the searcher 132 selects a predetermined number (for example, 20,000) of web document IDs in order of high score and generates the list. Alternatively, the searcher 132 selects a predetermined number of web document IDs of a higher rank (for example, 20,000) to generate a list (S211), and then, query broker 131 for information on the list of the corresponding web document IDs. ) Is authorized.

Accordingly, the query broker 131 receives information about the list of web document IDs from the searcher 132 and applies the summary information to the summary 133 provided in the search engine 130.

Then, the summary 133 receives information about the list of web document IDs from the query broker 131 and generates summary information on the search result of the searcher 132 (that is, each searched web document) to generate the query broker 131. Is applied.

In other words, the summary 133 receives a web document ID list from the query broker 131 and uses a web document ID of the authorized web document ID list to correspond to each web document ID from the database 140. After reading (ie, searching) information on the body of the document (S212), a summary information of each web document including a query input by the user is generated according to the corresponding authorized web document ID list (S213). Then, the summary information of each generated web document is authorized to the query broker 131.

Then, the query broker 131 receives the summary information of each web document from the summary 133 and transmits the summary information of each web document to the user terminal 110 through the wired / wireless communication network 120 as a search result (S214). Thus, the user terminal 110 may provide a user with a search result received from the search engine 130, thereby displaying web documents having a very high correlation with keywords that the user really wants.

In the above-described embodiment of the present invention, a search engine inputs a keyword that can sufficiently reflect the user's search intention with respect to a search word input by the user by using the collective intelligence of the user, and is input by the user at the time of indexing. It is a general index that indexes documents containing these keywords through stemming on search terms that can be searched for, and a search index indexing documents containing suggestions that correspond to search terms that users can enter without stemming. After the index structure is formed, the search word corresponding to the search term is read when the search term is entered, the document ID containing the search term is searched in the general index, the scoring and ID list for the document is generated, and the search index The score is weighted based on whether or not the searched document ID exists. By reordering the text and providing summary information based on the reordered ID list when generating summary information on search results, you can display web documents that are highly correlated with the keywords you really want. The keyword search method has been described.

However, embodiments of the present invention are not implemented only through the above-described apparatus and / or method, but implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention, a recording medium on which the program is recorded, and the like. Such an implementation may be easily implemented by those skilled in the art to which the present disclosure pertains based on the description of the above-described embodiments.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

1 is a schematic block diagram illustrating a system for keyword search according to an embodiment of the present invention.

2 is a flowchart illustrating a keyword search method according to an exemplary embodiment of the present invention.

Claims (8)

(a) reading the recommended keyword set corresponding to the keyword to be searched by the user, (b) searching the web document containing the keyword to be searched, (c) searching a web document containing the suggested keyword, (d) generate a ranking list by scoring the web documents including the keyword to be searched, and apply a weight to the score of each web document including the recommendation keyword among the web documents including the keyword to be searched. To make the ranking list, and (e) generating a summary information of each web document including the keyword to be searched and providing it to a user according to the ranking list. The method of claim 1, (f) setting and databaseting information on the ID of each web document including the keyword to be searched, and making a database; (g) setting and recommending a database of the recommended keywords corresponding to the keyword to be searched to reflect the search intention of the user; (h) setting and databaseting information about the ID of each web document including the recommendation keyword, and (i) setting the information on the body of the web document corresponding to the web document ID and databaseizing the keyword. The method of claim 2, (j) each web document ID including the keyword to be searched, recommended keywords corresponding to the keyword to be searched, each web document ID including the recommended keyword, and a web document body corresponding to the web document ID. Periodically resetting the information about the keyword search method. The method of claim 2, Step (a) reads the set recommended keyword corresponding to the keyword input by the user from the database, In step (b), the ID of each web document including the keyword to be searched is searched from the database. Step (c) is a keyword retrieval method for retrieving the ID of each web document containing the recommended keyword from the database. The method of claim 4, wherein In step (d), Scoring each web document including a keyword input by the user according to the degree of matching with the keyword input by the user or page ranking; Determining whether there are web document IDs including the recommended keyword among web document IDs including the keyword input by the user; If there is a web document ID including the recommendation keyword, adding a weight to a score of each web document corresponding thereto; Readjusting the score of each web document including the keyword entered by the user; And generating a list of predetermined number of web document IDs in order of increasing the readjusted score. The method of claim 5, In step (e), Reading information about a body of a web document corresponding to each web document ID of the web document ID list from the database; Generating summary information of a web document corresponding to each web document ID of the web document ID list according to the read web document body; And transmitting the summary information of each generated web document to a user terminal according to the web document ID list and providing the result to the user as a search result. The method of claim 4, wherein In step (d), Determining whether there are web document IDs including the recommended keyword among web document IDs including the keyword input by the user; If there is a web document ID including the recommendation keyword, applying a weight to each web document corresponding thereto; Performing scoring based on whether the weight is applied to each web document including the keyword input by the user, matching degree with the keyword input by the user, or page ranking; And generating a list of predetermined number of web document IDs in order of increasing score. (a) setting and databaseting information about the ID of each web document containing keywords to be searched by the user; (b) setting a database by recommending keywords that reflect a user's search intention in response to the keyword to be searched; (c) setting database information on the ID of each web document including the recommendation keyword to make a database; (d) setting information about a body of a web document corresponding to the web document ID and making a database; (e) retrieving the ID of each web document including the keyword entered by the user from the database when the user searches for the keyword, (f) reading a recommended keyword corresponding to the keyword input by the user from the database, (g) retrieving from the database an ID of each web document containing a recommended keyword corresponding to the keyword entered by the user; (h) determining whether there are web document IDs including the recommendation keyword among web document IDs including the keyword input by the user; (i) applying a weight to each web document corresponding to the web document ID including the recommendation keyword, (j) performing scoring based on whether the weight is applied to each web document including the keyword input by the user, the degree of matching with the keyword input by the user, or page ranking; (k) generating a predetermined number of web document IDs in a list in order of increasing score; and (l) generating a summary information of each web document including the keyword input by the user according to the generated web document ID list and providing the same to the user.

Images