KR101026986B1 - Method for searching information based on language expression analysis and method for providing information - Google Patents

Abstract

An information retrieval method and an information providing method based on language expression interpretation are provided. In this information retrieval method, the input search word is divided into meaningful words, and the information is searched using a search key generated by analyzing what roles the separated words play in the sentence. As a result, it is possible to provide a retrieval method having a much higher reproducibility and accuracy of information than the conventional retrieval method of retrieving a search word using a combination of logical operators AND, OR, and NOT.

Search, stemming, parsing, intent

Description

Method for searching information based on language expression analysis and method for providing information}

The present invention relates to a multifunctional information retrieval method and system that provides information and searches for words, phrases and sentences of a natural language as a search word, and provides accurate information and retrieval according to a user's intention through language expression analysis. .

Recently, there is a high demand for development of convergence technologies of information and contents providers, consumers, and various information providing media. For example, convergence technology of wired / wireless broadcasting communication technology, integrated information web service technology, and development of user interface technology convenient for various user layers are representative ones. The specific contents of these technologies are to automatically interpret the provided or provided contents, the quality and meaning of the information with a computer, so that the information contents are presented and provided in a convenient and various manner according to the taste and intention of the information contents user. Market technology is expected.

For example, Korean Patent Publication No. 10-2006-0043333 (system and method for determining the intent of data and responding to data based on the intent) is a system for facilitating data handling, comprising: a component for receiving data; For purposes of a system comprising a data manager for determining an intent of the data, reorganizing at least a subset of the data based on the intent, and automatically presenting the reorganized data to a user, A system for analyzing the data, classifying the data into one or more sets having related features, and extracting features from at least one set of the data sets to form a subset of the data; The meta of the data, to facilitate the data analysis Uses at least one of data, attributes, content, context, keywords, history, heuristics, reasoning, rules, demarcation, time of day, day of the week, associated handling costs, associated handling gains, and sources. A system that is extracted based at least in part on one of syntactic, syntactic, linguistic, and linguistic properties, wherein the linguistic features include words that occur in sentences, double-word words that occur in sentences, and An object is to include at least one of the word trigrams generated in a sentence.

Korean Patent Publication No. 2002-0028593 (harmful word blocking method) is a step in which a computer user enters member information and accesses a management server, and the computer user registers / transmits the information input screen displayed by the management server. Inputting information, the harmful word blocking module of the management server determining whether the harmful word is included in the information input by the computer user, and the management server is configured to determine whether the harmful word blocking module Characterized by the harmful word blocking method characterized in that the connection to the user, or the step of registering / transmitting the information, characterized in that the definition of the harmful word refers to slang, profanity, sexual insults or personal insults. If the message you send does not contain harmful words, please It consists of a processing module in progress. The patent is based on the presence of harmful words in the message on the chat.

Korean Patent Publication No. 10-2006-0062300 (a multi-step text filtering method for blocking harmful sites) divides the text to be used as a material for blocking pornographic sites into at least one set by dividing it into at least one set and then database And collecting text from a web site being accessed and comparing the text with text in the set to perform at least one multi-step filtering to determine whether to block the web site being accessed. Multi-level text filtering method for blocking harmful sites is to block harmful sites according to the degree of lewdness in the text.

Korean Patent Laid-Open Publication No. 10-2006-0087735 (system and method for providing improved spam message filtering) is a system for providing spam message filtering, wherein the wireless network receives a message for transmission from a sender terminal to a user terminal. The server, the use cases of the natural language sentences, and accessory information on these use cases are stored, and a large-scale language database (Corpus DB) which can be classified into categories according to each use purpose, and the corpus DB in the wireless network server. A lexical analysis unit that detects a reception and extracts at least one noun type keyword from a sentence included in the received message, and selects any one of the extracted noun type keywords and sets the user to a spam message In sentences contained in each of the user categories in the Corpus DB. A corpus search unit for searching spam frequencies used and a general frequency used in sentences included in a category other than the user category, and when the at least one spam frequency and the general frequency are input, the selected noun keyword is entered. Probability calculation unit for calculating the probability that the included sentence is included in the user category, and control the lexical analysis unit and the corpus search unit to calculate the spam message inclusion probability for each noun type keyword from the probability calculation unit, calculate And a filtering control unit for calculating whether or not the received message is a spam message by calculating statistical values of the probabilities, and transmitting the determination result to the wireless network server. Assess and analyze the spam of a message, A methodology that kind.

Korean Patent Publication No. 10-2008-0000416 (harmful message filtration system and filtering method thereof and a recording medium recording the same) has a database storing messages received from clients connected through a network, and a system for filtering harmful messages. The method may include: a message receiving means for receiving the message, a word extracting means for extracting a plurality of words from the received message, and determining whether the message is harmful by using the extracted plurality of words, and storing the message in the database; Means for evaluating whether the message is harmful using the word evaluation value stored in the database and subdividing the message into a plurality of harmful message classifications. Of the message A technology to classify it or not.

Korean Patent Registration No. 10-0484944 (morphological semantic automatic tagging device based on local syntactic relation and semantic air dictionary) is a basic syntax structure generation rule, dependency syntax rule, stemming semantic electronic dictionary, semantic framework and Short phrase recognition means for recognizing short sentences by means of lexical phrases, semantic frame and lexical phrases Means of restoring the auxiliary fire using auxiliary air dictionaries, lexical argument using lexical phrase air dictionaries Determining the meaning of nouns in noun phrases and compound nouns in a noun phrase using a noun phrase meaning determination means, and a noun phrase that consists of nouns / investigations / nouns and noun / noun air information Means for determining the meaning of a noun phrase, comprising a morpheme based on local syntax and meaning By using the correct syntax air information context information than other existing ambiguity it means that devices using the syntax structure information from the linguistic point of view on auto-tagging devices presents a way to resolve the ambiguity meaning of the noun.

Korean Patent Registration No. 10-0757951 (a search method through stemming analysis of a web page) is a client of a computer, a notebook, a mobile phone, a PDA, and the like, and an analysis server for providing information to the client, wherein the client user taps When the web site is accessed through a browser, the content of the connected site is displayed in a tab browser, and the analysis server analyzes the context of the content displayed in the tab browser of the client through context morpheme analysis. Identifying a frequency, selecting a sentence having a frequency above a reference value among the identified words, selecting a word as a keyword as an important word, outputting the selected important word as a title of each tab browser, Search using the title as a query of the portal site, and search results in each tab The present invention relates to a search method through morphological analysis of a web page comprising the step of outputting the contents of a web page. The patent is a methodology of a search system that calculates the number of sentences and words appearing in text by morphological processing. How to output it as a tab web browser title as a keyword.

Korean Patent Registration No. 10-0691400 (a method for analyzing a morpheme using additional information and a morpheme analyzer for performing the method), in a morpheme analysis method, includes a morphological analysis target and a morphological analysis target from a search index data. Obtaining an associated additional information, generating a key based on the additional information, and performing a morphological analysis on the morphological analysis object using the key. It is a way.

Korean Patent Publication No. 10-2007-0029389 (a method for providing an advertisement service using a core keyword, a system and a recording medium on which a program for implementing the same) is recorded is provided to a digital processing apparatus to provide a keyword advertisement service using a core keyword. A program of instructions that can be executed by a program is tangibly embodied and can be read by a digital processing device. The method includes extracting keywords by analyzing the morphemes of content texts, and key keywords through interfacing with an advertisement keyword DB server. Determining an advertisement, receiving an advertisement list matching the core keyword through interfacing with an advertisement DB server, and inserting at least one of advertisements included in the advertisement list into the content text. Recording media.

Korean Patent Publication No. 10-2006-0011333 (Regional Information Providing System and Method through Message Analysis) is a database for storing advertisement information for each region and industry, message recognition means for analyzing stored short messages or text messages. Requesting the message recognition means and the voice recognition means to store the voice recognition means for analyzing the stored voice message, the location information collecting means for confirming the current location of the subscriber station, and the message related to the service subscriber. Confirming the message analysis result transmitted from the recognition means and the voice recognition means, optionally request the current location information of the subscriber station to the location information collecting means, and desired according to the message analysis result to deliver the desired information to the subscriber terminal Services that provide industry and local information Message generation means for retrieving information from the database using the business type and area information transmitted from the service control means and the service control means, and including the search result in a short message and transmitting the result to the subscriber station. Local information provision system through.

Korean Patent Publication No. 10-2007-0015752 (Advertising-induced messaging service system and method thereof) is an advertisement-induced messaging service system that detects an advertisement-inducing identifier by analyzing a message transmitted from a calling terminal, and then designates a predetermined advertisement message. After transmitting to the calling terminal, and the message server for transmitting the contents of the message to the receiving terminal.

Korean Patent Registration No. 10-0775680 (Method and system for providing advertisement contents using a chat of a mobile communication terminal) is a method of exchanging messages with each other through a messenger between a mobile communication terminal and a server connected through a wireless communication network. Determining whether the mobile communication terminal connected to the server selects a virtual chat partner to be chatted with, and the server extracts a chat scenario of an advertisement content product according to user preference information of the mobile communication terminal. Transmitting a message according to a scenario rule to the mobile communication terminal, transmitting a response message to the message of the server in the mobile communication terminal, and analyzing the message of the mobile communication terminal received from the server. Advertising content information according to the intention and the above scenario rule Transmitting a message requesting the cloth permission to the mobile communication terminal; and transmitting the advertisement content information to the mobile communication terminal when the server transmits a message for recommending the advertisement content information from the mobile communication terminal. The transmitting of the message to the mobile terminal by the server may include: previously, analyzing the message of the mobile terminal by the server and adjusting a character state of the messenger according to the intention of the user. Advertising content providing method using a chat of the mobile communication terminal further comprising.

Korean Patent Registration No. 10-0597435 (Hybrid-based Question Classification System and Method in Information Retrieval and Question Answer System) is a question title processing unit that recognizes the name of the work included in the input question, a morphological analysis of the question, and an individual name. Question language analysis unit that converts individual vocabularies into meaningful codes through recognition and lexical semantic tagging process, and rules that classify the correct type of answers required by the question using meaningful LSP form codes of the questions and predefined question classification rules. Based question classification unit, statistics based question classification unit for classifying questions LSP code using statistical information constructed from learning document tagged correct question type, result of rule based question classification and statistics based question classification And a question correct answer type determiner which finally determines the type of the correct answer to the user's question using Hybrid based question classification system in information retrieval and question answering system.

Korean Patent Registration No. 10-0361166 (Information Retrieval System and Method) includes a plurality of information classified by a field, a code is assigned to each classified field, and a plurality of information included in the corresponding field in association with a code of each field. If a user who wants to search the data is coded with a database representing words of information and a user who wants to search the data is connected through a network using a communication device, and inputs a word indicating the information to be searched or indicates a field to which the information belongs, A processing device for encoding the coded words and searching the database based on the coded words to find information corresponding to the coded words, wherein the information stored in the database is classified into a plurality of information areas and the information area. Each is assigned an information classification code and contains one or more details. The detailed information areas are coded in association with an information classification code of the corresponding information area, and words having the same or similar meaning among the words representing the information are stored with the same code, and each word code These are information retrieval system methods arranged in a set order.

Korean Patent Publication No. 10-2005-0092955 (Online Advertising System and Advertising Method) includes a first storage device for storing information to be advertised, a second storage device for storing original content, and original content from a second storage device. At least one server to take a and analyze the natural language processing technique, the information to be advertised corresponding to the analysis result from the first storage device and insert some of the advertisement information into the original content and by the at least one server And a third storage device storing the content in which the advertisement information is inserted.

Korean Patent Registration No. 10-0669534 (document summary method and system using sentence abstraction and probability rule and sentence meaning analysis and expression method) includes a sentence input step of inputting and storing document contents to be summarized; A parsing step for parsing and storing the data stored in the step; A sentence main component extraction step of sequentially reading the syntax data in the step, extracting main components from each sentence, and storing the ontology data; The ontology data value in the step is compared with the cognitive upper category information, which is a predetermined classification criterion for the meaning of each of the selected main elements, and the data value identified as the abstract meaning for each of the selected main elements. Sentence abstracting step of storing; A topic selection step of reading a stored value in the step and inputting it into a probability rule program operating unit to select a plurality of topics having probability between phrases, and storing identification values for the selected plurality of topics respectively; A summary sentence preparation step of reading the corresponding values of the abstract meanings identified for the selected main elements stored in the above step, combining them into grammatical summary sentences, and storing the summary values; And a summary sentence output step of judging whether there is an output control signal and, if so, outputting and displaying the data value stored in the step through the operation of the output means.

Korean Patent Registration No. 10-0836878 (a subject or field allocating apparatus and method thereof in an information retrieval system) is connected to a user terminal, a web portal site, or a web site through the Internet to search for information and provide searched information. A subject or field assignment apparatus in a search system, the information search server comprising: a search engine for performing an information search on a document based on a query word or an index word corresponding to a document; A database for storing and managing information retrieved from the search engine; A thesaurus matching unit for extracting an index word from the original text stored in the database and performing a thesaurus matching by the index word to assign a subject or a field to the original text; A subject in an information retrieval system comprising a; taxonomy processing unit for extracting a term from a definition sentence stored in the database and assigning a subject using the extracted term and assigning a subject using a taxonomy A field allocation device.

In addition, the information retrieval system used by large information search portals such as Google and Yahoo is mainly based on the search key method of "word" unit, even though the "old" unit input is possible, Rather than the linguistic processing method, Ngram-based index keys are connected by logical operators (AND, OR, NOT) to present search results. Therefore, current technology has a limitation in improving reproducibility and accuracy of information retrieval. In particular, 'mobile information retrieval' technology is attracting attention as the next generation information retrieval field, and the new web retrieval technology is being proposed and discussed.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to expand the search key by phrase unit and sentence unit in an information retrieval system using a word unit as a search word to view a user interface of the information retrieval system. It is not only convenient, but also provides a multifunctional integrated information retrieval / provision system capable of integrated processing of information provision, search, classification, evaluation, and monitoring.

In accordance with an aspect of the present invention, an information retrieval method includes: receiving a search word; Separating the search term into meaningful words; Interpreting what roles separate words play in the sentence; Generating a search key based on the analysis result of the analyzing step; And searching for information using the search key generated in the generating step.

In the input step, the search word may be input in units of phrases and sentences.

In addition, the interpreting step refers to a syntax grammar rule dictionary in order to interpret what roles the separated words play in the sentence, and the syntax grammar rule dictionary includes a syntax, grammar, dependent grammar and vocabulary. At least one of the grammars is preferably contained.

In the separating step, it is preferable to separate the search word by parts of speech.

In addition, in the separating step, it is preferable to divide the search word by parts of speech by referring to the information on the parts of speech of each word in the morpheme dictionary.

In the search key generation step, it is preferable to generate a syntax expression obtained by analyzing the grammatical and semantic relations of the words forming the search word in the search key.

In addition, the syntax expression is preferably at least one of a logical expression, a formula-defense expression, and a syntax list item.

The interpreting may be performed even after the words are replaced with synonyms.

In addition, the information retrieval step, it is preferable to perform at least one of a complete search, a partial search, an association search, and a similarity search.

The method may further include adding an additional service to the search result in the information retrieval step.

The additional service may be a content context awareness service, and the content context awareness service may include at least one of a guide service, a trading service, an advertisement service, an education service, a counseling service, a recommendation service, an auction service, and an administrative service. It is preferable to include one.

On the other hand, the information retrieval method according to the invention, the step of collecting text; Extracting at least one of an important phrase and an important sentence of the text; Generating an index key using at least one of the key phrase and the key sentence; And generating and storing an index file used for information retrieval using the index key.

In addition, the text is preferably text included in content accessible through a network.

The information retrieval method may further include: separating the text collected in the collecting step into meaningful words; Extracting important words from the separated words; And analyzing what roles the important word plays in a sentence, wherein the extracting step comprises at least one of an important phrase and an important sentence of the text, based on an analysis result of the important word and the interpretation step; It is preferable to extract.

In the extracting step, it is preferable to extract the noun phrases containing the important words before and after as important phrases, and extract the words appearing in the sentence together with the important words as important sentences.

In addition, the important word is a word having a high frequency of appearance in the text among the separated words.

And, the interpreting step, to interpret what role the important words play in the sentence, refers to a syntax grammar rule dictionary, the syntax syntax rule dictionary, a syntax structure grammar, grammar, dependent grammar and vocabulary At least one of the grammars is preferably contained.

In addition, in the separating step, it is preferable to separate the text by parts of speech.

In the separating step, it is preferable to divide the text by parts of speech by referring to the information on the parts of speech of the words that are databased in the morpheme dictionary.

In addition, the index key generation step may generate a syntax expression obtained by analyzing the grammatical and semantic relations in the important phrase and the important phrase and the important phrase in the important phrase and the important phrase as the index key. It is preferable.

In addition, the syntax expression is preferably at least one of a logical expression, a formula-defense expression, and a syntax list item.

In addition, the interpretation of the grammatical and semantic relations in the important phrases and the important phrases in the important phrases and the important phrases, the important words contained in the important phrases and important sentences It is also desirable to perform phrases and sentences replaced with synonyms.

The information retrieval method may further include determining an intention of the creator of the content including the text using the separated words.

In the identifying step, it is preferable to identify the intention of the person who created the content by referring to semantic information on what meaning the words in the database are used in the morpheme dictionary.

In the identifying step, it is preferable to grasp the intention of the creator of the content through qualitative and quantitative analysis of semantic information of the words constituting the text.

In addition, it is preferable to grasp the intention of the person who created the content for each sentence constituting the text, and to grasp the intention common to a plurality of sentences as the intention of the person who created the content.

The information retrieval method may further include providing a result of determining the intention of the person who created the content for each sentence constituting the text in a table form or a graph form.

The information retrieval method may further include: obtaining a band foreign language for the important word; And registering the band foreign language.

As described above, according to the present invention, it is possible to search at the phrase and phrase level, so that 'search word' is a keyword and another search word is expanded as its logical operator (AND, OR, NOT). The reproducibility and accuracy of the information is much higher than that of the conventional retrieval method which searches by using a combination of.

Hereinafter, with reference to the drawings will be described the present invention in more detail.

1 is a diagram illustrating a search system according to an exemplary embodiment of the present invention. The blocks constituting the search system according to the present embodiment can be implemented in S / W as well as in H / W. In addition, some of the blocks constituting the search system may be implemented in S / W and the others in H / W.

As shown in FIG. 1, the search system according to the present embodiment includes a text crawler 101, a text stemmer 102, a stemmer dictionary 103, a content provider intention determiner 104, and an information monitor 105. , Key word extractor 106, text parser 107, syntax grammar rule dictionary 108, key phrase extractor 109, verb extractor 110, index key generator 111, index file constructor 112 , Index file store 113, search term inputter 114, search term stemmer 115, searcher intent determiner 116, search term parser 117, search key generator 118, searcher 119, Search result generator 120, supplementary service generator 121, supplementary service database 122, search result indicator 123, band foreign language obtainer 124, foreign language search engine connector 125 and information search window 126 It is provided.

The text crawler 101 collects text included in content accessible through a network. The text collected by the text crawler 101 is passed to a text stemmer 102, which will be described later.

Here, the network includes not only wired networks such as the Internet and intranets, but also mobile networks and wireless networks such as WLAN and WMAN. In addition, the content may be anything that is accessible through the network. In FIG. 1, a content or a storage in which a content is stored includes an image 11, an email 12, a database 13, a file system 14, a website 15, a blog 16, and a document storage 17. ), But other contents or storage may be applied.

In the morpheme dictionary 103, the part-of-speech information of each word is made into a database. In addition, in the morpheme dictionary 103, extended information such as semantic information of each word, synonym information, and band foreign language is also made into a database. Here, since there is no restriction on the words registered in the morpheme dictionary 103, the parts of speech other than the nouns (pronouns, adjectives, investigations, endings, etc.) other than the nouns may be registered.

The part-of-speech information contained in the morpheme dictionary 103 is referenced to separate text by parts-of-speech, and the semantic information is used to grasp the intention of the content creator or searcher. In addition, synonym information is used to perform a wider range of extended searches, and band foreign languages are used to perform multilingual searches.

Since the information stored in the morpheme dictionary 103 is illustrated in FIG. 2, the morpheme dictionary 103 will be described in detail later with reference to FIG. 2.

The text stemmer 102 separates the text received from the text crawler 101 into units of meaningful words. Here, the term 'sense of words' refers to a part-of-speech, and accordingly, the text morph processor 102 may be understood to divide the text into parts of speech.

To this end, the text morpheme processor 102 refers to the part-of-speech information on words that are databased in the morpheme dictionary 103 to separate text by parts of speech.

The content provider intention determiner 104 determines the intention (eg, friendly or unfavorable) of the creator of the content including the text by using the words separated by the parts of speech in the text morph processor 102.

On the other hand, the semantic information on the words databased in the morpheme dictionary 103 includes information on what meaning the word is used. Accordingly, the content provider intention determiner 104 refers to the semantic information contained in the morpheme dictionary 103 to determine the intention of the content provider.

In detail, the content provider intention determiner 104 determines the intention of the content provider through qualitative analysis and quantitative analysis of semantic information of words constituting the text included in the content. It will be described later.

The information monitor 105 outputs the intention of the content provider identified by the content provider intention determiner 104 for the user to see. To this end, the information monitor 105 may use graphical tools.

Meanwhile, the important word extractor 106 extracts an important word from words separated from the text by the text morpheme processor 102. In detail, the important word extractor 106 may extract a word having a high occurrence frequency among the words separated from the text by the text morph processor 102.

The text parser 107 interprets the important words extracted by the key word extractor 106 in a sentence. For this interpretation, the text parser 107 refers to the grammar rules contained in the syntax grammar rule 108.

The key phrase extractor 109 extracts noun phrases containing important words as key phrases based on the syntax analysis result of the text parser 107. Here, the key phrase extracted by the key phrase extractor 109 is a noun phrase in which important words are recorded before and after.

The verb extractor 110 extracts a verb (air word) appearing in a sentence together with an important word extracted by the important word extractor 106 along with an important word.

The index key generator 111 constructs a syntax for the 'important word' extracted from the key phrase extractor 109 and the 'important word and verb' (hereinafter, referred to as 'important sentence') extracted from the verb extractor 110. Perform the analysis.

Here, syntactic analysis is a process of interpreting what grammatical / meaning relationships the words constituting the 'important' and 'important sentences' have within the phrase / sentence. As a result of parsing, a syntax expression is obtained, which can be expressed as a logical expression, a formula-defense expression, or a syntactic list expression.

By such a procedure, the index key generator 111 generates a syntax expression obtained as a result of syntax analysis as an index key. In this case, the index key generator 111 may parse phrases and sentences in which important words and important sentences included in important phrases are replaced with synonyms, and generate index keys for the phrases.

The index file configurator 112 links the index key generated by the index key generator 111 to content information related to each index key to generate an index file, and stores the index file in the index file storage 113. Since the index file generated by the index file configurator 112 is illustrated in FIGS. 5 and 6, the index file will be described in detail in the detailed description of FIGS. 5 and 6.

The configurations described so far are the configurations used to generate the index keys needed for information retrieval. Hereinafter, when the user requests a search by inputting a search word, the configuration of performing an information search using the index file generated and stored above will be described in detail. Here, the search word input by the user includes not only words but also phrases or sentences.

The information search window 126 provides an input window through which a user can input a search word as a word, phrase, or sentence. In addition, the search term inputter 114 transmits the search term input by the user to the search term stem processor 115 through the information search window 126.

The search term stem processor 115 divides the search term received from the search term inputter 114 into units of meaningful words. Here, the term 'sense of words' refers to a part-of-speech, and accordingly, the search term morpheme processor 115 may be understood as separating the search terms by parts of speech in the same manner as the text morpheme processor 102.

To this end, the search term morpheme processor 115 refers to the part-of-speech information on words that are databased in the morpheme dictionary 103 to separate text by parts of speech.

The searcher intention determiner 116 detects the intention of the searcher who inputs the search word using the words separated by the parts of speech in the search term morpheme processor 115. The searcher intention determiner 116 refers to the semantic information contained in the morpheme dictionary 103 to determine the searcher's intention.

The search term phrase interpreter 117 interprets the roles of words separated by parts of speech in the search term morpheme processor 115 in the sentence. For this interpretation, the search term parser 117 refers to the grammar rules contained in the syntax grammar rule 108.

The search key generator 118 refers to a syntax interpretation result interpreted by the search term syntax interpreter 117 and obtains a syntax expression into words separated by parts of speech in the search term stem processor 115. The search key generator 118 generates the syntax expression obtained thereby as a search key.

The searcher 119 searches the index file storage 113 for the same index key as the search key generated in the search key generator 118. At this time, the searcher 119 searches the index file storage 113 for the same index key as the search key.

Meanwhile, the searcher 119 replaces the words constituting the search key generated by the search key generator 118 with synonyms, and performs an extended search for searching the index file storage 113 for the same index key as the generated phrase / phrase. Of course it is also possible to perform.

In addition, the searcher 119 can search not only an index key having the same structure and words as the search key, but also an index key having only a part of the structure and words. That is, the searcher 119 can perform a partial search or a related search as well as the same search.

The additional service database 122 stores a site information or a related advertisement service suitable for a searcher's search intent.

The additional service generator 121 generates an additional service related to the search key. To this end, the supplementary service generator 121 uses an supplementary service suitable for a searcher's search intention among the supplementary services stored in the supplementary service database 122.

The additional service provided by the additional service database 122 and the additional service generator 121 may be implemented as a content context awareness service. In this case, the content context-sensitive service includes a guide service, a trading service, an advertisement service, an education service, a counseling service, a recommendation service, an auction service, an administrative service, and the like.

The search result generator 120 adds the additional service generated by the additional service generator 121 to the search result performed by the searcher 119. In addition, the additional service generator 121 may determine the degree of match between the search result and the searcher's search term, and add the determination result to the search result. If the search result is the same search, the matching degree is high, but in the case of partial or related search, the matching degree is lower than the same search.

The search result indicator 123 outputs the search result to which the additional service generated by the search result generator 120 is added to the user through the screen.

On the other hand, it is possible to provide only the search results to the user without additional services. In this case, the additional service generator 121 may not generate the additional service, or the search result generator 120 may not add the generated additional service even if the additional service generator 121 generates the additional service.

Meanwhile, the band foreign language obtainer 124 accesses a foreign language search engine using the foreign language search engine connector 125 to obtain a band foreign language for the important word extracted by the key word extractor 106. Then, the band foreign language obtainer 124 registers the band foreign language for the important word in the morpheme dictionary 103.

Hereinafter, the morpheme dictionary 103 mentioned above is demonstrated in detail with reference to FIG. 2 illustrates an example of a morpheme dictionary in Korean.

As shown in FIG. 2, the morpheme dictionary 103 includes 1) a word in a first column, 2) a part-of-speech information in a second column, and 3) a semantic information of a word in a third column. In column 4, a set of synonyms, synonyms, and / or representatives, 5) band English in column 5, 6) band Japanese in column 6, and 7) band Chinese in column 7. , Each listed.

In the illustrated morpheme dictionary 103, the text morpheme processor 102 is used to morph text, and the search term morpheme 115 is used to morph the search term.

As mentioned above, morphological processing is to indicate what meaningful words the elements of a given sentence consist of. For example, when stemming "go to school", it is morpheme processing to divide the word boundary as "school + goes to + goes to school."

At this time, the information to be used is the part-of-speech information stored in the morpheme dictionary 103 to examine the possible relations between each part-of-speech and as a result of processing, such as "school / noun + to / investigative + verb / verb / mother". Output

At the same time, linguistic statistics such as "word frequency" and "sentence number" can be obtained as secondary information of morphological processing, and it is possible to simultaneously output semantic information, synonym information, and foreign language.

On the other hand, the semantic information recorded in the morpheme dictionary 103 is,

1) In the case of nouns, things | nature | materials | elements | tools | tools | human relations | society | crime | organization | culture | animals | humans | plants | water | abstract | composition | | Location | Interval | Number of times | Ratio | Accuracy | Quantity | Emphasis | Authenticity |

2) In the case of verbs, they are classified into acts | instantaneous | continued | moving | change | sentiment | thinking | perception | existence | judgment | involuntary | expression

3) In the case of adjectives, it is classified into state | relationship | property | degree | highlight | crushing | favorite | chosen | choice | obscene |

4) In the case of investigation, ward | means / method | raw materials / home | cause / reason | source / origin | point of arrival / direction | purpose / aim | Result | giver | receiving | place | time | case | contents | role | Control | range / frequency | accuracy | quantity | ratio | emphasis | authenticity |

5) In the case of adverbs, it is classified as |

6) In the case of a mother indicating an aspect, the past ｜ present ｜ future ｜ negative ｜ possible ｜ self-specific ｜ duty ｜ necessity ｜ unavoidable ｜ request ｜ request ｜ assertion ｜ certainty estimation ｜ uncertainty estimation ｜ expectation ｜ hope ｜ enforcement | It is classified into permission, recommendation, ministry, referral, manual, profit, humility and respect.

However, the classification presented above is only one example and is not fixed. As the meaning of words changes from time to time, the classification of meanings is not fixed but can change according to the change of the times. In addition, when classifying semantic information, the lexical dictionary by thesaurus classification may be referred.

As illustrated in FIG. 2,

1) "Rhee Syngman (103-1)", the part-of-speech information is "noun", the meaning information is "person name", the synonym information can be seen that the "Korean President",

2) "Cold noodles 103-2", the part-of-speech information "noun", meaning information "food name", synonym information can be seen that "Korean representative noodles food,"

3) "carbon dioxide (103-3)", the part-of-speech information "noun", meaning information "harmful substance", synonym information "Shiotu, global warming",

4) "Please (103-4)", the part-of-speech information is "mother", meaning information "request, respect", synonym information is "command",

5) "Nom 103-5" can know that the part-of-speech information is "noun", the semantic information is "relax", and the synonym information is "over".

Semantic information and synonym information may be used to expand and expand the search. For example, when "Rhee Syngman" is used as a search word, it is possible to further provide a search result related to "Rhee Syngman" as well as a search result related to "Korean President".

In addition, "carbon dioxide measures" if one with the query, "Carbon dioxide measures" search results, as well as "toxic substances", "CO ₂ measures", it is possible to also provide better search results relating to "Global Warming" on Do.

Similarly, since the meaning information such as "beef, pork, kimchi" is "food" and the synonym information is "meat", "vegetables," it is possible to perform an extended search using the semantic information and synonym information.

In this way, since the morpheme dictionary 103 contains semantic information and synonym information in addition to the part-of-speech information, rich information retrieval is possible.

Hereinafter, a process of generating an expression of an index key or a search key by stemming a phrase / phrase will be described with reference to FIG. 3.

Syntactic analysis is the process of interpreting the structure of each meaningful word that constitutes a sentence. Syntactic analysis uses grammar theory to describe the natural language. Grammar theories describing natural languages include grammar, grammar, dependence, and lexical grammar.

Since the grammar theory is applicable to all languages such as English, Korean, Chinese, and Japanese, the Korean syntax interpretation methodology presented in this embodiment can be applied to other languages.

The input sentence that has completed the morpheme processing 301 shown in FIG. 3 is separated into meaningful word units, and the parts of speech information is added 302 to the separated words. Part-of-speech information includes {nouns, pronouns, investigations, adjectives, adverbs, investigations… } There are about 10-12 parts of speech.

In order to interpret how each word is combined with each other to form correct phrases, clauses, and statements, a generation rule (303) for the correct phrases, phrases, and statements in each language is required. 108, it is extracted from the syntax grammar rule dictionary (108).

For example, the grammatical rule of knowledge that "a sentence consists of a housewife (noun phrase) and a predicate phrase (verb phrase)" is expressed as S (Sentence) ← NP (Noun Phrase) and VP (Verb Phrase). Where these rules are defined is the creation rule 303 for phrases, clauses, and statements.

To explain these rules briefly, 1) the grammatical form is defined based on the rules that are generated from word units to phrase, clause, and sentence units. Mathematical relations are defined as rules. 3) The lexical grammar form takes the form VP = cat (verb, end,…, Sub, SEM) and the correct combination between each word is performed by unification.

As a specific example, the rules of grammar and how to handle them

(1) S ← PPV

(2) PP ← NP

If you define and parse as "I'm from Seoul," the words of the morphological results (Seoul / N (noun), E / P (probe), and / V (verb) are generated), and the parsing is performed. When "in Seoul" is "NP", "in Seoul" is interpreted as "PP (prepositional phrase)" according to the grammar rule (2), and "come" is "V (verb)", so PP and V are grammar rules. According to (1), it is interpreted as 'S (Sentence)' and the result is output in Korean as a tree or list structure.

On the other hand, when the above example sentence is interpreted as a dependent grammar rule, the phrase "synonym from Seoul" is output as an expression "coming from Seoul".

And, the lexical grammar form outputs the result of syntax interpretation (come, [who], [in Seoul]) as a semantic expression. In the present invention, each syntax expression may be output 304 according to a grammar theory. Accordingly, the expression 305 based on the part-of-speech information, the expression 306 based on the semantic information, and the expression 307 based on the synonyms / phrases can be output, which are used as the aforementioned index key and search key.

Hereinafter, a search method using the search system shown in FIG. 1, specifically, a process of generating an index key and indexing the file, and performing a search on a search word input by a user using the index file, FIG. It will be described in detail with reference to 4 to 8.

4 is a flowchart provided to explain a search method according to another embodiment of the present invention, and FIGS. 5 and 6 are views provided to further explain the search method illustrated in FIG. 4.

First, the text crawler 101 designates the content for which the search key is to be secured (S401), and designates the text for which the search key is to be secured among the texts included in the content (S402), and extracts the specified text (S402). S403).

5A illustrates an example of text extracted through steps S401 to S403.

Thereafter, the text morpheme processor 102 separates the text extracted by the text crawler 101 by sentence, and separates the separated sentences by parts of speech (S404). The morpheme processing in step S404 is performed with reference to the morpheme dictionary 103.

Thereafter, the text morpheme processor 102 outputs the result of adding the part-of-speech information to the words separated in operation S404 together with the number of sentences and the number of words of the text (S405).

In FIG. 5B, the text is divided into parts of speech, and the results are expressed in units of words. In FIG. 5C, the number of words and sentences in the text are displayed together with the separated words and parts of speech. The results are shown. According to FIG. 5C, it can be seen that the text shown in FIG. 5A is composed of 127 words including punctuation in five sentences.

Thereafter, the key word extractor 106 extracts a key word based on the output in step S405 (S406). Extraction of important words in step S406 can be performed by a quantitative statistical calculation method, which is as follows.

First, formula tfidf (w _i , t) = freq (w _j , t) * idf (w _j ).

Here, freq (w _j , t) represents the frequency of occurrence of the word w _j in the text t and is defined as idf (w _i ) = log (n / uf (w _j )). N represents the number of sentences and uf (w _j ) represents the number of sentences containing the word w _j .

Then, tfidf (w _i , t) for all of the words included in the text is calculated, and the word having the highest calculated value is the important word.

When the above equation is applied to the result shown in (c) of FIG. 5, since "kimchi" is extracted as an important word, the important word is shown in FIG.

Thereafter, the text syntax interpreter 107 performs syntax analysis on the sentence including the important word extracted in step S406 (S407).

Then, the key phrase extractor 109 extracts key phrases that are noun phrases containing key words by using the result of the syntax analysis performed in step S407 (S408).

The verb extractor 110 extracts a verb (air word) appearing in the sentence together with the important word extracted in step S408 together with the important word (S409).

The 'important request' and 'important sentence' (important words and verbs) obtained as a result of performing the steps S408 and S409 are shown in FIG.

Among the phrases shown in (e) of FIG. 5, "kimchi imported", "kimchi imported amount of Korea", "exported kimchi", "imported kimchi", "expensive kimchi", "cheap kimchi" corresponds to an important phrase. (Kimchi imports) surge, (kimchi imports) surge, (expensive kimchi) exports and (cheap kimchi) imports, respectively, important phrases "kimchi imports surged", "kimchi imports surged", "expensive kimchi Export, "and" import cheap kimchi. "

Thereafter, the index key generator 111 parses the 'important statement' extracted in step S408 and the 'important sentence' extracted in step S409 (S410), and generates a syntax expression obtained as a result of the parse as an index key. (S411).

Then, the index file configurator 112 links the content information associated with each index key to the index key generated by the index key generator 111 to generate an index file (S412), and the index file storage 113 Store in (S413).

In Figure 5 (f), the syntactic expression obtained when performing the syntax analysis in the dependent syntax in step S410, kimchi → import, Korea → kimchi imports, exported → kimchi, import → kimchi, expensive → kimchi, cheap → The result of generating Kimchi with an index key is shown.

In addition, (g) of FIG. 5, (kimchi (import)), (kimchi (import amount), (exported (kimchi)), which are the syntax expressions obtained when the syntax analysis is performed in the old structure syntax in step S410, ( The results of generating index keys for import (kimchi)), (expensive (kimchi) export), (cheap (kimchi) import), etc. are shown.

5 (f) and (g), it can be seen that the content information related to the index key is linked. The appearance frequency, the appearance sentence, and the appearance position information of the index key correspond to this.

On the other hand, it is also possible to generate another index key having the same meaning as the index key by using the semantic information and to store the index file for them. An index file listing the resulting index keys is illustrated in FIG.

Hereinafter, a process of searching for a search word input by a user using an index file generated and stored in the above manner will be described in detail with reference to FIGS. 7 and 8.

7 is a flowchart provided to explain a search method according to another embodiment of the present invention, and FIG. 8 is a view provided to further explain the search method shown in FIG. 7.

First, when a user inputs a search word through the information search window 126 (S701), the query morpheme processor 115 separates the search word input in parts S701 and outputs it in word units, but the part-of-speech information for each word is used. Also output (S702).

In FIG. 8, it is assumed that a search word input by a user, “exported kimchi,” 801 is input, and the result of morphological processing is “exported / verb, b / mother, kimchi / noun” 802 and semantic information. Represents “verb / export, mother / past, noun / food” (803).

Subsequently, the search term syntax interpreter 117 performs syntax analysis with the output in step S702 (S703), and the search key generator 118 generates a search key by referring to the syntax analysis result performed in step S703 (S703). S704).

The "exported → kimchi" 804 and the "exported → food" 805 shown in FIG. 8 represent search keys that can be generated in step S704. The former is a search key represented by an expression by part-of-speech information, and the latter is a search key represented by an expression by semantic information.

Thereafter, the searcher 119 searches the index file storage 113 for the search key generated in step S704 (S705).

The search in step S705 includes a full search (same search), a partial search, a related search, and a similar match search for the purpose of increasing information reproducibility.

Here, "complete search" refers to a search by correct syntax interpretation of "exported → kimchi" (805) (806).

On the other hand, utilizing the meaning information of "kimchi" as "food" and using the search key "exported → food" (807), all the text information with the meaning information as "food" can be obtained as a search result. . That is, "beef" and "pork" means that if the meaning information is all "food", as well as "exported → kimchi" in the entire text with the search key "exported → food" (807), as well as "exported → beef All information given as "food" such as "exported pork" can be obtained as a search result, which allows for more extensive information search, which corresponds to 'associated search'.

In addition, it is also possible to perform a partial search using only a partial search key, that is, "kimchi" 808 and "export" 809 in "exported-to-kimchi".

Further, by utilizing synonym information stored in the morpheme dictionary 103 (see FIG. 2), it is also possible to change the "exported" to "outbound" and search.

That is, with reference to the synonym information of the morpheme dictionary 103, it is possible to search using a search key such as "going out → kimchi", "going out → food", which corresponds to the "current search".

As in the present embodiment, when the index key is extended and generated in various ways, an effect superior to the conventional method can be obtained in information reproducibility and accuracy. In the existing search system, 'search word' is a keyword, and as an extension thereof, another search word is searched using a combination of logical operators (AND, OR, and NOT). As a sentence "level search method, the search method according to the present embodiment has a much higher reproducibility and accuracy of information.

Hereinafter, a method of determining the intention of the content provider by the content provider intention determiner 104 illustrated in FIG. 1 with reference to the morpheme dictionary 103 will be described in detail with reference to FIG. 9.

9 is a diagram illustrating a process of identifying the intention of the content provider who has created the "review" content.

In FIG. 9, the text 901 is collected by the text crawler 101, and the part-of-speech information and / or semantic information is output along with the words, the number of words, and the number of sentences while the stemming process 902 is performed by the text morpheme processor 102. The result (903) is shown.

The content provider intention determiner 104 constructs the determinant 904 using the information contained in '903'. The determinant 904 is a sentence-by-state information on what semantic information is stored.

Thereafter, the content provider intention determiner 104 analyzes and evaluates the semantic information contained in the created determinant 904 for each sentence to determine the number of friendly sentences and the number of hostile sentences. Specifically, sentences with a high frequency of crush and respect are regarded as friendly sentences, and sentences with high frequency of affinity, negativeness, and abusive language are identified as hostile sentences.

According to the determinant 904 of FIG. 9, since the number of sentences having a high frequency of "favorite" and "respect" among the six sentences is five, the text 901 is evaluated as being extremely friendly, and an icon indicating this ( It is also possible to output 906.

FIG. 10 is a diagram illustrating a process of identifying an intention of a content provider who has created the "comment" content. By this process, it is possible to evaluate and classify whether the "comment" which is the intention grasping target is a sunfleet or an affliction.

In FIG. 10, the text 1001 is collected by the text crawler 101, and the part-of-speech information and / or semantic information is output along with the words, the number of words, and the number of sentences while the stemming process 1002 is performed by the text morpheme processor 102. The result (1003) is shown.

The content provider intention determiner 104 creates a determinant 1004 using the information contained in '1003'. The content provider intention determiner 104 analyzes and evaluates the semantic information contained in the created determinant 1004 for each sentence to determine the number of friendly sentences and the number of hostile sentences.

According to the determinant 1004 of FIG. 10, since the number of sentences having a high frequency of "unfavorable" and "depressed" among four sentences is four sentences, the text 1001 is evaluated as extremely hostile.

According to this, it is possible to perform a monitoring function that can analyze, evaluate and classify friendly and unfriendly product reviews in real time.

Meanwhile, the content provider intention determiner 104 may generate information indicating how distributed words having specific semantic information are in sentences, and output the generated distribution information by the information monitor 105. .

Here, the generated distribution information may be in the form of a table 1005 as well as in the form of a graph 1006.

On the other hand, by modifying the present embodiment, it is possible to extract the information provider intention to filter the harmful information, spam mails, and the like. To this end, "sexually explicit material", "harmful information" or "spammy", etc., are recorded as the meaning information in the words most frequently appeared in youth harmful information or spam mails, and the harmful information and spam mails are automatically processed through the intention grasping procedure. It is possible to perform filtering functions such as evaluation, automatic classification and automatic deletion.

For example, the above meaning information is given to words such as "agent driving", "lowest loan", "hit game", "internet minimum", "plastic surgery", and "viagra" that frequently appear as words of spam mail. In this case, harmful information can be evaluated and classified.

FIG. 11 is a diagram of a geographic guidance service stored in the supplementary service database 122 and an example of an additional service generated by the supplementary service generator 121.

To this end, in FIG. 11, it is assumed that the sentence “Nearly delicious Italian restaurant” 1101 is input through the information search window 126.

The input sentence 1401 is morphologically processed by the search term stem processor 115, and is separated into "near neighborhood / delicious / island / italy / restaurant / silver" (1102), and the "near neighborhood / to which semantic information is added." Noun / Current Location, Delicious / Adjective / Food Evaluation, Silver / Investigation / Formula, Italy / Noun / Country, Restaurant / Noun / Restaurant, Silver / Investigation / Presentation "(1103) is output.

Then, the searcher intention determiner 116 detects the searcher's intention through the morpheme processing result. In detail, the searcher intention determiner 116 recognizes the search request as the contextual information of the word 'silver' indicates 'present', and the contextual word "nearby". Is understood to mean that the current location that can be identified by the GPS should be acquired, and the Italian restaurant is intended to require a search of a restaurant database.

Accordingly, the supplementary service generator 121 determines the current position using the GPS (1104), the searcher 119 performs a search for a delicious Italian restaurant (1105), and the supplementary service generator 120 maps the map. In operation 1106, a map displaying a delicious Italian restaurant searched for is displayed on a map showing a current location (1107).

12 is a diagram provided to explain a method of providing another type of additional service. The illustrated additional service is to provide various additional information related to a specific word and a subject included in the text output as a search result. To this end, various kinds of information such as organization information 1201, advertisement 1202, person information 1203, topic topic information 1204, company information 1205, etc. stored in the additional service database 122 may be utilized. Of course, it can be obtained from other databases connected through the network.

FIG. 13 illustrates a process in which the band foreign language obtainer 124 accesses a foreign language search engine using the foreign language search engine connector 125 to obtain a band foreign language for an important word extracted by the key word extractor 106. It is a drawing used for describing in detail.

The band foreign language for the key word extracted by the key word extractor 106 may be obtained through the morpheme dictionary 103. Also, using the obtained band foreign language, as shown in FIG. 13, information search is also performed on foreign languages such as English search 1302, Japanese search 1303, and Chinese search 1304 through multilingual cross search using a foreign language search engine. can do. Foreign language information search engines can use existing search engines such as Google or Yahoo.

14 is a diagram illustrating an example of an information providing / search window provided with an information search window and an information providing window. In the illustrated information providing / search window, input window 1401 is a window used to input a search word, information providing button 1402 is a button used to register information, and information search button 1403 is an input window. It is a button used to instruct an information search for a search word entered in 1401.

The buttons 1404 provided below are buttons used to set a search method and a search result providing method. For example, the "exact search" button shown in FIG. 14 is a button used to set the search method to a perfect search, and the "ad allow" button is used to allow advertisements related to the search results to be displayed together with the search results. The "Allow related service provision" button is a button used to allow additional information on words and phrases included in a search result.

The embodiments described so far can be implemented in mobile information devices such as mobile phones as well as other information devices.

In addition, although the preferred embodiment of the present invention has been shown and described above, the present invention is not limited to the specific embodiments described above, but the technical field to which the invention belongs without departing from the spirit of the invention claimed in the claims. Of course, various modifications can be made by those skilled in the art, and these modifications should not be individually understood from the technical spirit or the prospect of the present invention.

1 is a diagram illustrating a search system according to an embodiment of the present invention;

2 is a diagram showing a Korean morpheme dictionary;

3 is a view provided in the description of the process of performing syntax analysis after stemming the phrase / sentence,

4 is a flowchart provided to explain a search method according to another embodiment of the present invention;

5 and 6 are views provided to further explain the search method shown in FIG.

7 is a flowchart provided to explain a search method according to another embodiment of the present invention;

8 is a view provided for further explanation of the search method illustrated in FIG. 7;

9 is a view illustrating a process of identifying the intention of the content provider who has created the "review" content for the content;

FIG. 10 is a diagram illustrating a process of determining the intention of a content provider who wrote a comment about content.

FIG. 11 is a diagram of a geographic information service stored in an additional service database and provided as an example of an additional service generated by an additional service generator;

12 is a view provided to explain a method of providing an additional service of a different type from FIG. 11;

13 is a view provided in a detailed description of a process of obtaining a band foreign language;

14 is a diagram illustrating an example of an information providing / search window provided with an information search window and an information providing window.

Explanation of symbols on the main parts of the drawings

101: Text Crawler 102: Text Stemmer

103: stemming dictionary 104: content provider intention grasp

105: information monitor 106: key word extractor

107: text parser 108: syntax syntax rules dictionary

109: key phrase extractor 110: word extractor

111: index key generator 112: index file organizer

113: index file storage 114: search term input

115: query stemmer 116: searcher intent grasp

117: query parser 118: search key generator

119: Searcher 120: Search Result Generator

121: additional service generator 122: additional service database

123: search result indicator 124: band foreign language obtainer

125: foreign language search engine connector 126: information search box

Claims (28)

Receiving a search term; Separating the search term into meaningful words; Interpreting what roles separate words play in the sentence; Generating a search key based on the analysis result of the analyzing step; And And searching for information using the search key generated in the generating step. The search key generation step, And a phrase expression obtained by analyzing what grammatical and semantic relations of the words forming the search word have in the search word. The method of claim 1, The input step, Information search method characterized in that for receiving the search word in phrase units and sentence units. The method of claim 1, The analysis step, To interpret what roles the separated words play in the sentence, refer to a syntax grammar dictionary, The syntax syntax rule dictionary, An information retrieval method characterized in that at least one of a grammar, a grammar, a dependent grammar, and a lexical grammar is recorded. The method of claim 1, The separation step, Information retrieval method characterized in that for separating the search word by parts of speech. The method of claim 4, wherein The separation step, An information retrieval method comprising: separating the search word by parts of speech by referring to information on parts of speech of each word in a morpheme dictionary. delete The method of claim 1, The syntax expression is And at least one of a logical expression, an expression-defense expression, and a syntactic list expression. The method of claim 1, The interpreting step, The information retrieval method, characterized in that performed even after replacing the words with synonyms. The method of claim 1, The information retrieval step, And at least one of a complete search, a partial search, an associated search, and a similarity search. The method of claim 1, And adding an additional service to a search result in the information retrieval step. The method of claim 10, The additional service is a content context awareness service, The content context perceptual is, An information retrieval method comprising at least one of a guide service, a trading service, an advertising service, an education service, a counseling service, a recommendation service, an auction service, and an administrative service. Collecting text; Extracting at least one of an important phrase and an important sentence of the text; Generating an index key using at least one of the key phrase and the key sentence; And And generating and storing an index file used for information retrieval using the index key. The index key generation step, Searching information providing method characterized in that for generating a phrase expression obtained by analyzing the grammatical and semantic relationship between the important phrases and the important phrases in the important phrases and important phrases and important sentences with the index key; . The method of claim 12, The text above is Search information providing method, characterized in that the text contained in the content accessible through the network. The method of claim 12, Separating the text collected in the collecting step into meaningful words; Extracting important words from the separated words; And Interpreting what roles the important words play in the sentence; The extraction step, And at least one of an important phrase and an important sentence of the text is extracted based on the important word and an analysis result of the interpreting step. 15. The method of claim 14, The extraction step, And extracting a noun phrase containing the important word before and after as an important phrase, and extracting a word appearing in a sentence together with the important word as an important sentence. 15. The method of claim 14, The important word, Search information providing method, characterized in that the occurrence frequency of the separated words in the text is more than the reference frequency. 15. The method of claim 14, The analysis step, To interpret what roles the important word plays in a sentence, refer to a syntax grammar dictionary, The syntax syntax rule dictionary, At least one of a grammar, a grammar, a dependent grammar, and a lexical grammar. 15. The method of claim 14, The separation step, Search information providing method characterized in that the text is separated by parts of speech. The method of claim 18, The separation step, A method for providing information for retrieval, characterized in that the text is separated for each part of speech with reference to information on parts of speech of each word in the morpheme dictionary. delete The method of claim 12, The syntax expression is Search information providing method, characterized in that at least one of a logical expression, formula-defense expression and syntax list item. The method of claim 12, The interpretation of the grammatical and semantic relations between the important phrases and the important phrases in the important phrases and the important phrases is as follows: Search information providing method, characterized in that performed for phrases and sentences in which the important words contained in the important phrases and important sentences are replaced with synonyms. 15. The method of claim 14, Identifying the meaning of the separated words by referring to the morpheme dictionary to determine the intention of the creator of the content including the text; delete 24. The method of claim 23, The grasping step, Searching information providing method characterized in that to grasp the intention of the creator of the content through the qualitative and quantitative analysis of the semantic information of the words constituting the text. 24. The method of claim 23, The method of claim 1, wherein the intention of the creator of the content is identified for each sentence constituting the text, and the intention common to a plurality of sentences is grasped as the intention of the creator of the content. The method of claim 26, And providing a result in which the intention of the creator of the content is determined for each sentence constituting the text in a table form or a graph form. The method of claim 12, Obtaining a band foreign language for an important word; And Registering the band foreign language; the search information providing method further comprising.

Images