KR102226742B1 - Information providing method and apparatus using reserved word - Google Patents

Abstract

An embodiment of the present specification relates to a method and apparatus for providing information using reserved words, and an information providing method according to an embodiment of the present specification includes storing an object-keyword association corresponding to an object item and a representative attribute keyword pair ; Storing a basic reserved word-keyword association corresponding to the reserved word and the representative attribute keyword pair by using the association weight corresponding to the representative attribute keyword and the sub-keyword pair, and the basic reserved word-sub-keyword association degree corresponding to the reserved word and the sub-keyword pair. ; Obtaining a reception reserved word selected by a user; Obtaining a reserved word-object association degree corresponding to the received reserved word and each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree; And providing an object item according to the reserved word-object association degree corresponding to the received reserved word and each object item pair. The reserved word includes an expression expressed by a degree of association with the representative attribute keywords or weights of the representative attribute keywords.

Description

Information provision method and device using reserved words {INFORMATION PROVIDING METHOD AND APPARATUS USING RESERVED WORD}

At least some embodiments of the present specification relate to a method and apparatus for providing information using reserved words.

According to a conventional search method, a user can search and find a desired web document by entering a search keyword in a search window. For example, the user may search for information on the movie "Interstellar" by entering the movie title "Interstellar" in a search box. However, if the user cannot remember the title of the movie he is looking for, a different method of providing information is required. For example, a user may try to search by entering an actor who appeared in a movie he is looking for, a director, or a producer of the movie as a keyword. Movie information sites or movie review sites often provide information on cast members along with movie information, so if you are not lucky, users may be able to search for a movie you want with keywords such as actors, directors, and producers.

However, furthermore, if information is provided based on informal language, for example, emotional language, not formal information as described above, the conventional search method cannot be used. For example, the response from conventional search engines to search terms such as "Fun movies" or "Movies you watch when you are sad" is a search for documents written by someone including the keywords "Fun Movies" or "Movies you watch when you are sad." It will be nothing more than doing. However, atypical language requires a different approach from standardized information such as actors, (standardized) movie genre, and release year. Even if someone doesn't include the keywords "fun movies" or "movies to watch when you're sad," there are actually movies that many people find "fun" or "nice to watch when you're sad." Furthermore, for other fields other than movies, a different approach may be needed to requests for information in an unstructured language.

KR 10-2009-0080822 A

An embodiment of the present specification has an object to provide a method and apparatus for providing efficient information using reserved words.

An information providing method according to an embodiment of the present specification is an information providing method of an information providing apparatus, comprising: storing an object-keyword association corresponding to an object item and a representative attribute keyword pair; Storing a basic reserved word-keyword association corresponding to the reserved word and the representative attribute keyword pair by using the association weight corresponding to the representative attribute keyword and the sub-keyword pair, and the basic reserved word-sub-keyword association degree corresponding to the reserved word and the sub-keyword pair. ; Obtaining a reception reserved word selected by a user; Obtaining a reserved word-object association degree corresponding to the received reserved word and each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree; And providing an object item according to the reserved word-object association degree corresponding to the received reserved word and each object item pair. The reserved word includes an expression expressed by a degree of association with the representative attribute keywords or weights of the representative attribute keywords.

An information providing device according to an embodiment of the present specification is a storage unit that stores an object-keyword association degree corresponding to an object item and a representative attribute keyword pair, and a basic reserved word-keyword association degree corresponding to a reserved word and a representative attribute keyword pair ; A communication unit that obtains a reception reserved word selected by a user; And a controller for obtaining a reserved word-object association degree corresponding to the received reserved word and each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree, wherein the controller comprises a representative attribute keyword and a lower level By using the association weight corresponding to the keyword pair and the basic reserved word corresponding to the reserved word and the lower keyword pair-the basic reserved word-keyword association degree corresponding to the reserved word and the representative attribute keyword pair, are stored in the storage unit, and the The reserved word includes an expression expressed by a degree of association with the representative attribute keywords or weights of the representative attribute keywords.

According to an embodiment of the present specification, a method and apparatus for efficiently providing information using reserved words may be provided.

1 is a network configuration diagram of an information providing system using an attribute language according to an embodiment of the present specification.
2 is a block diagram of a terminal 200 according to an embodiment of the present specification.
3 is a block diagram of an information providing apparatus 300 according to an exemplary embodiment of the present specification.
4 is a flowchart of a process of providing information through an information providing interface according to an embodiment of the present specification.
5 is a detailed flowchart of step 910 according to an embodiment of the present specification.
6 is a detailed flowchart of step 510 according to an embodiment of the present specification.
7 is a detailed flowchart of step 530 according to an embodiment of the present specification.
8 is a flowchart of a process of providing information according to another embodiment of the present specification.
9 is a flowchart of a process of providing information according to an embodiment of the present specification.
10 is an example of a stored object-keyword association diagram according to an embodiment of the present specification.
11 is an example of a basic reserved word-keyword association diagram according to an embodiment of the present specification.
12 is a detailed flowchart of step 940 according to an embodiment of the present specification.

Hereinafter, embodiments of the present specification will be described in detail with reference to the accompanying drawings.

In describing the embodiments, descriptions of technical contents that are well known in the technical field to which the present specification pertains and are not directly related to the present specification will be omitted. This is to more clearly convey the gist of the present specification by omitting unnecessary description.

For the same reason, some elements in the accompanying drawings are exaggerated, omitted, or schematically illustrated. In addition, the size of each component does not fully reflect the actual size. The same reference numerals are assigned to the same or corresponding components in each drawing.

Hereinafter, embodiments of the present specification will be described with reference to the drawings.

1 is a network configuration diagram of an information providing system using an attribute language according to an embodiment of the present specification.

Referring to FIG. 1, an information providing system according to an embodiment of the present specification may include a terminal 200, an information providing device 300, and a communication network 150.

The terminal 200 may be implemented as an electronic device or a similar device capable of performing communication, receiving user input, and outputting a screen, for example, a smart phone, a PDA, a tablet PC, a notebook computer, a laptop computer, and a personal computer. I can.

The information providing device 300 may be implemented as, for example, a workstation, a server, a general-purpose computer, an electronic device capable of performing other communication, or a similar device.

The terminal 200 and the information providing device 300 are connected through the communication network 150 and communicate with each other through the communication network 150.

The communication network 150 is, for example, LTE (Long Term Evolution), LTE-A (LTE-Advanced), WI-FI, LAN (Local Area Network), WAN (Wide Area Network), CDMA (Code Division Multiple Access), TDMA (Time Division Multiple Access), WiBro (Wireless Broadband), GSM (Global System for Mobile Communications), etc. It may be implemented using at least some of the communication methods developed in the past, present, or available in the future. Hereinafter, for convenience, it will be described as if the terminal 200 and the information providing device 300 directly communicate with each other without referring to the communication network 150.

Detailed operations and configurations of the terminal 200 and the information providing apparatus 300 will be described later with reference to FIGS. 2 to 8.

2 is a block diagram of a terminal 200 according to an embodiment of the present specification.

Referring to FIG. 2, a terminal 200 according to an exemplary embodiment of the present specification may include an input unit 210, a display unit 220, a communication unit 230, a storage unit 240, and a control unit 250.

The input unit 210 converts a user's input operation into an input signal and transmits it to the control unit 250. The input unit 210 may be implemented as, for example, a keyboard, a mouse, a touch sensor on a touch screen, a touch pad, a keypad, a voice input, and other input processing devices that are possible in the present, in the past, or in the future. The input unit 210 may receive, for example, a user's information provision request input and transmit it to the control unit 250.

The display unit 220 outputs a screen under the control of the controller 250. The display unit 220 may be implemented as, for example, a liquid crystal display (LCD), a light emitting diode (LED), an organic light emitting diode (OLED), a projector, and other display devices that are possible in the present, in the past, or in the future. The display unit 220 may display, for example, an interface page for providing information or an information providing result page. Depending on the embodiment, the display unit 220 may be used instead of the display unit 220 instead of the display unit 220, but instead of the display unit 220, a component using a voice output or a vibration or other method capable of transmitting information to the user may be used.

The communication unit 230 exchanges data with the information providing device 300 and/or other external devices. The communication unit 230 transmits the data received from the information providing device 300 to the control unit 250. In addition, the communication unit 230 transmits data to the information providing device 300 under the control of the controller 250. The communication technology used by the communication unit 230 may vary depending on the type of communication network 150 or other circumstances.

The storage unit 240 stores data under the control of the controller 250 and transmits the requested data to the controller 250.

The controller 250 controls the overall operation of the terminal 200 and each component. The control unit 250 transmits an information provision request and other data to the information providing device 300 according to the information input from the input unit 210 as described later, and according to the page information received from the information providing device 300. The result page and/or the interface page are displayed through the display unit 220.

The operation performed by the control unit 250 may be distributedly processed by several physically separated computing devices. Some of the operations performed by the control unit 250 are performed by the first server and other operations are performed by the second server. In this case, the controller 250 may be implemented as a total number of physically separated computing devices.

The storage unit 240 may be implemented as a total of physically separated storage devices.

When the controller 250 or the storage unit 240 is implemented as a total of several devices that are physically separated, communication between several devices may be required. Here, for the sake of simplicity, the description will be made on the assumption that the storage unit 240 or the control unit 250 is implemented as a single object.

When the terminal 200 transmits and receives data, it may be expressed that the communication unit 230 transmits and receives data according to the control of the controller 250 according to the viewpoint, or that the controller 250 controls the communication unit 230 to transmit and receive data. You can also express it.

Detailed operations of each component of the terminal 200 will be described later with reference to FIGS. 4 to 8.

3 is a block diagram of an information providing apparatus 300 according to an exemplary embodiment of the present specification.

Referring to FIG. 3, an information providing apparatus 300 according to an embodiment of the present specification may include a communication unit 310, a control unit 320, and a storage unit 330.

The communication unit 310 exchanges data with the terminal 200 and/or other external devices. The communication unit 310 transmits the data received from the terminal 200 to the control unit 320. In addition, the communication unit 310 transmits data to the terminal 200 under the control of the controller 320. The communication technology used by the communication unit 310 may vary depending on the type of communication network 150 or other circumstances.

The storage unit 330 stores data under the control of the controller 320 and transmits the data requested from the controller 320 to the controller 320.

The controller 320 controls the overall operation of the information providing device 300 and each component. When receiving an interface page request, an information provision result page request, and other data through the communication unit 310, the controller 320 loads necessary data from the storage unit 330 and generates page information. The page information is transmitted to the terminal 200 through the communication unit 310.

When the information providing device 300 transmits and receives data, it may be expressed that the communication unit 310 transmits and receives data according to the control of the controller 320 depending on the viewpoint, and the controller 320 controls the communication unit 310 to transmit data. It can also be expressed as sending and receiving.

Detailed operations of each component of the information providing apparatus 300 will be described later with reference to FIGS. 4 to 8.

According to another embodiment, instead of a page for visually providing information, data for providing information may be transmitted/received by voice or other methods.

4 is a flowchart of a process of providing information through an information providing interface according to an embodiment of the present specification.

In step 410, the control unit 320 of the information providing device 300 generates interface page information. The interface page information is information necessary to create an interface page. The interface page is a page for inducing a user's input, receiving the user's input, and transmitting it to the information providing device 300. For example, the interface page information may be in the form of an HTML document or other markup language document. In another embodiment, the terminal 200 has the form information of the interface page in advance, and only a matter corresponding to the content may be transmitted from the information providing device 300 to the terminal 200. Hereinafter, for convenience, it is assumed that interface page information or other page information is delivered in an HTML document format. However, the scope of the rights of the present specification is not limited thereto.

In step 420, the communication unit 310 of the information providing device 300 transmits the interface page information to the terminal 200.

In step 430, the control unit 250 of the terminal 200 constructs an interface page using the interface page information. For example, the controller 250 may drive a web browser to analyze an HTML document and configure an interface page in the form of a web page. A separate application may be used instead of a web browser.

In step 440, the display unit 220 of the terminal 200 displays an interface page to the user 400. The interface page may include, for example, an interface through which the user 400 requests information provision, inputs and/or selects a keyword for information provision, and sets for other information provision.

In step 450, the input unit 210 of the terminal 200 receives the selection input of the user 400 inputted through the interface page and transmits it to the control unit 250.

In step 460, the communication unit 230 of the terminal 200 transmits input information for identifying a user's selection input to the information providing device 300 under the control of the controller 250.

In step 470, the control unit 320 of the information providing device 300 generates result page information using a user's input (eg, keyword and/or other information provision setting). A preparation process for generating result page information and a process of generating result page information will be described later with reference to FIGS. 5 to 8. The result page information may be configured in, for example, an HTML document format and/or an image format.

In step 480, the communication unit 310 of the information providing device 300 transmits the result page information to the terminal 200.

In step 490, the control unit 250 of the terminal 200 constructs a result page using the result page information received by the communication unit 230. For example, the controller 250 may construct a result page by analyzing result page information in HTML format.

In step 495, the display unit 220 of the terminal 200 provides the result page to the user 400.

In the embodiment of FIG. 4, it is assumed that a page in a visual form is provided to the user, but it is also possible to provide an interface or result information through voice. In this case, an audio output unit may be used instead of the display unit 220. Instead of the visual/auditory method, another type of interface method that is currently possible or will be possible in the future may be used in the relationship with the user 400. In this case, the information providing apparatus 300 may provide the terminal 200 with information converted in a different manner instead of page information to fit the interface method.

In the embodiment after FIG. 5, the user 400 wants to be provided with information on an object of a specific field of interest that the user 400 is interested in. However, the scope of the rights of the present specification is not limited thereto.

The area of interest can be, for example, the type of object. For example, if the field of interest is'celebrity', objects corresponding to the field of interest may include'Citizen Yoo','Jaeseok Yoo', and'Taylor Swift'. For example, if the field of interest is'movie', objects corresponding to this field of interest may include'Dunkirk','Spider-Man: Homecoming', and'Superbad 3'. For example, if the field of interest is'broadcasting program', objects corresponding to this field of interest may include'Infinite Challenge','American Idol','Game of Thrones', and the like.

In the following embodiments, a document is collected to evaluate the relationship (relevance/weight, etc.) between keywords. All collected documents may be evaluated as having an equal value, or the latest documents may be evaluated as having a higher value. That is, the association between the age of the document at the time of evaluation and the keywords appearing in the document may have a negative correlation.

In the process after FIG. 5 to be described later, the value may be evaluated differently according to the freshness of the document. For example, a case where two keywords appear together in a document 1 day after the evaluation point may be evaluated with a 10 times higher association than a case where two keywords appear together in a document 10 days after the evaluation point. The age of a document may be evaluated in units of seconds/minutes/hours, or in units of days/months/years, for example. The control unit 320 extracts a partial relevance value by dividing a partial relevance value of the document evaluated before the age of the document is reflected by the age of the document, and accumulates the partial relevance between the two keywords. The degree of association of can be extracted.

The creation time of a document to check the age of a document can be determined using, for example, the posting time and/or metadata included in the document, or through periodic crawling, when a new document not found in the previous crawl is found. In this case, it may be determined that a new document was added at the time of a new crawl.

9 is a flowchart of a process of providing information according to an embodiment of the present specification.

In step 910, the control unit 320 stores an object-keyword association degree corresponding to the object item and the representative attribute keyword pair in the storage unit 330.

10 is an example of a stored object-keyword association diagram according to an embodiment of the present specification.

In the embodiment of FIG. 10, there are m number of object items (i ₁ to i _m ) and n number of representative attribute keywords (k ₁ to k _n ).

For example, the object-keyword association of the _{object item i 5} and the representative attribute keyword k _{3 is w 5,3} .

The process of step 910 may be performed according to, for example, some of the embodiments of FIGS. 5 to 8, a process similar thereto, or a process corresponding thereto. According to another embodiment, the process of step 910 may be performed as an input of an administrator or may be performed by receiving an object-keyword association degree determined in an external system through a network or a storage medium.

5 is a detailed flowchart of step 910 according to an embodiment of the present specification.

Referring to FIG. 5, in step 510, the controller 320 extracts a representative attribute keyword candidate set from the first set of documents. The controller 320 may collect, for example, keywords frequently appearing in documents corresponding to the field of interest among the first set of documents as a representative attribute keyword candidate set.

6 is a detailed flowchart of step 510 according to an embodiment of the present specification.

The control unit 320 selects keywords appearing in a document, such as an object keyword representing an object item belonging to a specific field, and keywords appearing in a document, such as a field keyword representing a specific field, as a first attribute keyword candidate set and a second attribute keyword candidate. Can be set as a set.

For example, if the field of interest to provide information service is a celebrity, the field keywords may include'celebrity','celebrity','movie actress','star','celebrity','celeb', and the like. The field keyword may be set by an administrator or recommended and set by the controller 320. After acquiring some field keywords, the controller 320 may recommend/set a similar keyword, which is analyzed to have a degree of association with the field keywords or higher than a preset value, as an additional field keyword.

If the field of interest to provide the information providing service is a celebrity, the object keyword may be an individual person belonging to the field of interest. For example, "Jae-Seok Yoo", "Taylor Swift", and "Stephan Curry" may be object keywords corresponding to celebrities in the field of interest.

To explain the relationship between the field keyword and the object keyword, for example, the field keyword may correspond to the attribute or type of the object keyword. If the field keyword indicates a set, the object keyword may indicate elements belonging to the set.

Object keywords may be set by an administrator or may be selected in a similar manner to field keywords. According to another embodiment, the controller 320 may analyze the context of the collected documents and select keywords that are determined to be elements belonging to the set indicated by the field keywords as object keywords.

Popular object keywords and non-popular object keywords can be classified according to the amount of search/collection of the corresponding object keywords. The controller 320 may search/collect a document including each object keyword, set an object keyword having a collection amount equal to or greater than a specific threshold value as a popular object keyword, and set the remaining object keywords as an unpopular object keyword.

Popular field keywords and non-popular field keywords may be classified according to the amount of search/collection of keywords in the field. The controller 320 may search/collect a document including each field keyword, set a field keyword whose collection amount is greater than or equal to a specific threshold value as a popular field keyword, and set the remaining field keywords as an unpopular field keyword. However, a threshold value for distinguishing between popular object keywords and non-popular object keywords and threshold values for distinguishing popular field keywords and non-popular field keywords may be different values. Hereinafter, for convenience, the popular object keywords and popular field keywords are collectively referred to as popular field object keywords. Also, for convenience, unpopular object keywords and unpopular field keywords are collectively referred to as unpopular field object keywords.

In a modified embodiment, only a popular field keyword or a popular object keyword may be used instead of the popular field object keyword. In a modified embodiment, only an unpopular field keyword or an unpopular object keyword may be used instead of the unpopular field object keyword.

In step 610, the controller 320 sets keywords that appear together in the same document as the popular field object keyword as a first attribute keyword candidate set.

The controller 320 may search/collect documents including popular field object keywords, and set keywords included in the collected documents as a first attribute keyword candidate set. According to another embodiment, the controller 320 may exclude a field keyword and an object keyword from among keywords included in the collected document from the first attribute keyword candidate set. In addition, the controller 320 may exclude a preset meaningless keyword, for example, a survey/article, etc. from the first attribute keyword candidate set. According to another embodiment, the controller 320 may include a keyword registered in a preset dictionary among keywords included in the collected document in the first attribute keyword candidate set.

According to another embodiment, the control unit 320 searches/collects a document containing a popular field object keyword, and searches for keywords arranged within a preset distance from a popular field object keyword or a sentence containing the keyword in the collected document. It may be included in the first attribute keyword candidate set. According to another embodiment, the controller 320 searches for/collects a document containing a keyword of a popular field object, analyzes a context, and selects keywords used for formulating/descripting a keyword of a popular field object as a first attribute keyword candidate. It can also be included in the set.

The distance between keywords or between a keyword and a sentence is, for example, the number of two keywords or sentences placed between a keyword and a sentence, the number of words placed between two keywords or a keyword and a sentence, and between two keywords or a keyword and a sentence. It may be determined based on one or more of the number of positioned words, two keywords, or the number of characters between a keyword and a sentence.

The controller 320 may first perform morpheme analysis for keyword analysis.

In step 620, the controller 320 sets keywords that appear together in the same document as the non-popular field object keywords as the second attribute keyword candidate set.

The controller 320 may search/collect a document including an unpopular field object keyword, and set keywords included in the collected document as a second attribute keyword candidate set. According to another embodiment, the controller 320 may exclude a field keyword and an object keyword from among keywords included in the collected document from the second attribute keyword candidate set. In addition, the controller 320 may exclude a preset meaningless keyword, for example, a survey/article, etc. from the second attribute keyword candidate set. According to another embodiment, the controller 320 may include a keyword registered in a preset dictionary among keywords included in the collected document in the second attribute keyword candidate set.

According to another embodiment, the control unit 320 searches/collects a document containing an unpopular field object keyword, and searches keywords arranged within a preset distance from the unpopular field object keyword or a sentence containing the keyword in the collected document. It may be included in the second attribute keyword candidate set. According to another embodiment, the control unit 320 searches for/collects a document containing an unpopular field object keyword, analyzes a context, and selects keywords used for formulating/descripting the unpopular field object keyword as a second attribute keyword candidate. It can also be included in the set.

The distance between keywords or between a keyword and a sentence is, for example, the number of two keywords or sentences placed between a keyword and a sentence, the number of words placed between two keywords or a keyword and a sentence, and between two keywords or a keyword and a sentence. It may be determined based on one or more of the number of positioned words, two keywords, or the number of characters between a keyword and a sentence.

The controller 320 may first perform morpheme analysis for keyword analysis.

In step 630, the controller 320 may set keywords belonging to both the first attribute keyword candidate set and the second attribute keyword candidate set as the representative attribute keyword candidate set. That is, keywords that are used as expressions for modifying popular field object keywords and also as expressions for non-popular field object keywords may be collected in the representative attribute keyword candidate set.

According to another embodiment, in step 510, the controller 320 may include keywords that appear together with object keywords and/or field keywords in the representative attribute keyword candidate set regardless of whether they are popular or unpopular.

Returning to FIG. 5, in step 520, the controller 320 extracts two or more sub-keywords associated with each representative attribute keyword included in the representative attribute keyword candidate set from the second set of documents.

The second set document used for extracting the lower keyword in step 520 and the first set document used for extracting the representative attribute keyword candidate set in step 510 may be different document sets or may be the same document set. For example, the first set of documents may be a set including all collectable documents, and the second set of documents may be a set including only documents in which a specific field of interest to provide an information provision service is used as a main keyword. . The controller 320 may analyze collectible documents and analyze whether each document is a document used as a major keyword in a specific field of interest to provide an information providing service based on keywords that frequently appear. According to another embodiment, both the first set document and the second set document may be a set including all collectable related documents. According to another embodiment, the first set of documents may be a set including all of the collectable related documents, and the second set of documents may be a set including only documents related to a specific field of interest to be provided with an information providing service. According to another embodiment, the second set of documents may be a set including all of the collectable related documents, and the first set of documents may be a set including only documents related to a specific field of interest for which an information providing service is to be provided.

For step 520, the control unit 320, for example, in order to generate a set including only documents related to a specific field of interest to be provided with an information providing service, documents including a field keyword indicating the field of interest itself and/ Alternatively, a set including only documents related to a specific interest may be generated by collecting documents including object keywords belonging to a corresponding interest field, and extracting documents in which the weight of the field keyword/object keyword is greater than or equal to a preset value. The weight of the field keyword/object keyword can be determined based on the frequency, location, and context of the field keyword/object keyword. For example, a document in which a field keyword/object keyword frequently appears, a field keyword/object keyword is used as a title of a corresponding document, or displayed in large letters or a font for emphasis may be classified as a document related to a specific field of interest.

In step 520, the control unit 320 analyzes at least a part of the second set of documents, extracts a preset number of sub-keywords with the highest relevance to each representative property keyword, and extracts two or more sub-keywords associated with each representative property keyword. You can extract keywords.

The controller 320 may determine, for example, a degree of association between the representative attribute keyword and the lower keyword in consideration of the frequency at which the lower keyword appears in the same or similar context as the representative attribute keyword. For example, it can be said that words appearing around keyword A in a specific sentence may appear around words related to keyword A in other documents.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on a trip with a big heart, but because it was July, the weather was too humid, so I suffered."

Looking at the above two sentences, the word "because it's hot" has been replaced by the word "because it's wet" in the same context. The controller 320 may infer that “hot” and “wet” are words related to each other.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on vacation with a big heart, but because it was July, the weather was too hot and I suffered."

Likewise, the controller 320 may infer that “travel” and “vacation” are related words through the above two sentences.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on a trip with a big heart, but because it was in August, the weather was too hot and I suffered."

Similarly, the controller 320 may infer that “July” and “August” are related words through the above two sentences.

The control unit 320 indicates that "hot" and "wet" are related words, "July" and "August" are related words, and "travel" and "vacation" are related words through the previously collected documents. It can be saved. Assume that the following sentences are collected.

"I went on vacation with a big heart, but because it was July, the weather was too humid, so I suffered."

"I went on a trip with a big heart, but it was hard because the weather was too hot because it was August."

Although the two sentences are not the same context, if it is known that “hot” and “wet” are related words, “July” and “August” are related words, and “travel” and “vacation” are related words, the controller 320 ) From the above sentence, you will be able to learn that "I had a hard time" and "I had a hard time" are also related words.

Keyword pairs with a high frequency appearing in the same/similar context may be determined to have a high correlation with each other. In addition, the higher the similarity of the context in which the two keywords appear, the higher the correlation between the two keywords can be determined. The control unit 320 performs learning using the collected documents to set the degree of association between keywords, and sets the degree of relevance of keywords appearing in the corresponding sentence by using the relationship between the set keywords and the context of the sentence. The accuracy of judging the relationship between the liver can also be improved.

As similar learning methods, NNLM (Neural Net Language Model), RNNLM (Recurrent Neural Net Language Model), word2vec, skipgram, and CBOW (Continuous Bag-of-Words) methods are known. In particular, when word2vec is used, word2vec learns using documents and associates each keyword with a vector, and the similarity between the two keywords can be determined by calculating the cosine similarity of the two vectors.

In this or similar manner, the controller 320 may analyze at least a part of the second set of documents and extract sub-keywords having the highest correlation with each representative attribute keyword as many as a preset number.

In step 530, the controller 320 may extract an association weight corresponding to each representative attribute keyword and sub-keyword pair in the representative attribute keyword candidate set from the second set of documents.

7 is a detailed flowchart of step 530 according to an embodiment of the present specification.

In step 710, the controller 320 may analyze at least a part of the second set of documents to extract a degree of correlation between the lower keywords. For example, it is assumed that those collected as related sub-keywords for the representative attribute keyword A1 are 50 sub-keywords of _{B1 1} to B1 _50. In this case, the controller 320 may extract a degree of correlation between the two sub-keywords by using the frequency at which the two sub-keywords appear together in the same document for these 50 sub-keywords. B1 is also interconnected between the _first B1 and B1 ₂ and B1 is determined in accordance with the _first frequency to the _second appearing in the same document. According to another embodiment, B1 ₁ and B1 ₂ are The distance between the addition to the frequency appearing in the same document that affects the degree interrelated, B1 ₁ and B1 if the _two appeared in the same document both keywords B1 ₁ and B1 distance between _two (or a statement that both keywords appear The closer the) is, the higher the degree of correlation can be recognized. In a similar way, correlations between sub-keywords can be extracted. The distance between keywords or between a keyword and a sentence is, for example, the number of two keywords or sentences placed between a keyword and a sentence, the number of words placed between two keywords or a keyword and a sentence, and between two keywords or a keyword and a sentence. It may be determined based on one or more of the number of positioned words, two keywords, or the number of characters between a keyword and a sentence.

In step 720, the controller 320 may extract an association weight between each representative attribute keyword and the lower keyword based on the degree of correlation between the lower keywords. The control unit 320, for example, for a sub-keyword set corresponding to each representative attribute keyword, a correlation degree between a specific sub-keyword in the sub-keyword set and another sub-keyword in the sub-keyword set, and the specific sub-keyword and the The association weight between the specific lower keyword and the representative attribute keyword may be set so that the association weight between the representative attribute keywords has a positive correlation with each other.

For example, as the correlation between _{the sub-keyword B1 1} of the representative attribute keyword A1 and other sub-keywords B1 ₂ to B1 ₅₀ of A1 is higher, the association weight between _{A1 and B1 1 may be set higher.} For example, an arithmetic mean (or sum) of correlations between _{B1 1} and other sub-keywords of A1 (B1 ₂ to B1 ₅₀ ) may be an association weight between _{B1 1 and A1.} Instead of a simple arithmetic mean, a geometric mean/harmonized mean may be used. A cutting average that is averaged out of the correlations between the sub-keywords B1 ₁ and the other sub-keywords of A1 (B1 ₂ to B1 ₅₀ ) may be used, excluding the two highest (example) and two lowest (example). have. Instead of the arithmetic mean of the degree of correlation, a median may be used.

According to some embodiments, _{"B1 1} and B1 _{2, which} are used to calculate the association weight of _{B1 1 with respect to A1, are} "The frequency that appears in the same document" simply means that B1 ₁ and B1 ₂ It does not depend on the number of documents appearing together (or appearing in the same sentence, or appearing in close proximity), but B1 ₁ and B1 ₂ The number of documents appearing together (or appearing in the same sentence, or appearing in close proximity) can be obtained by dividing the number of documents in _{which B1 1} appears and/or _{the number of documents in which B1 2 appears.} In a similar way, "B1 ₁ and B1 ₂ "The frequency of appearance in the same document" means that B1 ₁ and B1 ₂ Appeared to as (to appear or are in the same sentence, or close to the emergence of) correlation of the number and the notes of the document to the number and the positive correlation between the B1 ₁ appeared number of documents and / or to have B1 ₂ appeared in the document It can be set to have a relationship. It is simply a kind of normalization to prevent commonly used words from having a high association weight to the representative attribute keyword A1.

Returning to FIG. 5, in step 540, the controller 320 may extract a sub-relationship between the object item and the sub-keyword from the first set of documents.

Among the first set of documents, a document such as an object keyword indicating an object item (for example, "Taylor Swift"), or sub-keywords frequently appearing in the same sentence or adjacent sentence may be determined to be related to the object item. The controller 320 may collect documents in which an object keyword of a corresponding object item appears, and extract a lower association degree between the lower keyword and the object keyword according to the frequency of appearing together in the documents. Particularly, when the lower keyword appears in the same sentence as the object keyword, the controller 320 may set the association between the lower keyword and the object item to be higher than when the lower keyword appears in a sentence different from the object keyword.

The controller 320 may set a higher degree of association between the lower keyword and the object item of the corresponding object keyword as the sentence in which the lower keyword appears is closer to the sentence in which the object keyword appears. The proximity of two sentences is, for example, any of the number of sentences between the two sentences, the number of words between the two sentences, the number of words between the two sentences, and the number of letters between the two sentences. It may be determined based on more than one.

The controller 320 may set a higher degree of association between the lower keyword and the object item of the corresponding object keyword as the position where the lower keyword appears is closer to the position where the object keyword appears. The proximity between the sub-keyword and the object keyword is, for example, the number of sentences between the sub-keyword and the object keyword, the number of words between the sub-keyword and the object keyword, and the number of words between the sub-keyword and the object keyword. It may be determined based on any one or more of the number and the number of characters positioned between the lower keyword and the object keyword.

In step 550, the controller 320 may extract an object-keyword association degree between the object item and each of the representative attribute keywords by using the sub-relationship in step 540 and the association weight in step 530.

For example, the object-keyword association between object item C and the representative attribute keyword A1 uses the sub-keywords of C and A1 (for example, B1 ₁ to B1 ₅₀ ) and the association weight of each sub-keyword. Can be extracted. For example, the object-keyword association degree between the object item C and the representative attribute keyword A1 may be set higher as the lower degree of association between the _{object item C and B1 1} to B1 _{50 is higher.}

In addition, if the sub-keyword with a higher association weight in the relationship with A1 has a higher sub-relationship with the object item C, the sub-keyword with the lower association weight is compared with the object C and the representative attribute keyword A1 compared to the case with a higher sub-keyword. The object-keyword association between them may be set higher. For example, in the case of Table ₁ , the sub-relationship of the one with higher association weight (B1 1) than in the case of Table 2 is higher, so that in the case of Table 1, the object between object C and the representative attribute keyword A1- Keyword relevance may be set to be high.

Associated weight with A1 Degree of sub-relation with C B1 ₁ 0.5 0.5 B1 ₂ 0.2 0.2

Associated weight with A1 Degree of sub-relation with C B1 ₁ 0.2 0.5 B1 ₂ 0.5 0.2

According to an embodiment, the object-keyword association degree between the object C and the representative attribute keyword A1 may be obtained as the sum of the product of the association weight corresponding to each sub-keyword and the sub-relation degree (or using this total). In the case of Table 1, 0.5×0.5+0.2×0.2=0.29, and in the case of Table 2, 0.2×0.5+0.5×0.2=0.20, so that the case of Table 1 is between the object C and the representative attribute keyword A1 than the case of Table 2. The object-keyword association degree of may be set to be high. The above-described method of calculating the object-keyword association is only an example, and if the sub-relationship with C in step 540 and the association weight with A1 in step 530 are positively correlated with the object-keyword association between C and A1. Other methods may be used.

Thereafter, when the communication unit 310 receives a request to provide information related to a specific representative attribute keyword, the control unit 320 may provide a result item through the communication unit 310 based on the object-keyword association degree extracted in step 550. For example, when receiving a request to provide information including any one of the representative attribute keywords, the controller 320 may provide information on the object items in the order of the highest object-keyword association in the relationship with the corresponding representative attribute keyword. have.

In another embodiment, when receiving an information provision request including two or more representative attribute keywords and weights corresponding thereto, the controller 320 associates object-keywords with representative attribute keywords included in the information provision request for each object item. Information on object items may be provided in the order of the highest sum (or average) of values multiplied by weights (or similarly weighted) to degrees.

8 is a flowchart of a process of providing information according to another embodiment of the present specification.

The example of FIG. 8 further includes two steps 523 and 526 between steps 520 and 530 in addition to the same processes as the example of FIG. 5. Here, redundant description is avoided and only steps 523 and 526 are described.

In step 523, the controller 320 determines whether each of the sub-keywords extracted in step 520 corresponds to an emotion language (emotional language). To this end, the storage unit 330 or an external server may hold a dictionary of appraisal words. The appraised word dictionary is a tool for determining whether a certain word (keyword) is an appraised word, and may hold a list of appraised words, for example. A keyword included in the list of appraised words may be referred to as an appraised word, and a keyword not included in the appraised word list may be determined to be not an appraised word. However, this judgment is based on dictionary meaning, and may not reflect the popular use of words that change with the times. Accordingly, the control unit 320 does not determine whether the representative attribute keyword itself is an appraised word, but determines whether to use the representative attribute keyword based on whether the subkeywords related to the representative attribute keyword are appraised words.

In another embodiment, the controller 320 may add another word registered as an appraised word in the appraised word dictionary and learned to have a high degree of correlation (with a pre-set value or more) with a pre-set number of words or more, to the appraised word dictionary.

In step 526, the control unit 320 may leave only a preset number of representative attribute keywords in the representative attribute keyword candidate set in the order in which the emotional language ratio (or number) of the associated lower keyword is high, and remove the remainder. Through this process, it is possible to prevent keywords that are far from emotional language from being treated like emotional language.

Returning to FIG. 9, in step 920, the controller 320 stores a basic reserved word-keyword association degree corresponding to the reserved word and the representative attribute keyword pair in the storage unit 330.

The reserved word may include expressions that can be expressed by weights of representative attribute keywords. For example, "bored" can be a reserved word, and "pretty" can be a reserved word.

For the reserved word "bored", the basic reserved word-keyword representative attribute keywords with high correlation may include representative attribute keywords that can solve a boring situation, such as "interesting", "exciting", "for getting out of time". .

For the reserved word "pretty", the default reserved word-keyword representative attribute with high correlation is a representative attribute that is similar to "pretty" or describes "pretty" such as "beautiful", "cute", "eye-catching" Can include keywords.

The process of step 920 may be performed, for example, as an input of an administrator or may be performed by receiving a basic reserved word-keyword association degree determined in an external system through a network or a storage medium. According to another embodiment, the process of step 920 may be performed in a manner similar to the process of FIGS. 5 to 8 by analyzing collectible documents such as the Internet, SNS, and news. In addition, the process of step 920 may include a process of reflecting a user's feedback, as described later.

11 is an example of a basic reserved word-keyword association diagram according to an embodiment of the present specification.

In the embodiment of FIG. 11, all of the reserved words are q (C ₁ to C _q ), and all of the representative attribute keywords are n (k ₁ to k _n ).

For example, the basic reserved word-keyword association of the reserved _{word C 5} and the representative attribute keyword k _{3 is v 3,5} .

In step 930, the communication unit 310 receives and obtains a reception reserved word from the terminal 200, and transmits the reception reserved word to the control unit 320.

The reception reserved word is a reserved word received by the terminal 200 from a search user. The terminal 200 may convert the voice input into an electrical signal (voice signal) and transmit it to the search device 300. The control unit 320 of the search device 300 may analyze the voice signal, convert it into text, and match the converted text to the reserved word. The controller 320 may also analyze the voice signal to analyze the intonation, pitch, speed, breathing state, etc. of the voice and use it as context information.

According to another embodiment, the terminal 200 may convert a voice input into text and transmit the converted text to the search device 300. The terminal 200 may analyze the intonation, height, speed, breathing state, etc. of the received voice input and transmit the analysis information to the search device 300. The search device 300 may utilize the analysis information as a kind of context information.

In step 940, the controller 320 acquires the received reserved word and the reserved word-object correlation corresponding to each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree.

12 is a detailed flowchart of step 940 according to an embodiment of the present specification.

Referring to FIG. 12, in step 1210, the controller 320 acquires a coordinated object-keyword association corresponding to each object item and a representative attribute keyword pair for a reception reserved word.

According to an embodiment, for each object item and representative attribute keyword pair, the control unit 320 is configured to correspond to the reception reserved word and the representative attribute keyword pair in an object-keyword association degree corresponding to the object item and the representative attribute keyword pair. By applying the basic reserved word-keyword association degree, an adjusted object-keyword association degree corresponding to each object item and representative attribute keyword pair for the received reserved word may be obtained.

In particular, for each object item and representative attribute keyword pair, the control unit 320 includes a basic reserved word corresponding to the received reserved word and the representative attribute keyword pair in the object-keyword association degree corresponding to the object item and the representative attribute keyword pair. The coordinated object-keyword association corresponding to each object item and representative attribute keyword pair for the received reserved word may be obtained by using a value multiplied by the degree of association.

In addition, the control unit 320 has an object-keyword association degree corresponding to the object item and the representative attribute keyword pair with respect to each object item and representative attribute keyword pair, and has a positive correlation, and the received reserved word and the representative attribute keyword pair The coordinated object-keyword association degree corresponding to each object item and representative attribute keyword pair may be set for the received reserved word so as to have a positive correlation with a corresponding basic reserved word-keyword association degree.

In the present specification, it is assumed that values representing object-keyword association, basic reserved word-keyword association, coordinated object-keyword association, and other association relationships have a tighter association as the values increase. In another embodiment, in some cases, when the association value is smaller, the association is tight, and for other association values, when the association value is larger, the association is tight, so that the positive correlation-negative correlation is appropriate. It could be used instead.

For example, when the reception reserved word is C ₂ , in order to obtain a coordinated object-keyword association corresponding to the object item i ₄ and the representative attribute keyword k _{3 pair, the control unit 320 is the object item i 4} and the representative attribute keyword k ₃ pair. The coordinated object-keyword association can be obtained by applying a basic reserved word-keyword association v _3,4 corresponding to a pair of the received reserved word C ₂ and the representative attribute keyword k ₃ to the object-relationship w _{4,3 corresponding to.} .

In particular, the method of applying the degree of association may be a method of multiplying the degree of object-relatedness and the basic reserved word-keyword association. For example, when the reception reserved word is C ₂ , in order to obtain a coordinated object-keyword association corresponding to the object item i ₄ and the representative attribute keyword k _{3 pair, the control unit 320 is the object item i 4} and the representative attribute keyword k ₃ pair. _{(W 4,3} ×v _3,2 ), which is the product of the object-relationship w _4,3 corresponding to the received reserved word C ₂ and the basic reserved word-keyword association v _3,2 corresponding to the pair of representative attribute keywords k ₃ The coordinated object-keyword association can be obtained by using. In another embodiment, instead of multiplying, the controller 320 _{uses a function f(w 4,3)} to which a different calculation/use method is applied so that the _{coordinated object-keyword association has a positive correlation with w 4,3} and v _3,2. , v _3,2 ) can also be used to obtain the coordinated object-keyword association. In addition, _{a method of using (w 4,3} ×v _3,2 ) as a coordinated object-keyword association, and (w _4,3 ×v _3,2 ) are corrected based on other factors as a coordinated object-keyword association. Any method you use can be used.

를 이용하여 획득될 수 있다. f(w_4,j,v_j,2)는 객체 항목 i₄ 및 수신 예약어 C₂, 키워드 k_j에 대응하는 조정 객체-키워드 연관도이다.In operation 1220, the controller 320 may obtain a reserved word-object association degree using a value obtained by accumulating the coordinated object-keyword association degree for a specific object item. For example, the control unit 320 may determine the received reserved word and the reserved word-object correlation corresponding to the specific object item pair so as to have a positive correlation with the accumulated value of the coordinated object-keyword correlation for a specific object item. Can be set. The reserved word-object association corresponding to the object item i ₄ and the received reserved word C _{2 pair is, for example,}

It can be obtained using f(w _4,j ,v _j,2 ) is a coordinated object-keyword association corresponding to the object item i ₄ , the reception reserved word C ₂ , and the keyword k _j.

이 될 수 있다. 다른 예에서 객체 항목 i₄ 및 수신 예약어 C₂ 쌍에 대응하는 예약어-객체 연관도는

에 대해 기타 요인에 의한 보정을 가한 값이 될 수 있다.For example, the reserved word-object association corresponding to the pair of object item i ₄ and received reserved word C _{2 is}

Can be In another example, the reserved word-object association corresponding to the pair of _{object item i 4} and received reserved word C _{2 is}

It may be a value that has been corrected for by other factors.

Returning to FIG. 9, in step 950, the controller 320 may provide an object item according to a reserved word-object association degree corresponding to the received reserved word. For example, if the reserved word _{corresponding to the received reserved word C 2} -the object association is as shown in Table 3 below, the controller 320 may provide the object items in the order of Table 4.

Object item Reserved word-object association with received reserved word i ₁ 0.23 i ₂ 0.33 i ₃ 0.99 i ₄ 0.84

order Object item Reserved word-object association with received reserved word One i ₃ 0.99 2 i ₄ 0.84 3 i ₂ 0.33 4 i ₁ 0.23

That is, the control unit 320 may provide the object items in the order of a high degree of relation between the reserved words corresponding to the reception reserved words and the object. The terminal 200 receiving the object item may provide information on the _{object item i 3 to the user through the display unit 220.} The terminal 200 may also provide information on other object items in a lower order, if necessary. The terminal 200 may provide information on the _{object item i 3} to the user through a voice through a speaker instead of the display unit 220.

13 is a flowchart of a process of providing information according to another embodiment of the present specification.

The processes of FIGS. 13 to 17 may be performed by partially using the processes of FIGS. 5 to 12 or partially changing the processes of FIGS. 5 to 12. While describing the process of FIGS. 13 to 17, descriptions of FIGS. 5 to 12 may be used if necessary.

Referring to FIG. 13, in step 1310, the controller 320 extracts a representative attribute keyword candidate set from first set of documents. The controller 320 may collect, for example, keywords frequently appearing in documents corresponding to the field of interest among the first set of documents as a representative attribute keyword candidate set. The process of step 1310 may be the same as or similar to the process of step 510 of FIG. 5, for example. The process of step 1310 may be performed through the same or similar process as that of FIG. 6. The description of the process of FIG. 6 is not repeated again.

In step 1320, the controller 320 sets a reserved word set. For example, a set of reserved words can be set by manual input by an administrator. According to a modified example, the controller 320 may provide an interface for setting a word/phrase suitable for a reserved word as a reserved word candidate, and setting one or more of the reserved word candidates as a reserved word.

18 is a detailed flowchart of step 1320 according to a modified example of the present specification.

In step 1810, the controller 320 acquires the number of appearances of one language unit or two or more consecutive language units in a document set. Here, the document set may be the same document set as the document set used in step 510 of FIG. 5 or may be a different document set.

The language unit may be, for example, any one of words/words/morphemes/syllables/letters. A unit divided by several criteria for classifying other sentences may be the language unit in the present embodiment.

Before step 1810, the controller 320 may divide documents included in each document of the document set into word units and store them in the form of an arrangement or a list. Depending on the embodiment, the control unit 320 may delete meaningless words in each word, for example, some research in Korean, denoting adjectives such as'i','me', and other words that are not necessary for analysis from the word phrase or from the arrangement/list. Can be removed. In addition, according to an embodiment, when a word consists of one word, the controller 320 may convert the corresponding word into a basic form (or a preset form).

According to a modified example, before step 1810, the controller 320 may divide documents included in each document of the document set into words and store them in the form of an arrangement or a list. The controller 320 may convert each word into a basic type (or a preset type) according to an embodiment. According to an embodiment, the controller 320 may remove meaningless words, for example, some surveys in Korean, denoting adjectives such as'i' and'me', and other words that are not necessary for analysis from the arrangement/list.

A modified example in which the control unit 320 divides the document into morpheme units, syllable units, or character units is also possible.

In the following embodiments, for convenience, it is assumed that the control unit 320 divides a document into word units, and each word becomes a language unit.

A single language unit can be a reserved word. According to a modified embodiment, two or more consecutive language units may also be reserved words. For example, "clean" (single language unit) can be a reserved word, and "atmosphere" (two consecutive language units) can be a reserved word. However, since two or more consecutive language units are difficult to appear more often than a single language unit, in order to select two or more consecutive language units as reserved words according to embodiments, weights or additional points for two or more consecutive language units when selecting a reserved word Can give. According to a modified example, when selecting a reserved word, a reference value for selecting a reserved word for two or more consecutive language units may be more generously set. For example, if a single language unit must appear at least a times to be recommended as a reserved word candidate, two consecutive language units can be set to be recommended as reserved word candidates even if only b appears, which is a much smaller number of times than a. The language unit may be set to be recommended as a reserved word candidate even if it appears only c times less than b. Hereinafter, in the present specification, two or more consecutive language units are referred to as continuous language units.

In addition, if a continuous language unit is recommended/selected as a reserved word, the language unit included in the reserved word (candidate) or included in the reserved word (candidate), and consecutive language units shorter than the reserved word (candidate) are not recommended as reserved words or recommended as reserved words. Points can be deducted when calculating the number of points to become. This is to prevent multiple similar reserved words from being selected or recommended. Hereinafter, the description of the continuous language unit is omitted to simplify the description, but the description of the single language unit may be applied equally or similarly to the continuous language unit.

The number of appearances of the language unit in step 1810 may be, for example, the number of documents in which the language unit appears. Even if the language unit appears several times in a document, only one appearance is recognized. According to another embodiment, a case in which a corresponding language unit appears multiple times in a document may be recognized as the number of appearances, and the number of appearances may be the number of appearances of the language unit.

According to another embodiment, when a corresponding language unit repeatedly appears more than once in a document, the second and subsequent appearances may be recognized as having a lower value than the first appearance. Also, if the language unit appears repeatedly in a document, the later appearance can be recognized as a lower score. As the number of appearances increases, the score increases, but the slope gradually decreases. For example, 1/r power of the number of appearances (r is a real number greater than 1) may be used as the appearance score of the corresponding language unit in the document. For example, (log value of the number of appearances) + 1 (however, if the number of appearances is 0, the number of appearances is 0), etc. may be used. In addition, the appearance score of a language unit in a document may be limited so as not to exceed a preset upper limit value. The value obtained by accumulating the appearance score of the corresponding language unit for all documents may be the appearance score according to the number of appearances of the corresponding language unit. And these appearance scores can be used in step 1830.

Hereinafter, for convenience, it is assumed that the number of appearances of a language unit is the number of documents in which the corresponding language unit appears.

In step 1820, the controller 320 acquires the number of times the appraised word is located within a preset distance from the language unit. The distance between the language unit and the appraised word is, for example, among the number of words located between the language unit and the appraised word, the number of words between the language unit and the appraised word, and the number of characters between the language unit and the appraised word. It may be determined based on one or more.

In addition, when the language unit and the appraised word belong to different sentences, the controller 320 determines that the appraised word is not located within a preset distance from the language unit, regardless of the number of words/words/characters located between the language unit and the appraised word. May be. According to another variation, when the language unit and the appraised word belong to different sentences, the controller 320 may calculate the distance by adding a certain number to the distance calculated by excluding the judgment on the sentence. This is because if the language unit and the appraised words belong to different sentences, the probability of not being related to each other is relatively high, so it is desirable to evaluate the distance further than the number of words/words/characters located between appraised words.

Whether a specific word (word phrase) is an appraised word can be checked by inquiring into a pre-registered appraised word dictionary.

The number of times the appraised word is located within a certain distance from the language unit may be, for example, the number of documents in which the language unit and the appraised word are co-located within a certain distance. Even if the language unit and the appraised word appear together within a certain distance in a document, the number of appearances is only recognized once. According to another embodiment, when there are multiple language units in which an appraised word is located within a certain distance among the corresponding language units of a document, all of the multiple language units may be recognized as the number of times the appraised word is located within a certain distance from the language unit. Hereinafter, the language unit in which the appraised word is located within a certain distance is referred to as the appraised word location language unit.

According to another embodiment, when an appraised word location language unit repeatedly appears more than once in a document, the second and subsequent appearances may be recognized as having a lower value than the first appearance. In addition, if the appearance of the corresponding appraisal position language unit in a document is repeated, the later appearance can be recognized as a lower score. As the number of appearances increases, the score increases, but the slope gradually decreases. For example, 1/r power of the number of appearances (r is a real number greater than 1) may be used as the appearance score of the appraisal position language unit in the document. For example, (the log value of the number of appearances) + 1 (however, if the number of appearances is 0, the number of appearances is 0), etc. may be used. In addition, the appearance score of the appraised word location language unit in a document may be limited so as not to exceed a preset upper limit value. A value obtained by accumulating the appearance score of the corresponding appraised word location language unit for all documents may be the appearance score according to the number of appearances of the corresponding language unit. And this appearance score can be used in step 1830.

According to another embodiment, the higher the number of appraised words located within a certain distance for one language unit, the higher the appearance score may be recognized. In addition, if there is an appraised word within a closer distance for one language unit, the appearance score may be recognized higher. In addition, the appearance score (the number of appearances) may be recognized only when there are two or more pre-set number of appraised words for one language unit instead of one.

Hereinafter, for convenience, it is assumed that the number of appearances of the appraised word location language unit is the number of documents in which the appraised word appears within a preset distance from the corresponding language unit.

In step 1830, the controller 320 sets a reserved word candidate in consideration of the number of appearances of the language unit and the number of times the appraised word is located within a predetermined distance from the language unit.

For example, the controller 320 may multiply the number of appearances of the language unit and the number of times the appraised word is located within a certain distance (or according to an operation method having a positive correlation for the other two variables) and convert it into an appraised word score. In addition, a predetermined number of preset words may be set as reserved word candidates in the order of the highest score. Alternatively, a language unit corresponding to a preset score or higher may be set as a reserved word candidate.

Language unit Number of appearances Number of appraised words score First language unit 3003 1122 3369366 2nd language unit 2001 1820 3641820 3rd language unit 3121 1300 4057300 4th language unit 200 110 22000

In the example of Table 5, reserved words may be candidates in the order of a third language unit -> a second language unit -> a first language unit -> a fourth language unit. If the controller 320 recommends two candidate reserved words, the third language unit and the second language unit will be recommended. If the control unit 320 recommends a language unit with a score of 3 million or more as a reserved word candidate, the third language unit, the second language unit, and the first language unit will be recommended as the reserved word candidate in this order. According to another embodiment, the control unit 320 extracts the first number of language units in the order of the number of appearances of the language units, and then the number of times (or the score) of the appraised words appear within a certain distance of the extracted language units. A certain number of reserved word candidates can be extracted in one order. In the example of Table 5, if three language units are selected in the order of the number of appearances, the first to third language units may be extracted. Based on the number of appearances of the appraised word, candidates for reserved words may be recommended in the order of the second language unit, the third language unit, and the first language unit.

According to another embodiment, the controller 320 extracts the first number of language units in the order of the number of appearances of the language units, and then the number of times (or the score) of the appraised words appear within a certain distance of the extracted language units. A second number of reserved word candidates (however, the second number is less than the first number) may be extracted in one order. In the example of Table 5, if three language units are selected in the order of the number of appearances, the first to third language units may be extracted. If two language units are extracted based on the number of appearances of the appraised word, candidates for reserved words may be recommended in the order of the second language unit and the third language unit.

In a manner similar to or somewhat different from the other methods described above, the control unit 320 has a positive correlation with the number of appearances of the language unit in step 1810 in a manner similar to or somewhat different from the above-described method, and the appraised word is located within a preset distance of step 1820. A score for selecting a candidate for a reserved word can be set to have a positive correlation with the number of appearances in a language unit, and the score can be used to recommend a candidate for a reserved word.

In addition, the controller 320 may process a language unit already included in the reserved word set so that it is not added as a reserved word candidate. Also, the controller 320 may process a language unit that is substantially the same as a reserved word included in the reserved word set so that it is not added as a reserved word candidate.

In step 1840, the information providing apparatus 300 provides the terminal 200 with interface information for generating a reserved word selection interface including reserved word candidate information. The interface information may be, for example, an html format document. According to another embodiment, the interface information includes only dynamic information (recommended reserved word candidates, etc.) necessary to create an interface, and the terminal 200 reflects this dynamic information in the form of a page previously stored in the terminal 200. In this way, a page including an interface can be provided to the user.

The controller 320 may generate page information for generating a page including a reserved word selection interface including reserved word candidate information, and the communication unit 310 may provide the page information to the terminal 200. The terminal 200 may render and display a page including a corresponding interface to the user. According to a modified embodiment, instead of an interface included in a visual page, an interface by sound or an interface by a technology known to date or to be known in the future may be provided. In the following description, it is assumed that an interface included in a visual page is provided for convenience.

19 is an example of an interface 1900 generated according to the interface information provided in step 1840.

Referring to FIG. 19, the interface 1900 includes a table including a check box column 1910, a reserved word candidate column 1920, and a detail view column 1930. In addition, the interface 1900 may include a reserved word add button 1940, a candidate delete button 1950, and an put button 1960. The user selects the desired reserved word candidate(s) in the check box column 1910, and then selects any one of the add reserved word button 1940, the delete candidate button 1950, and the put in library button 1960 to process the reserved word candidate. I can. When one of the buttons is selected, the terminal 200 may transmit input information obtained by converting the user's input to the information providing device 300.

The information providing device 300 may process the reserved word candidate according to input information received from the terminal 200. For example, when the user selects the check box 1910 of some reserved word candidates (hereinafter referred to as'selected candidates') and selects the reserved word add button 1940, the control unit of the information providing device 300 that has received input information related thereto ( 320) may add selected candidates to the reserved word set and delete selected candidates from the reserved word candidate set. The controller 320 controls not to recommend the language unit(s) included in the reserved word set and the language unit(s) substantially the same as the language unit(s) included in the reserved word set as the reserved word candidates when recommending the reserved word candidate in the future.

According to another example, when the user selects the check box 1910 of some reserved word candidates and selects the reserved word candidate deletion button 1950, the control unit 320 of the information providing device 300 receiving related input information excludes reserved words. Selected candidates may be added to the set and selected candidates may be deleted from the reserved word candidate set. The controller 320 controls not to recommend the language unit(s) included in the reserved word exclusion set and the language unit(s) substantially the same as the language unit(s) included in the reserved word exclusion set as reserved word candidates when recommending future reserved word candidates. do.

According to another example, when the user selects the check box 1910 of some reserved word candidates and selects the put button 1960, the control unit 320 of the information providing device 300 receiving related input information Selected candidates may be added to the candidate storage set and selected candidates may be deleted from the reserved word candidate set. The controller 320 does not recommend the language unit(s) included in the reserved word candidate storage set and the language unit(s) substantially the same as the language unit(s) included in the reserved word candidate storage set as reserved word candidates when recommending a reserved word candidate in the future. Control to avoid.

Instead of the buttons 1940, 1950, and 1960, another interface similar to a button or another interface capable of performing the role of a button may be used.

In addition, the controller 320 may provide an interface for partially deleting reserved words from the reserved word set. The controller 320 may provide an interface for removing a language unit from the reserved word exclusion set so that some language units of the reserved word exclusion set are not excluded from recommendation. The control unit 320 provides the language units of the reserved word candidate storage set in the form of a list interface similar to that of FIG. 19, and allows the user to add some language units of the reserved word candidate storage set to the reserved words through the list interface. In addition, the user may include some language units of the reserved word candidate storage set in the reserved word exclusion set and delete it from the reserved word candidate storage set through the list interface. In this case, the corresponding language unit is no longer provided through the list interface of the reserved word candidate storage set, and is not recommended as a reserved word candidate even through the interface 1900 of FIG. 19. In addition, the user may simply delete some language units of the reserved word candidate storage set from the reserved word candidate storage set through the list interface. In this case, the corresponding language unit is no longer provided through the list interface of the reserved word candidate storage set, but may be recommended as a reserved word candidate through the interface 1900 of FIG. 19.

In addition, the interface 1900 may include a previous page button 1970 and/or a next page button 1980 for moving a page in case all of the reserved word candidates are not displayed on one page. The previous page button 1970 and/or the next page button 1980 may be selectively provided according to the number of actual candidates and the current page position. In addition, instead of the previous page button 1970 and/or the next page button 1980, an interface extending through scrolling may be provided. In some interfaces, only tables including items 1910, 1920, and 1930 are scrolled, and buttons 1940, 1950, 1960, 1970, and 1980 may be excluded from scrolling.

The user may select the detailed view 1930 to inquire about the background or related information in which the reserved word candidate is recommended. When the detail view 1930 is selected, the interface provided by the control unit 320 includes information on the corresponding reserved word candidate and an interface for adding the corresponding reserved word candidate as a reserved word, to the reserved word candidate storage set, or to the reserved word exclusion set. can do.

An interface for managing other reserved word candidates may be provided to the user through the terminal 200 by the controller 320.

Returning to FIG. 18, in step 1850, the controller 320 may add the selected reserved word candidate to the reserved word set according to an input for selecting the reserved word.

Returning to FIG. 13, in step 1330, the controller 320 stores an object-keyword association diagram corresponding to an object item and a representative attribute keyword pair.

10 is an example of a stored object-keyword association diagram according to an embodiment of the present specification.

In the embodiment of FIG. 10, there are m number of object items (i ₁ to i _m ) and n number of representative attribute keywords (k ₁ to k _n ).

For example, the object-keyword association of the _{object item i 5} and the representative attribute keyword k _{3 is w 5,3} .

The process of step 1330 may be performed according to, for example, some of the embodiments of FIGS. 5 to 8, a process similar thereto, or a process corresponding thereto. According to another embodiment, the process of step 1330 may be performed as an input of an administrator or may be performed by receiving an object-keyword association degree determined in an external system through a network or a storage medium.

Since the embodiments of FIGS. 5 to 8 have been described above, repeated descriptions will be omitted. However, the process performed in step 510 of FIGS. 5 and 8 is substantially the same as the process of step 1310 of FIG. 13. Therefore, in performing the process of step 1330, even if the embodiments of FIGS. 5 to 8 are used, the process of step 510 is not performed again and the result of step 1310 may be recycled.

In step 1340, the control unit 320 uses the association weight corresponding to the representative attribute keyword and the lower keyword pair, and the basic reserved word corresponding to the reserved word and the lower keyword pair-the basic reserved word corresponding to the representative attribute keyword pair. -The keyword association is stored in the storage unit 330. The process of step 1340 may be performed according to an input of an administrator, or may be performed according to one or more embodiments of FIGS. 15 to 17.

Prior to the process of step 1340 or during the process of step 1340, a sub-keyword is determined, an association weight corresponding to a representative attribute keyword and a sub-keyword pair is determined, and a basic reserved word-sub-keyword association corresponding to the reserved word and the sub-keyword pair Should be determined.

The lower keyword used in the process of step 1340 may be determined while step 520 is performed during the process of step 1330. In this case, the lower keyword of step 520 may be used in step 1340. If the lower keyword is not determined in step 1330, the lower keyword may be determined through the same or similar process as in step 520 of FIG. 5 and the previous process.

The association weight used in the process of step 1340 may be determined while step 530 is performed during the process of step 1330. In this case, the association weight of step 530 may be used in step 1340. When the association weight is not determined in step 1330, the association weight may be determined through the same or similar process as in step 530 of FIG. 5 and the previous process.

The degree of association between the basic reserved word and the sub-keyword may be calculated, for example, in consideration of the frequency in which the reserved word and the sub-keyword appear in the same or similar context.

Hereinafter, in the description of FIGS. 15 to 17, an example of acquiring a basic reserved word-keyword relationship v _{3 and 2 between the reserved word C 2} and the representative attribute keyword k 3 will be described. For example, it is assumed that the sub-keywords of the _{representative attribute keyword k 3 are B3 1} to B3 _50. For a reserved word, a representative attribute keyword, and a basic reserved word-keyword association, refer to the example described above with reference to FIG. 11. The sub-keyword association degree is expressed as x _{j, h} - reserved word reserved word base corresponding to C _j and sub-keyword pair Bg _h. The association weight corresponding to the pair of the lower keyword Bg _h and the representative attribute keyword k _g is expressed as _{y g,h.} The sub-keyword association degree is expressed as x _{j, g, h} - reserved word C _j, k _g represent attribute keywords and keyword sub Bg _h combination adjustment reserved word corresponding to.

15 is a detailed flowchart of step 1340 according to the first embodiment of the present specification.

Referring to FIG. 15, in step 1510, the controller 320 obtains an adjusted reserved word-sub-keyword correlation by applying an association weight to the basic reserved word-sub-keyword association degree.

In step 1510, the control unit 320 applies an association weight corresponding to the lower keyword and the representative attribute keyword pair to the basic reserved word-sub-keyword association degree corresponding to the reserved word and the lower keyword pair for each reserved word and sub-keyword pair. An adjusted reserved word-sub-keyword association degree corresponding to each reserved word and sub-keyword pair for the representative attribute keyword may be obtained.

For example, controller 320 is C ₂ to reserved words that correspond to the sub-keyword B3 ₄ and shorthand keyword k ₃ pairs when C ₂ days Adjustment between B3 _{4 In} order to obtain the reserved word-sub-keyword correlation, the basic reserved word-sub-keyword correlation x _{2,4 corresponding to the reserved word C 2} and the sub-keyword B3 ₄ pairs, and the sub-keyword B3 ₄ and the representative attribute keyword k ₃ pairs. It is possible to obtain an adjusted reserved word-sub-keyword correlation x _2,3,4 by applying a corresponding association weight y _3,4.

In particular, the method of applying the association weight may be a method of multiplying the basic reserved word-sub-keyword association degree x _2,4 by the association weight y _3,4 corresponding to the lower keyword B3 ₄ and the representative attribute keyword k _{3 pair.} For example, controller 320 is C ₂ to reserved words that correspond to the sub-keyword B3 ₄ and shorthand keyword k ₃ pairs when C ₂ days Adjustment between B3 _{4 In} order to obtain the reserved word-sub-keyword correlation x _2,3,4 , the basic reserved word-sub-keyword correlation x _{2,4 corresponding to the reserved word C 2} and the sub-keyword B3 ₄ pair x 2,4 and the sub-keyword B3 ₄ and representative An adjusted reserved word-sub-keyword association degree x _2,3,4 may be obtained _{by using (x 2,4} Хy _3,4 ), which is a value obtained by multiplying the association weight y _3,4 corresponding to the attribute keyword k _{3 pair.} In another embodiment, instead of multiplying, the control unit 320 applies another calculation/use method in which the _{adjusted reserved word-sub-keyword correlation x 2,3,4} has a _{positive correlation with x 2,4} and y _3,4. It is also possible to obtain an adjusted reserved word-sub-keyword association x _2,3,4 using the function f(x _2,4 ,y _{3,4 ).} In addition, (x _2,4 Хy _3,4 ) is used as an adjusted reserved word-sub-keyword association x _2,3,4, and adjusted reserved words by applying corrections based on other factors to (x _2,4 Хy _3,4) -As sub-keyword relevance x _2,3,4, both methods can be utilized.

가 될 수 있다. 즉, 하위 키워드 B3₁ 내지 B3₅₀에 대해 예약어 C₂와의 기본 예약어-하위키워드 연관도 x_2,f를 구한 뒤 각각에 대해 해당 하위 키워드에 대한 연관 가중치 y_3,f를 반영하여 x_2,3,f를 구하고 x_2,3,f를 누적하여 예약어 C₂ 및 대표 속성 키워드 k₃ 간의 기본 예약어-키워드 연관도를 구할 수 있다. 다른 실시 예에 따르면, 예약어 C₂ 및 대표 속성 키워드 k₃ 간의 기본 예약어-키워드 연관도는

에 기타 요인에 의한 보정을 가한 값이 될 수 있다. 다른 실시 예에 따르면, 예약어 C₂ 및 대표 속성 키워드 k₃ 간의 기본 예약어-키워드 연관도는

과 양의 상관 관계를 가지는 값이 될 수 있다. 여기서는 하나의 대표속성 키워드에 연결돼 있는 하위 키워드가 50개인 것으로 가정하였으나, 대표속성 키워드에 연결돼 있는 하위 키워드의 개수가 달라지면 수식에서 f의 누적 범위가 50이 아닌 다른 값이 될 수 있을 것이다.In step 1520, the controller 320 may set a basic reserved word-keyword association between the reserved word C ₂ and the representative attribute keyword k ₃ using the accumulated value of the adjusted reserved word-sub-keyword correlation x _2,3,f. That is, the basic reserved word-keyword association between the reserved _{word C 2} and the representative attribute keyword k _{3 is}

Can be. In other words, for the sub-keywords B3 ₁ to B3 ₅₀ , the basic reserved word-sub-keyword association x _{2,f with the reserved word C 2 is obtained, and then x 2,3} by reflecting the _{association weight y 3,f} for the sub-keyword for each. _, to obtain a _2,3 x _{f, f} by accumulating the reserved words C ₂ and the representative keyword attribute default reserved word between k ₃ - can be obtained also associated keywords. According to another embodiment, the basic reserved word-keyword association between the reserved _{word C 2} and the representative attribute keyword k _{3 is}

It may be a value that has been corrected by other factors. According to another embodiment, the basic reserved word-keyword association between the reserved _{word C 2} and the representative attribute keyword k _{3 is}

It can be a value that has a positive correlation with. Here, it is assumed that there are 50 sub-keywords connected to one representative attribute keyword, but if the number of sub-keywords connected to the representative attribute keyword is different, the cumulative range of f in the formula may be a value other than 50.

16 is a detailed flowchart of step 1340 according to another embodiment of the present specification. The processes of steps 1510 and 1520 of FIG. 16 are the same as those described above with reference to FIG. 15 and thus are not repeated.

Referring to FIG. 16, in step 1530, the control unit 320 selects a representative keyword whose relationship between the reserved word and the corresponding representative keyword pair is less than or equal to the reference basic reserved word-keyword correlation among representative keywords corresponding to a specific reserved word. You can delete the default reserved word-keyword association. Representative keywords corresponding to a specific reserved word refer to keywords for which a basic reserved word-keyword association degree is set in the relationship with the specific reserved word. The reference basic reserved word-keyword association may be set in advance. According to another embodiment, the reference basic reserved word-keyword association is set using the average value of the basic reserved word-keyword association corresponding to a specific reserved word, or when the basic reserved word-keyword association corresponding to a specific reserved word is arranged in order of size. It can be set by using the basic reserved word-keyword association of the ranking. A specific value having a positive correlation with basic reserved word-keyword associations corresponding to other specific reserved words may be the reference basic reserved word-keyword association. The reference basic reserved word-keyword association degree may be set differently according to the reserved words, or may be set identically for all reserved words. When the basic reserved word-keyword association is deleted, there is no association between the reserved word and the representative keyword. The control unit may set the degree of association to 0, or the basic reserved word-keyword association degree of the reserved word and the corresponding representative keyword pair in a list (which can be replaced by an array or other data structure) indicating a basic reserved word-keyword association degree. It is also possible to delete the basic reserved word-keyword association by deleting information related to it. Other basic reserved word-keyword associations may be deleted (or a method of adding information indicating that the association has been deleted) may be used.

After going through the process of step 1530, the basic reserved word-keyword association, which has a relatively fine degree of association, is deleted (i.e., it is set as no association), so that an excessively complex operation is performed or an association that has no practical meaning to the user/administrator is created. It can be suppressed.

17 is a detailed flowchart of step 1340 according to another embodiment of the present specification.

The processes of steps 1510 and 1520 of FIG. 17 are the same as those described above with reference to FIG. 15, so the description will not be repeated. The process of step 1530 of FIG. 17 is the same as the process described above with reference to FIG. 16, so the description will not be repeated.

In step 1540, the controller 320 may normalize the basic reserved word-keyword association that is not deleted and remains. For example, in order to include the average value of the basic reserved word-keyword associations stored connected to a specific representative attribute keyword in a specific range, the basic reserved word-keyword associations stored connected to a specific representative attribute keyword are multiplied by a certain coefficient to increase or Can be reduced. For another example, in order to ensure that the total of the basic reserved word-keyword associations stored connected to a specific representative attribute keyword is included in a specific range, the basic reserved word-keyword associations stored connected to a specific representative attribute keyword are multiplied by a certain coefficient to increase. It can be reduced or reduced. In other words, in order to prevent the case that only the specific representative attribute keyword is recommended/used or the specific representative attribute keyword is rarely used even if a certain reserved word is selected because the high-value basic reserved word-keyword association is concentrated on a certain representative attribute keyword. Appropriate adjustments can be made.

According to another embodiment, in step 1540, the controller 320 may perform normalization by adding a predetermined coefficient or combining operations such as log and square instead of multiplying the basic reserved word-keyword association by a predetermined coefficient. According to another embodiment, normalization may be performed by partially reducing only the basic reserved word-keyword associations equal to or higher than a specific reference value or partially increasing the basic reserved word-keyword associations equal to or lower than a specific reference value.

According to another embodiment, in step 1540, the controller 320 may normalize the average value (or the sum) of the basic reserved word-keyword associations stored in connection with the specific reserved word to be included in a specific range.

In the process of FIG. 17, step 1530 is omitted, and a modified example in which the basic reserved word-keyword association is also normalized is also possible.

In step 1350, the communication unit 310 receives and obtains a reception reserved word from the terminal 200, and transmits the reception reserved word to the control unit 320.

The reception reserved word is a reserved word received by the terminal 200 from a search user. The terminal 200 may convert the voice input into an electrical signal (voice signal) and transmit it to the search device 300. The control unit 320 of the search device 300 may analyze the voice signal, convert it into text, and match the converted text to the reserved word. The controller 320 may also analyze the voice signal to analyze the intonation, pitch, speed, breathing state, etc. of the voice and use it as context information.

According to another embodiment, the terminal 200 may convert a voice input into text and transmit the converted text to the search device 300. The terminal 200 may analyze the intonation, height, speed, breathing state, etc. of the received voice input and transmit the analysis information to the search device 300. The search device 300 may utilize the analysis information as a kind of context information.

In step 1360, the controller 320 acquires the received reserved word and the reserved word-object correlation corresponding to each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree. The process of step 1360 may be performed according to the method of step 940 of FIG. 9 or the method of FIG. 12. Description of the same content will be omitted.

In step 1370, the controller 320 may provide an object item according to a reserved word-object association degree corresponding to the received reserved word. The process of step 1370 may be performed in the same manner as the process of step 950. The same description is omitted.

14 is a flowchart of a process of providing information according to another embodiment of the present specification.

Steps 1310, 1320, 1330, 1340, 1350, 1360, and 1370 of FIG. 14 are the same as steps 1310, 1320, 1330, 1340, 1350, 1360, and 1370 of FIG. 13, so the same description will not be repeated.

Step 1333 added in FIG. 14 may be performed at any time after the representative attribute keyword and the lower keyword are determined. For example, step 1333 may be performed simultaneously/in parallel with step 1330, or the process of step 1333 may be performed during step 1330.

In step 1333, the controller 320 stores an association weight between the representative attribute keyword and the lower keyword. If the process of setting the association weight is not performed in step 1330, the association weight between the representative attribute keyword and the lower keyword may be set through a process similar to that of step 530 of FIG. 5. According to another embodiment, the controller 320 may store the association weight between the representative attribute keyword and the sub-keyword in a manner that loads the association weight set in step 1330.

In step 1337, the controller 320 obtains a basic reserved word-low keyword association between the lower keyword and the reserved word.

The controller 320 may determine, for example, a degree of association between the reserved word and the lower keyword in consideration of the frequency in which the lower keyword appears in the same or similar context as the reserved word. For example, it can be said that words appearing around keyword A in a specific sentence may appear around words related to keyword A in other documents.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on a trip with a big heart, but because it was July, the weather was too humid, so I suffered."

Looking at the above two sentences, the word "because it's hot" has been replaced by the word "because it's wet" in the same context. The controller 320 may infer that “hot” and “wet” are words related to each other.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on vacation with a big heart, but because it was July, the weather was too hot and I suffered."

Likewise, the controller 320 may infer that “travel” and “vacation” are related words through the above two sentences.

"I went on a trip with a big heart, but because it was July, the weather was too hot and I suffered."

"I went on a trip with a big heart, but because it was in August, the weather was too hot and I suffered."

Similarly, the controller 320 may infer that “July” and “August” are related words through the above two sentences.

The control unit 320 indicates that "hot" and "wet" are related words, "July" and "August" are related words, and "travel" and "vacation" are related words through the previously collected documents. It can be saved. Assume that the following sentences are collected.

"I went on vacation with a big heart, but because it was July, the weather was too humid, so I suffered."

"I went on a trip with a big heart, but it was hard because the weather was too hot because it was August."

Although the two sentences are not the same context, if it is known that “hot” and “wet” are related words, “July” and “August” are related words, and “travel” and “vacation” are related words, the controller 320 ) From the above sentence, you will be able to learn that "I had a hard time" and "I had a hard time" are also related words.

Keyword pairs with a high frequency appearing in the same/similar context may be determined to have a high correlation with each other. In addition, the higher the similarity of the context in which the two keywords appear, the higher the correlation between the two keywords can be determined. The control unit 320 performs learning using the collected documents to set the degree of association between keywords, and sets the degree of relevance of keywords appearing in the corresponding sentence by using the relationship between the set keywords and the context of the sentence. The accuracy of judging the relationship between the liver can also be improved.

As similar learning methods, NNLM (Neural Net Language Model), RNNLM (Recurrent Neural Net Language Model), word2vec, skipgram, and CBOW (Continuous Bag-of-Words) methods are known. In particular, when word2vec is used, word2vec learns using documents and associates each keyword with a vector, and the similarity between the two keywords can be determined by calculating the cosine similarity of the two vectors.

20 is a diagram illustrating a term hierarchy according to an embodiment of the present specification.

When the process of FIG. 13 or 14 is completed, a hierarchical relationship between _{reserved words (C 1} to C _q ), representative attribute keywords (k ₁ to k _n ), and lower keywords (BX ₁ to BX _{50) is established.} .

A basic reserved word-keyword association degree is set between the reserved word and the representative attribute keyword, and an association weight value is set between the representative attribute keyword and the sub-keyword. Using this hierarchical relationship, the controller 320 may recommend an appropriate object according to the reserved word or perform an operation such as selecting a new reserved word candidate. In addition, the hierarchical relationship of FIG. 20 may be modified or supplemented through learning while repeating the process of FIGS. 13 and 14 by reflecting new data.

At this time, it will be appreciated that each block of the flowchart diagrams and combinations of the flowchart diagrams may be executed by computer program instructions. Since these computer program instructions can be mounted on the processor of a general purpose computer, special purpose computer or other programmable data processing equipment, the instructions executed by the processor of the computer or other programmable data processing equipment are described in the flowchart block(s). It creates a means to perform functions. These computer program instructions can also be stored in computer-usable or computer-readable memory that can be directed to a computer or other programmable data processing equipment to implement a function in a particular way, so that the computer-usable or computer-readable memory It is also possible for the instructions stored in the flow chart to produce an article of manufacture containing instruction means for performing the functions described in the flowchart block(s). Since computer program instructions can also be mounted on a computer or other programmable data processing equipment, a series of operating steps are performed on a computer or other programmable data processing equipment to create a computer-executable process to create a computer or other programmable data processing equipment. It is also possible for instructions to perform processing equipment to provide steps for executing the functions described in the flowchart block(s).

In addition, each block may represent a module, segment, or part of code that contains one or more executable instructions for executing the specified logical function(s). In addition, it should be noted that in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For example, two blocks shown in succession may in fact be executed substantially simultaneously, or the blocks may sometimes be executed in the reverse order depending on the corresponding function.

In this case, the term'~ unit' used in this embodiment refers to software or hardware components such as FPGA or ASIC, and'~ unit' performs certain roles. However,'~ part' is not limited to software or hardware. The'~ unit' may be configured to be in an addressable storage medium, or may be configured to reproduce one or more processors. Thus, as an example,'~ unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , Subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, database, data structures, tables, arrays, and variables. Components and functions provided in the'~ units' may be combined into a smaller number of elements and'~ units', or may be further separated into additional elements and'~ units'. In addition, components and'~ units' may be implemented to play one or more CPUs in a device or a security multimedia card.

Those of ordinary skill in the art to which the present specification pertains will understand that the present specification may be embodied in other specific forms without changing the technical spirit or essential features thereof. Therefore, it should be understood that the embodiments described above are illustrative and non-limiting in all respects. The scope of the present specification is indicated by the scope of the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and the concept of equivalents thereof are included in the scope of the present specification. It must be interpreted.

On the other hand, the present specification and drawings disclose preferred embodiments of the present specification, and although specific terms are used, these are merely used in a general meaning to easily describe the technical content of the present specification and to aid understanding of the present invention. It is not intended to limit the scope of the specification. In addition to the embodiments disclosed herein, it is apparent to those of ordinary skill in the technical field to which the present specification belongs that other modified examples based on the technical idea of the present specification can be implemented.

Claims (12)

As an information providing method of an information providing device,
Storing an object-keyword association degree corresponding to an object item and a representative attribute keyword pair;
Storing a basic reserved word-keyword association corresponding to the reserved word and the representative attribute keyword pair by using the association weight corresponding to the representative attribute keyword and the sub-keyword pair, and the basic reserved word-sub-keyword association degree corresponding to the reserved word and the sub-keyword pair. ;
Obtaining a reception reserved word selected by a user;
Obtaining a reserved word-object association degree corresponding to the received reserved word and each object item pair by using the object-keyword association degree and the basic reserved word-keyword association degree; And
Providing an object item according to the reserved word-object association degree corresponding to the received reserved word and each object item pair;
The reserved word is an information providing method including an expression expressed by a degree of association with the representative attribute keywords or weights of the representative attribute keywords.
The method of claim 1,
Using the object-keyword association degree and the basic reserved word-keyword association degree, the step of obtaining the reserved word-object association degree corresponding to the received reserved word and each object item pair,
For each object item and representative attribute keyword pair, the received reserved word and the basic reserved word corresponding to the representative attribute keyword pair are applied to the object-keyword association degree corresponding to the object item and the representative attribute keyword pair. Acquiring a coordinated object-keyword association corresponding to each object item and representative attribute keyword pair for the received reserved word; And
And obtaining the reserved word-object correlation by using a value obtained by accumulating the adjusted object-keyword correlation for a specific object item.
The method of claim 2,
For each object item and representative attribute keyword pair, by applying a basic reserved word-keyword association degree corresponding to the reception reserved word and the representative attribute keyword pair to the object-keyword association degree corresponding to the object item and the representative attribute keyword pair The step of obtaining a coordinated object-keyword association corresponding to each object item and representative attribute keyword pair for the received reserved word,
For each object item and representative attribute keyword pair, a basic reserved word corresponding to the received reserved word and the representative attribute keyword pair having a positive correlation with the object-keyword association degree corresponding to the object item and the representative attribute keyword pair- And setting an adjustment object-keyword association degree corresponding to each object item and representative attribute keyword pair with respect to the received reserved word so as to have a positive correlation with a keyword association degree.
The method of claim 2,
The step of obtaining the reserved word-object association degree by using the accumulated value of the coordinated object-keyword association degree for the specific object item,
Information comprising the step of setting the reserved word-object association degree corresponding to the received reserved word and the specific object item pair so as to have a positive correlation with the accumulated value of the coordinated object-keyword association degree for the specific object item Delivery method.
The method of claim 3,
For each object item and representative attribute keyword pair, by applying a basic reserved word-keyword association degree corresponding to the reception reserved word and the representative attribute keyword pair to the object-keyword association degree corresponding to the object item and the representative attribute keyword pair The step of obtaining a coordinated object-keyword association corresponding to each object item and representative attribute keyword pair for the received reserved word,
For each object item and representative attribute keyword pair, an object-keyword association degree corresponding to the object item and the representative attribute keyword pair multiplied by the received reserved word and a basic reserved word-keyword association degree corresponding to the representative attribute keyword pair And acquiring a coordinated object-keyword association corresponding to each object item and representative attribute keyword pair for the received reserved word.
The method of claim 4,
The step of obtaining the reserved word-object association degree by using the accumulated value of the coordinated object-keyword association degree for the specific object item,
And obtaining the reserved word-object correlation corresponding to the received reserved word and each object item pair by using a value obtained by accumulating the coordinated object-keyword association degree for the specific object item.
A storage unit for storing an object-keyword association degree corresponding to an object item and a representative attribute keyword pair, and a basic reserved word-keyword association degree corresponding to the reserved word and the representative attribute keyword pair;
A communication unit that obtains a reception reserved word selected by a user; And
And a control unit for obtaining a reserved word-object association degree corresponding to the received reserved word and each object item pair using the object-keyword association degree and the basic reserved word-keyword association degree,
The control unit calculates a basic reserved word-keyword association degree corresponding to a reserved word and a representative attribute keyword pair by using the association weight corresponding to the representative attribute keyword and the lower keyword pair, and the basic reserved word-sub-keyword association degree corresponding to the reserved word and the lower keyword pair. Stored in the storage unit,
The information providing apparatus including an expression expressed by a degree of association with the representative attribute keywords or a weight of the representative attribute keywords in the reserved word.
The method of claim 7,
The control unit,
For each object item and representative attribute keyword pair, by applying a basic reserved word-keyword association degree corresponding to the reception reserved word and the representative attribute keyword pair to the object-keyword association degree corresponding to the object item and the representative attribute keyword pair Acquiring a coordinated object-keyword association degree corresponding to each object item and representative attribute keyword pair for the received reserved word,
An information providing device for acquiring the reserved word-object correlation by using a value obtained by accumulating the adjusted object-keyword correlation for a specific object item.
The method of claim 8,
The control unit,
For each object item and representative attribute keyword pair, a basic reserved word corresponding to the received reserved word and the representative attribute keyword pair having a positive correlation with the object-keyword association degree corresponding to the object item and the representative attribute keyword pair- An information providing apparatus for setting an adjustment object-keyword association degree corresponding to each object item and a representative attribute keyword pair with respect to the reception reserved word so as to have a positive correlation with a keyword association degree.
The method of claim 9,
The control unit,
An information providing apparatus for setting the reserved word-object association degree corresponding to the received reserved word and the specific object item pair so as to have a positive correlation with the accumulated value of the coordinated object-keyword association degree for the specific object item.
The method of claim 10,
The control unit,
For each object item and representative attribute keyword pair, an object-keyword association degree corresponding to the object item and the representative attribute keyword pair multiplied by the received reserved word and a basic reserved word-keyword association degree corresponding to the representative attribute keyword pair An information providing device for obtaining a coordinated object-keyword association corresponding to each object item and a representative attribute keyword pair for the received reserved word by using.
The method of claim 11,
The control unit,
An information providing device for acquiring the reserved word-object association degree corresponding to the received reserved word and each object item pair by using a value obtained by accumulating the coordinated object-keyword association degree for the specific object item.

Images