KR101220080B1 - Method and system for providing reprsentation words of real-time popular keyword - Google Patents

Abstract

A method and system for providing representative phrases for real-time popular keywords are disclosed. The representative phrase providing method and system may generate the first candidate representative phrase based on the document and generate the second candidate representative phrase based on the broadcast data. In addition, a suitable representative phrase among the first and second candidate representative phrases may be provided in combination with a popular keyword.

Description

METHOD AND SYSTEM FOR PROVIDING REPRSENTATION WORDS OF REAL-TIME POPULAR KEYWORD}

The present invention relates to a method and system for providing a representative phrase for a real-time popular keyword, and more particularly, a method for extracting a representative phrase, which is the cause of the sudden increase in popularity of a keyword exposed on a site, based on documents and broadcast data. It is about the system.

In general, most web pages that provide keyword search expose popular keywords, which are search terms that are rapidly increasing in real time at the present time. At this time, the ranking of the selected popular keyword and the popular keyword is exposed on the web page.

In particular, the popular keyword is simply a celebrity name or exposed as an international or domestic corporate name on a web page. Accordingly, when the user wants to know what contents the popular keyword includes, the user can select a specific popular keyword by using a mouse click or the like. Then, articles, documents, etc. related to the selected popular keyword may be exposed on the web page.

In this way, not only if you want to know the details of the popular keywords, but also if you want to know the general contents such as why they became popular keywords, the user should select the popular keywords and read all the articles about the selected popular keywords. There is a hassle to do.

In addition, most of the users listen to the TV or radio and perform a search because they are curious about the information of the cast, or when the specific broadcast program starts or the content of a specific broadcast program that is broadcast today. Accordingly, most of the real-time popular keywords are celebrity names and broadcast program titles.

Therefore, there is a need for a method of providing a broadcast program corresponding to a real-time popular keyword to inform the user of the cause of the real-time popular keyword.

In addition, there is a need for a method capable of providing a reason for becoming a popular keyword in real time even if a user does not select a popular keyword and articles related to the popular keyword step by step.

The present invention provides a method and system that can solve a question about a reason given as a popular keyword by providing a first candidate representative phrase related to a cause of exposure of a popular keyword in real time using a document.

The present invention provides a method and system that can solve a question about a reason presented as a popular keyword by providing a second candidate representative phrase related to the cause of exposure of the popular keyword in real time using broadcast data.

The present invention provides a method and system for more accurately providing the cause of the popular keyword exposure by determining the final representative phrase for the popular keyword in consideration of the popularity and broadcast suitability for the first candidate representative phrase and the second candidate representative phrase. To provide.

In accordance with an embodiment of the present invention, a method of providing representative phrases may include generating a first candidate representative phrase for a popular keyword based on a document, and generating a second candidate representative phrase for a popular keyword based on broadcast data. And determining the final representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase.

In the determining of the final representative phrase, the final representative phrase may be determined to have a higher suitability for a popular keyword among the first candidate representative phrase and the second candidate representative phrase.

The method may further include calculating a popularity of the document used to generate the first candidate representative phrase.

The determining of the representative phrase may include determining the first candidate representative phrase as the final representative phrase by using the popularity of the document and a predetermined reference popularity.

The method may further include calculating a time score of the second candidate representative phrase by using the broadcast time of the broadcast data used to generate the second candidate representative phrase and the exposure time of the popular keyword.

In the determining of the representative phrase, the second candidate representative phrase which is determined to have a higher suitability for the popular keyword by using the time score of the second candidate representative phrase and a predetermined reference time score is finally represented. You can decide by phrases.

The generating of the first candidate representative phrase includes: determining a reference word of documents including the popular keyword, determining a representative reference word among the reference words, and successively before or after the representative reference word. And combining the representative reference word and the representative reference word to expand the representative reference word, and generating the first candidate representative phrase using the extended representative reference word.

The expanding of the representative criterion word may include calculating a conditional probability including successive words before or after the representative criterion word in documents including the popular keyword, and based on the conditional probability. And expanding the representative reference word.

The determining of the representative reference word may include analyzing a morpheme of documents including the popular keyword, counting a frequency of the reference word, and determining the representative reference word based on the counted frequency. It may include.

The method may further include collecting documents including the popular keyword, and performing a cluster on the collected documents. Then, in the generating of the first candidate representative phrase, the first candidate representative phrase may be generated by extending a representative reference word based on documents belonging to the cluster.

The generating of the first candidate representative phrase may further include adding a weight to the collected documents, and determining an exposure priority of the cluster using the weight.

The generating of the second candidate representative phrase may include: determining a broadcast program including the popular keyword in the broadcast data; and generating the second candidate representative phrase using a broadcast time of a broadcast program. It may include.

The generating of the second candidate representative phrase may include: when the popular keyword is a combination of several words, stemming the popular keyword into a plurality of words, and broadcasting the popular keyword. The method may include selecting any one broadcast program based on a plurality of words separated from the program, and generating the second candidate representative phrase using the broadcast time of the selected broadcast program.

The generating of the second candidate representative phrase may include adding a matching score, a time score, and a broadcast station weight to a broadcast program including the popular keyword, wherein the matching score, time score, broadcast station weight, and broadcasting program are included. Calculating a final matching score for the broadcast program using at least one of the viewer ratings, selecting one of the broadcast programs based on the final matching score among broadcast programs including the popular keyword, and selecting the selected broadcast program And generating the second candidate representative phrase by using the broadcast time.

The method may further include providing the final representative phrase by combining the popular keyword.

Representative phrase providing system according to an embodiment of the present invention, a first candidate representative phrase generation unit for generating a first candidate representative phrase for a popular keyword based on a document, a second candidate for a popular keyword based on broadcast data And a second candidate representative phrase generation unit generating a representative phrase, and a representative phrase determination unit determining a final representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase. .

In addition, the representative phrase determination unit may determine that the higher suitability of the popular keyword among the first candidate representative phrase and the second candidate representative phrase as the final representative phrase.

The apparatus may further include a popularity calculator configured to calculate a popularity of the document used to generate the first candidate representative phrase. Then, the representative phrase determination unit may determine, as the final representative phrase, the first candidate representative phrase determined to have a higher suitability for the popular keyword by using the popularity of the document and a predetermined reference popularity.

The apparatus may further include a time score calculator configured to calculate a time score of the second candidate representative phrase by using the broadcast time of the broadcast data used to generate the second candidate representative phrase and the exposure time of the popular keyword. Then, the representative phrase determination unit determines the second candidate representative phrase as the final representative phrase that is determined to have higher suitability for the popular keyword by using the time score of the second candidate representative phrase and a preset reference time score. Can be.

The first candidate representative phrase generation unit may further determine a reference word by analyzing the morphemes of documents including the popular keyword, and determine the representative reference word according to the determined frequency of the reference word, and the representative. And a reference word extension unit configured to expand the representative reference word by combining a continuous word before or after the reference word and the representative reference word, and generate the second candidate representative phrase using the extended representative reference word. have.

The apparatus may further include a document collecting unit collecting documents including the popular keyword, and a cluster performing unit performing a cluster on the collected documents.

The second candidate representative phrase generation unit may be configured to generate the second candidate representative phrase using the broadcast program determiner that determines a broadcast program including the popular keyword in the broadcast data, and the broadcast time of the determined broadcast program. It may include a generation unit.

The second candidate representative phrase generator may, when the popular keyword is in the form of a combination of words, separates the popular keyword into a plurality of words and separates the broadcast keyword including the popular keyword from the broadcast program including the popular keyword. The apparatus may include a broadcast program determiner that selects any one broadcast program based on a plurality of words, and a generator that generates the second candidate representative phrase using the broadcast time of the selected broadcast program.

The second candidate representative phrase generation unit may add a matching score, a time score, and a broadcast station weight to a broadcast program including the popular keyword, and at least one of the matching score, the time score, the broadcast station weight, and the viewer rating of the broadcast program. A matching score calculation unit that calculates a final matching score for the broadcast program using a broadcast program determination unit that selects any one broadcast program based on the final matching score among broadcast programs including the popular keyword, and selected It may include a generation unit for generating the second candidate representative phrase by using the broadcast time of the broadcast program.

In addition, the representative phrase may include a representative phrase providing unit for providing a combination of the popular keyword.

According to an embodiment of the present invention, the question about the reason for being presented as the popular keyword may be solved by providing a first candidate representative phrase related to the cause of the popular keyword being exposed in real time using the document.

According to an embodiment of the present invention, questions about the reason for being presented as a popular keyword may be solved by providing a second candidate representative phrase related to the cause of exposure of the popular keyword in real time using broadcast data.

According to an embodiment of the present invention, by determining the final representative phrase for the popular keyword in consideration of the popularity and broadcasting suitability for the first candidate representative phrase and the second candidate representative phrase provides a more accurate cause of the exposure of the popular keyword Can be.

1 is a view provided to explain a process of generating a representative phrase according to an embodiment of the present invention.
2 is a block diagram showing the configuration of a representative phrase providing system according to an embodiment of the present invention.
3 is a flowchart provided to explain the operation of the representative phrase providing system of FIG. 2 according to an embodiment of the present invention.
4 is a block diagram illustrating a detailed configuration of a first candidate representative phrase generation unit according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating a screen displaying a web page by combining a first candidate representative phrase and a popular keyword.
6 is a flowchart provided to explain a method of generating a first candidate representative phrase according to an embodiment of the present invention.
7 is a block diagram illustrating a detailed configuration of a second candidate representative phrase generation unit according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating a screen in which a second candidate representative phrase and a popular keyword are combined and displayed on a web page.
9 is a flowchart provided to explain a process of providing a second candidate representative phrase using broadcast time according to an embodiment of the present invention.
FIG. 10 is a flowchart provided to explain a process of generating a second candidate representative phrase for a popular keyword having a combination of several words.
11 is a flowchart provided to explain a process of generating a second candidate representative phrase using a matching score according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a view provided to explain a process of generating a representative phrase according to an embodiment of the present invention.

Referring to FIG. 1, the representative phrase providing system 30 may collect documents 10 related to the real time popular keyword 40 and broadcast data 20 related to the real time popular keyword 40. Here, the representative phrase providing system 30 may include a text document including a popular keyword, an article related to the popular keyword, and documents 10 such as a still image of the popular keyword, such as a portal site and a news site. From blogs, cafes, communities such as microblogs.

In this case, the representative phrase providing system 30 may generate the first candidate representative phrase for the popular keyword by analyzing the morphemes of the collected documents. Here, the first candidate representative phrase is a phrase indicating a reason that the popular keyword to be exposed on the web page becomes a real-time popular keyword, and means a phrase generated based on a document.

In addition, when collecting broadcast data, the representative phrase providing system 30 may periodically collect broadcast data from a broadcasting station at predetermined time intervals. Here, the representative phrase providing system 30 is broadcast data such as a broadcast schedule, an electronic program guide from a broadcasting station site, a news site, and the like related to airwave TV broadcasting, cable TV broadcasting, radio broadcasting, internet broadcasting, satellite TV broadcasting, and the like. Can be collected.

In this case, the representative phrase providing system 30 may determine a broadcast program including a popular keyword from the collected broadcast data. Then, the representative phrase providing system 30 may generate a second candidate representative phrase for the popular keyword based on the determined broadcast program. Here, the second candidate representative phrase is a phrase indicating a reason that the popular keyword to be exposed on the web page becomes a real-time popular keyword and means a phrase generated based on broadcast data.

For example, the representative phrase providing system 30 may generate the second candidate representative phrase using the determined broadcast time of the broadcast program. Here, the broadcast time may include a broadcast start time and broadcast end time of the broadcast program.

In addition, the representative phrase providing system 30 may determine the representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase. In this case, the representative phrase providing system 30 may determine that the higher suitability of the popular keyword among the first candidate representative phrase and the second candidate representative phrase is the final representative phrase. Here, the first candidate representative phrase and the second candidate representative phrase mean a candidate group having a high possibility of becoming the representative phrase which is the cause of the sudden rise in popularity keywords in real time.

In one example, the representative phrase providing system 30 may determine a goodness of fit of the first candidate representative phrase based on the popularity of the document used to generate the first candidate representative phrase. In addition, the representative phrase providing system 30 may determine the first candidate representative phrase as the final representative phrase based on the goodness of fit of the first candidate representative phrase.

In this case, when it is determined that the first candidate representative phrase is not suitable, the representative phrase providing system 30 may determine the suitability of the second candidate representative phrase based on a time score of the second candidate representative phrase. Subsequently, the representative phrase providing system 30 may determine the second candidate representative phrase as the final representative phrase based on the goodness of fit of the second candidate representative phrase.

Then, the representative phrase providing system 30 may combine the final representative phrase and the popular keyword to expose the web page.

2 is a block diagram showing the configuration of a representative phrase providing system according to an embodiment of the present invention.

2, the representative phrase providing system 200 includes a first candidate representative phrase generator 210, a second candidate representative phrase generator 220, a popularity calculator 230, a representative phrase determiner 240, The time score calculator 250 and the representative phrase provider 260 may be included.

First, the first candidate representative phrase generation unit 210 may generate a first candidate representative phrase for a popular keyword based on the document 110. Here, the first candidate representative phrase is a phrase indicating the basis for which the popular keyword is presented and may be extracted based on the document 110.

For example, the first candidate representative phrase generation unit 210 may generate a first candidate representative phrase by analyzing morphemes of documents related to a real-time popular keyword. Here, a detailed process of generating the first candidate representative phrase will be described later with reference to FIGS. 4 to 6.

Subsequently, the second candidate representative phrase generator 220 may generate a second candidate representative phrase for a popular keyword based on the broadcast data 120. Here, the second candidate representative phrase is a phrase indicating the basis for which the popular keyword is presented and may be extracted based on the broadcast data 120.

For example, the second candidate representative phrase generator 220 may determine a broadcast program including a popular keyword in the broadcast data 120, and generate the second candidate representative phrase based on the determined broadcast time of the broadcast program. . Here, a detailed process of generating the second candidate representative phrase will be described later with reference to FIGS. 7 to 11.

The popularity calculator 230 may extract a document used to generate the first candidate representative phrase from among documents collected in association with the popular keyword. The popularity calculator 230 may calculate the popularity of the extracted document.

For example, the popularity calculator 230 may calculate the popularity of the document using a click-to-search ratio (CTR) that combines the number of clicks and impressions of the extracted document. At this time, the popularity calculation unit 230 may calculate the number of clicks using the click log data of the extracted document. The popularity calculator 230 may calculate the number of exposures of the document using the search log data of the extracted document. Here, the click log data and the search log data may be limited to the last 30 minutes.

Then, the representative phrase determination unit 240 determines that the higher suitability for the popular keyword among the first candidate representative phrase and the second candidate representative phrase is higher based on the goodness of fit of the first candidate representative phrase and the second candidate representative phrase. It can be decided by a representative representative phrase.

For example, when the document popularity and the predetermined reference popularity are used for determining the goodness of fit, and there is only one extracted document, the popularity calculator 230 may calculate one document popularity as the final popularity. In this case, when there are a plurality of extracted documents, the popularity calculation unit 230 may calculate the final popularity by combining the plurality of document popularity.

Then, the representative phrase determination unit 240 may determine the suitability of the first candidate representative phrase using the final popularity and the reference popularity.

For example, when the reference popularity is preset to α, the representative phrase determination unit 240 may compare the popularity of the document and the reference popularity α. If the final popularity is greater than or equal to the reference popularity α, the representative phrase determination unit 240 may determine that the first candidate representative phrase has a higher suitability for the popular keyword than the second candidate representative phrase. Then, the representative phrase determination unit 240 may determine the first candidate representative phrase as the final representative phrase. That is, when the final popularity is greater than or equal to the reference popularity α, the representative phrase determination unit 240 may determine that the document used to generate the first candidate representative phrase has obtained popularity. Here, the reference popularity α may be changed.

As another example, when the final popularity is less than the reference popularity α, the representative phrase determination unit 240 may determine that the first candidate representative phrase is not suitable as the final representative phrase for the popular keyword.

Then, the time score calculator 250 may calculate the time score of the second candidate representative phrase. In this case, the time score calculator 250 may calculate the time score of the second candidate representative phrase based on 1) the broadcast time of the broadcast program used to generate the second candidate representative phrase, and 2) the exposure time of the popular keyword. Can be. Here, the broadcast time may include a broadcast start time and a broadcast end time.

For example, when the exposure time is greater than or equal to the broadcast start time and less than or equal to the broadcast end time, the time score calculator 250 may calculate the time score of the second candidate representative phrase as the predetermined reference score β. That is, when the popular keyword is exposed during the broadcast time of the broadcast program, the time score calculator 250 may calculate the time score of the second candidate representative phrase as the reference score β.

As another example, when the exposure time is before the broadcast start time, or when the exposure time is after the broadcast end time, the time score calculator 250 subtracts the error score δ for each error unit γ from the reference score β to form the second candidate representative phrase. The time score of can be calculated. Here, the error unit γ and the error score δ are preset and may be changed.

In this case, when the exposure time is before the broadcast start time, the time score calculator 250 may calculate a difference value between the exposure time and the broadcast start time. The time score calculator 250 may calculate the time score of the second candidate representative phrase by subtracting the error score from the reference score for each error unit based on the difference value. For example, when the difference value is 20 minutes, the error unit γ is set to 10 minutes, and the error score δ is 5, the time score calculator 250 may calculate the error value 2 and the error score 5 calculated based on the difference value. The product can be subtracted from the reference score β (β− (2 × 5)). In addition, the time score calculator 250 may calculate β-10, which is a result of the subtraction, as a time score of the second candidate representative phrase.

Then, the representative phrase determination unit 240 determines whether the second candidate representative phrase has a higher suitability for the popular keyword than the first candidate representative phrase by using the time score of the second candidate representative phrase and the preset reference time score τ. You can decide.

For example, when the time score of the second candidate representative phrase is not less than or equal to the reference time score, the representative phrase determination unit 240 may determine that the second candidate representative phrase has a higher suitability for the popular keyword. Accordingly, the representative phrase determination unit 240 may determine the second candidate representative phrase as the final representative phrase. In this case, when the time score of the second candidate representative phrase is less than or equal to the reference time score, the representative phrase determination unit 240 may determine that the second candidate representative phrase is not suitable as the final representative phrase for the popular keyword.

The representative phrase provider 260 may provide a combination of the representative phrase and the popular keyword finally determined for the popular keyword. Then, the user can confirm the reason why the popularity of the popular keyword is rapidly increased in real time through the representative phrase finally determined.

For example, when the first candidate representative phrase is "secret date with Hong Kong actress BBB," and the first candidate representative phrase is determined as the final representative phrase for the popular keyword, the representative phrase providing unit 260 as shown in FIG. 5, The representative phrase "Hong Kong actress BBB and secret date" (521) and the popular keyword "AAA" (510) can be combined to expose the web page. In this case, the representative phrase provider 260 may combine the photo 522 of the popular keyword “AAA” with the popular keyword 510 and the representative phrase 521 to expose the web page.

As another example, when the second candidate representative phrase is "starting broadcasting from 22:00" and the second candidate representative phrase is determined as the final representative phrase for the popular keyword, the representative phrase providing unit 260 as shown in FIG. 8, The representative phrase provider 260 may combine the popular keyword "XXX" 810 and the representative phrase "start broadcasting from 22:00" 821 and expose the web page. In this case, the representative phrase provider 260 may combine the pictorial photo 822 of the broadcast program of the popular keyword “XXX” with the popular keyword 810 and the representative phrase 821 and expose the web page.

In FIG. 2, the representative phrase providing system 200 includes a first candidate representative phrase generation unit 210, a second candidate representative phrase generation unit 220, a popularity calculation unit 230, a representative phrase determination unit 240, Although it has been described as including the time score calculation unit 250, and the representative phrase providing unit 260, this is an embodiment, the configuration of the representative phrase providing system used in the present specification is the user or operator's intention or the present invention It may vary depending on the practice of the field to which it belongs. Therefore, the configuration of the present representative phrase providing system is not limited or limited by FIG. 2.

In one example, the representative phrase providing system according to an embodiment of the present invention may be composed of a first candidate representative phrase generation unit, a second candidate representative phrase generation unit, and a representative phrase determination unit. In this case, the representative phrase providing system may further include a representative phrase providing unit.

As another example, the representative phrase providing system according to an embodiment of the present invention, may be composed of a first candidate representative phrase generation unit, a second candidate representative phrase generation unit, a popularity calculation unit, and a representative phrase determination unit. In this case, the representative phrase providing system may further include a representative phrase providing unit.

As another example, the representative phrase providing system according to an embodiment of the present invention, the representative phrase providing system, the first candidate representative phrase generation unit, the second candidate representative phrase generation unit, time score calculation unit, and representative phrase determination unit It may be configured. In this case, the representative phrase providing system may further include a representative phrase providing unit.

3 is a flowchart provided to explain the operation of the representative phrase providing system of FIG. 2 according to an embodiment of the present invention.

First, in operation 310, the first candidate representative phrase generator 210 may generate a first candidate representative phrase based on a document. In this case, the first candidate representative phrase generation unit 210 may generate the first candidate representative phrase by analyzing the morphemes of documents related to the real-time popular keyword.

In operation 320, the second candidate representative phrase generator 220 may generate the second candidate representative phrase based on the broadcast data. For example, the second candidate representative phrase generator 220 may determine a broadcast program including a popular keyword among a plurality of broadcast programs included in the broadcast data. In addition, the second candidate representative phrase generator 220 may generate the second candidate representative phrase based on the determined broadcast time of the broadcast program.

In operation 330, the popularity calculator 230 may calculate the popularity of one or more documents used to generate the first candidate representative phrase.

In this case, the popularity calculator 230 may extract a document used to generate the first candidate representative phrase from the documents collected in association with the popular keyword. The popularity calculator 230 may calculate the popularity of the document using a click-to-search ratio (CTR) that combines the number of clicks and impressions of the extracted document. At this time, the popularity calculation unit 230 may calculate the number of clicks using the click log data of the extracted document. The popularity calculator 230 may calculate the number of exposures of the document using the search log data of the extracted document. Here, the click log data and the search log data may be data generated within the last 30 minutes, but the present invention is not limited to the above-mentioned time.

As such, when there is only one document used to generate the first candidate representative phrase, the popularity calculator 230 may calculate one document popularity. In addition, when there are a plurality of documents used to generate the first candidate representative phrase, the popularity calculation unit 230 may calculate document popularity corresponding to the plurality of documents, respectively.

Accordingly, in operation 335, the popularity calculator 230 may calculate a final popularity by combining a plurality of document popularitys. In this case, when the calculated document popularity is one, the popularity calculator 230 may calculate one document popularity as the final popularity.

Subsequently, the representative phrase determination unit 240 determines that a higher degree of suitability for the popular keyword among the first candidate representative phrase and the second candidate representative phrase is higher based on the goodness of fit of the first candidate representative phrase and the second candidate representative phrase. It can be decided by a representative representative phrase.

In operation 340, the representative phrase determination unit 240 may determine whether the first candidate representative phrase has a higher suitability for the popular keyword by using the final popularity and the predetermined reference popularity.

In this case, when the final popularity is greater than or equal to the reference popularity, in step 350, the representative phrase determination unit 240 may determine that the first candidate representative phrase has a higher suitability for the popular keyword than the second candidate representative phrase. In addition, the representative phrase determination unit 240 may determine the first candidate representative phrase as the final representative phrase. That is, when the final popularity is greater than or equal to the reference popularity α, the representative phrase determination unit 240 may determine that the document used to generate the first candidate representative phrase has obtained popularity.

In addition, when the final popularity is less than the reference popularity, the representative phrase determination unit 240 may determine that the first candidate representative phrase is not higher suitability for the popular keyword. That is, the representative phrase determination unit 240 may determine that the first candidate representative phrase is not suitable as the final representative phrase.

Then, in step 360, the time score calculator 250 may calculate the time score of the second candidate representative phrase. In this case, the time score calculator 250 may calculate the time score of the second candidate representative phrase based on the broadcast time of the broadcast program used to generate the second candidate representative phrase, and the exposure time of the popular keyword. Here, the broadcast time may include a broadcast start time and a broadcast end time.

For example, when the exposure time is greater than or equal to the broadcast start time and less than or equal to the broadcast end time, the time score calculator 250 may calculate the time score of the second candidate representative phrase as the predetermined reference score β.

As another example, when the exposure time is before the broadcast start time, or when the exposure time is after the broadcast end time, the time score calculator 250 subtracts the error score δ for each error unit γ from the reference score β to form the second candidate representative phrase. The time score of can be calculated.

In this case, when the exposure time is before the broadcast start time, the time score calculator 250 may calculate a difference value between the exposure time and the broadcast start time. The time score calculator 250 may calculate the time score of the second candidate representative phrase by subtracting the error score from the reference score for each error unit based on the difference value. For example, when the difference value is 20 minutes, the error unit γ is set to 10 minutes, and the error score δ is 5, the time score calculator 250 may calculate the error value 2 and the error score 5 calculated based on the difference value. The product can be subtracted from the reference score β (β− (2 × 5)). In addition, the time score calculator 250 may calculate β-10, which is a result of the subtraction, as a time score of the second candidate representative phrase.

Similarly, when the exposure time is after the broadcast end time, the time score calculator 250 may calculate a difference between the exposure time and the broadcast end time. In addition, the time score calculator 250 may calculate the time score of the second candidate representative phrase by subtracting the error score from the reference score for each error unit based on the difference value.

In operation 370, the representative phrase determination unit 240 may determine whether the second candidate representative phrase has a higher suitability for the popular keyword than the first candidate representative phrase by using the time score and the preset reference time score τ. have.

In this case, when the time score of the second candidate representative phrase is less than or equal to the reference time score, the representative phrase determination unit 240 may determine that the second candidate representative phrase is not higher in suitability for the popular keyword. That is, the representative phrase determination unit 240 may determine that the second candidate representative phrase is not suitable as the final representative phrase.

If the time score of the second candidate representative phrase is not less than or equal to the reference time score, in step 380, the representative phrase determination unit 240 has a better suitability for a popular keyword than the first candidate representative phrase in the second candidate representative phrase 240. It can be determined to be high. That is, when the final popularity is greater than or equal to the reference popularity and the time score of the second candidate representative phrase is not less than or equal to the reference time score, it may be determined that the second candidate representative phrase has a higher suitability for the popular keyword than the first candidate representative phrase. have. Accordingly, the representative phrase determination unit 240 may determine the second candidate representative phrase as the final representative phrase for the popular keyword.

Then, in step 390, the representative phrase provider 260 may provide a combination of the representative phrase and the popular keyword.

For example, when the first candidate representative phrase is determined as the final representative phrase for the popular keyword, as shown in FIG. 5, the representative phrase provider 260 combines the first candidate representative phrase and the popular keyword and exposes the web page. Can be.

As another example, when the second candidate representative phrase is determined as the final representative phrase for the popular keyword, as shown in FIG. 8, the representative phrase provider 260 combines the second candidate representative phrase and the popular keyword to be displayed on a web page. Can be.

Meanwhile, in step 370 of FIG. 3, when the time score of the second candidate representative phrase is less than or equal to the reference time score, the representative phrase determination unit 240 may finalize all of the first candidate representative phrase and the second candidate representative phrase for the popular keyword. It may not be decided by a representative phrase. As such, if both the first candidate representative phrase and the second candidate representative phrase are not determined as the final representative phrase for the popular keyword, the representative phrase determination unit 240 selects any one of the first candidate representative phrase and the second candidate representative phrase. Selected randomly to determine the final representative phrase for popular keywords. Then, the representative phrase provider 260 may combine and expose a randomly selected representative phrase and popular keywords.

In addition, when both the first candidate representative phrase and the second candidate representative phrase are not determined as final representative phrases, the representative phrase determination unit 240 ends both the first candidate representative phrases and the second candidate representative phrases with respect to the popular keyword. It may be decided by a representative representative phrase. Then, the representative phrase providing unit may combine and expose the first candidate representative phrase, the second candidate representative phrase, and the popular keyword.

In addition, when neither the first candidate representative phrase nor the second candidate representative phrase is determined as the final representative phrase, the representative provider 260 may expose only the popular keyword without the representative phrase.

In FIG. 3, the configuration of determining the suitability of the first and second candidate representative phrases using the popularity calculation and the time score calculation, and determining and providing the final representative phrase based on the determined fitness, has been described. For example, the configuration of the representative phrase providing method used in the present specification may vary according to a user, an operator's intention, or a custom in the field to which the present invention belongs. Therefore, the configuration of the present representative phrase providing method is not limited or limited by FIG. 3.

For example, the method of providing a representative phrase according to an embodiment of the present invention may include generating a first candidate representative phrase for a popular keyword based on a document, and displaying a second candidate representative phrase for a popular keyword based on broadcast data. And generating a final representative phrase for the popular keyword using the first candidate representative phrase and the second candidate representative phrase. In this case, the representative phrase providing method may further include providing the final representative phrase by combining the popular keyword.

As another example, the representative phrase providing method may include generating a first candidate representative phrase for a popular keyword based on a document, generating a second candidate representative phrase for a popular keyword based on broadcast data, and first candidate representative Calculating the popularity of the document used to generate the phrase, and determining the final representative phrase for the popular keyword using the first candidate representative phrase and the second candidate representative phrase. In this case, the representative phrase providing method may further include providing the final representative phrase by combining the popular keyword.

As another example, the representative phrase providing method according to an embodiment of the present invention, generating a first candidate representative phrase for the popular keyword based on the document, the second candidate representative phrase for the popular keyword based on the broadcast data Calculating a time score of the second candidate representative phrase using the broadcast time of the broadcast data used to generate the second candidate representative phrase and the exposure time of the popular keyword, and the first candidate representative phrase and The method may include determining a final representative phrase for the popular keyword by using the second candidate representative phrase. In this case, the representative phrase providing method may further include providing a final representative phrase by combining the popular keyword.

In another example, generating a first candidate representative phrase for a popular keyword based on a document, generating a second candidate representative phrase for a popular keyword based on broadcast data, and generating a first candidate representative phrase Calculating a popularity of the used document, calculating a time score of the second candidate representative phrase using the broadcast time of the broadcast data used to generate the second candidate representative phrase and the exposure time of the popular keyword, and And determining the final representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase. In this case, the representative phrase providing method may further include providing a final representative phrase by combining the popular keyword.

Hereinafter, a configuration of generating the first candidate representative phrase and a configuration of combining the first candidate representative phrase and a popular keyword when the first candidate representative phrase is determined as the representative phrase will be described with reference to FIGS. 4 to 6. It will be described in detail.

4 is a block diagram illustrating a detailed configuration of a first candidate representative phrase generation unit according to an embodiment of the present invention.

According to FIG. 4, the representative phrase providing system of FIG. 2 may further include a document collecting unit and a cluster performing unit.

First, the document collection unit 410 may collect documents related to popular keywords to be exposed as a real-time rising search word on a web page through various portal sites, news sites, blogs, microblogs, and the like.

For example, the document collection unit 410 may collect the documents for the popular keyword AAA as shown in Table 1 below.

document NNN_news # 0 Secret date with AAA, Hong Kong, top actress, 'BBB'? NNN _news # 1 AAA, secretly dating Hong Kong actress BBB? MMM _news # 3 AAA-BBB, sweetheart? Concert always popular explosion MMM _news # 1 Secret date with AAA, Hong Kong actress BBB? LLL # 0 AAA and BBB Concert Videos Popular

The cluster performer 420 may perform a cluster on the collected documents using N-gram morpheme analysis. Here, N-gram is used to add up to N words in a row, 1-gram means one word, 2-gram means a form in which two words are combined in succession. As such, as the cluster is performed through N-gram morpheme analysis, the quality of the cluster may be improved.

In this case, in the case of 1-gram morphological analysis, an incorrect cluster may be generated, but when stemming using 5 grams, the probability that the documents related to the AAA keyword are related to each other increases. For example, the cluster execution unit 420 may generate cluster input data by using 5-gram morpheme analysis on NNN_news # 0 of Table 1. In this case, the generated cluster input data may be as shown in Table 2 below.

document Cluster input data NNN_news # 0 AAA Hong Kong AAA Hong Kong Top AAA Hong Kong Best Actor Hong Kong Best Actor BBB AAA Hong Kong Best Actor BBB and AAA Hong Kong Best Actor BBB Hong Kong Best Actor Hong Kong Best Actor Hong Kong Best Actor BBB Hong Kong Best Actor BBB Best Actress BBB and Middle East Hong Kong Best Actress BBB and Middle Secret Best Actor Best Actress BBB and Best Actress BBB Best Middle Actress BBB and Middle Best Actress BBB and Middle Secret Actress Actress BBB Actress BBB and Actress BBB and Middle Actress BBB and Chinese actressBBB and Chinese secret actressBBB and Chinese secret date BBB BBB and BBB and Chinese BBB and Chinese BBB and Chinese secret BBB and Chinese secret date Chinese and Chinese secret Secret Secret Date Secret Secret Date

In Table 2, although the cluster input data generated for the document NNN_news # 0 is described, the cluster execution unit 420 may generate cluster input data using N-gram morpheme analysis for all the collected documents.

Then, the cluster execution unit 420 may perform a cluster on all the collected documents using the generated cluster input data and the categorical clustering technique. In this case, the cluster execution unit 420 may perform a cluster using a hierarchical clustering technique based on a Jaccard coefficient. For example, the cluster execution unit 420 may generate two clusters as shown in Tables 3 and 4 by performing a cluster on the documents collected in Table 1 using the hierarchical clustering technique.

# Cluster A MMM _news # 1 Secret date with AAA, Hong Kong actress BBB? NNN_news # 0 Secret date with AAA, Hong Kong, top actress, 'BBB'? NNN _news # 1 AAA, secretly dating Hong Kong actress BBB?

# Cluster B LLL # 0 AAA and BBB Concert Videos Popular MMM _news # 3 AAA-BBB, sweetheart? Concert always popular explosion

According to Table 3, cluster A is a cluster containing documents related to the date between popular keywords AAA and BBB. In addition, according to Table 4, cluster B is a cluster including documents related to a concert between popular keywords AAA and BBB.

The first candidate representative phrase generation unit 430 may generate a first candidate representative phrase for a popular keyword by combining one sentence representing each cluster in the generated plurality of clusters. Here, the first candidate representative phrase generator 430 may include a reference word determiner 431 and a reference word expander 432.

The reference word determiner 431 may determine one or more reference words by analyzing the morphemes of all documents included in one cluster. For example, when analyzing the stems of the documents included in the cluster A, the reference word determiner 431 analyzes the stems of the documents MMM _news # 1, NNN_news # 0, and NNN _news # 1 "Hong Kong", "date "," BBB "," actress "," secret "," sneak "can be determined as the reference word.

In this case, the reference word determiner 431 may count the frequency in which the determined reference word is included in the documents MMM_news # 1, NNN_news # 0, and NNN_news # 1. For example, the frequency of the reference word and the counted reference word determined for the cluster A may be as shown in Table 5 below.

Reference word Frequency Hong Kong 232 Date 224 BBB 215 actress 152 Secret 142 stealthily 70

The reference word determiner 431 may determine two reference words as representative reference words in order of the highest frequency of counted reference words. For example, according to Table 5, the reference word determiner 431 may determine the top two "Hong Kong" and "date" having a high frequency as the representative reference word.

The reference word expansion unit 432 may calculate a conditional probability that the determined representative reference word and consecutive words are included in documents belonging to the cluster. In this case, the reference word expansion unit 432 may calculate a conditional probability including the representative reference word and a continuous word before or after using the bi-gram conditional probability.

In one example, the reference word expansion unit 232 may calculate a conditional probability of including the representative reference word "Hong Kong" and the consecutive words "actress" and "highest" in documents belonging to the cluster A. In this case, the reference word expansion unit 432 may calculate the conditional probability that the word "actress" consecutive after "Hong Kong" is included in documents belonging to the cluster A by 2/3 using Equation 1 below. More specifically, among documents MMM_news # 1, NNN_news # 0, and NNN_news # 1 belonging to cluster A, the document including "Hong Kong actress" is MMM_news # 1 and NNN_news # 1. Accordingly, the reference word expansion unit 232 may calculate the conditional probability of "Hong Kong actress" to 2/3.

In the same way, the reference word extension 432 may calculate a 1/3 conditional probability that documents belonging to cluster A include the word "highest" after "Hong Kong". More specifically, since the document including "Hong Kong Highest" among the documents belonging to Cluster A is NNN_news # 0, the reference word expansion unit 432 uses the Equation 1 below to conditional probability of "Hong Kong Highest". Can be calculated as 1/3.

The reference word expansion unit 432 may extend the representative reference word to a word corresponding to a probability having a large probability value if the probability of the calculated conditional probability is greater than or equal to a predetermined reference probability.

For example, when the reference probability is set to 1/2, the reference word expansion unit 432 may compare the calculated conditional probability and select a conditional probability with a probability value. Then, the reference word expansion unit 432 may select 2/3 by comparing the conditional probability 2/3 with 1/3. The reference word expansion unit 432 may determine whether the selected conditional probability 2/3 is greater than or equal to the reference probability 1/2. In this case, since the conditional probability 2/3 is greater than or equal to the reference probability 1/2, the reference word expansion unit 432 may expand the representative reference word to the word “Hong Kong actress” corresponding to the conditional probability 2/3. That is, the representative criterion word can be extended from "Hong Kong" to "Hong Kong actress".

In this case, the reference word expansion unit 432 may continuously expand the representative reference word until the conditional probability of the word to be extended becomes less than the reference probability.

In one example, the reference word expansion unit 432 may calculate a conditional probability that the subsequent words will be included after the extended word "actress" in the documents belonging to the cluster A. In this case, the reference word expansion unit 432 may calculate the conditional probability of P (BBB and actress) as 1 in the same manner as in Equation 1 above. More specifically, since all documents belonging to cluster A contain the word "BBB" consecutively after the actress, the reference word extension 432 determines the conditional probability of P (BBB and | actress) 3/3 = 1. Can be calculated as The reference word expansion unit 432 calculates the representative reference word from "Hong Kong actress" to "Hong Kong actress BBB" because the conditional probability 1 of the calculated P (BBB and actress) is equal to or larger than the preset reference probability 1/2. Can be extended.

In this case, if the conditional probability of the expanded word is less than the reference probability, the reference word expansion unit 432 may terminate the expansion of the representative reference word. In one example, when using the Baigram conditional probability, the reference word expansion unit 432 has a conditional probability of including the extended word " BBB " and subsequent words in documents belonging to the cluster A. In other words, since the third probability is less than the reference probability 1/2, the expansion of the representative reference word can be terminated.

More specifically, document MMM _news # 1 contains "secret" as the word "BBB" contiguous, NNN_news # 0 contains "Middle West" as the word "BBB" contiguous, and NNN_news # 1 includes "sneak" as the word "continued with BBB". Accordingly, the reference word expansion unit 432 may calculate conditional probabilities including words consecutive with " BBB " after the documents belonging to the cluster A as 1/3, 1/3, and 1/3. At this time, since all of the calculated conditional probabilities are the same, the reference extension unit 432 may select one of the 1/3, and compare the selected conditional probability 1/3 with the reference probability 1/2. Since the selected conditional probability 1/2 is less than the reference probability 1/2, the reference expansion unit 432 may terminate the expansion of the representative reference word. Accordingly, the reference word expansion unit 432 may extend the representative reference word "Hong Kong" to "Hong Kong actress BBB" by using the Baigram conditional probability.

Subsequently, when the expansion of the first high frequency representative reference word "Hong Kong" ends, the reference word expansion unit 432 may expand the second high frequency representative reference word "date". In this case, the reference word expansion unit 432 may expand the "date" using the Baigram conditional probability like the "Hong Kong".

For example, the reference word extension 432 may calculate conditional probabilities of consecutive words “secret date” and “secret date” before 2 " date " have. In addition, the reference word expansion unit 432 may select 2/3 having a greater probability value among the calculated conditional probabilities. Subsequently, since the selected conditional probability 2/3 is equal to or greater than the reference probability 1/2, the reference word expansion unit 432 may extend the representative reference word with the word “secret date” corresponding to 2/3.

In this case, like the "Hong Kong", the reference word expansion unit 432 may continue to expand the representative reference word until the Bygram conditional probability of the expanded word is less than the reference probability. Then, the reference word expansion unit 432 may extend the representative reference word "date" to finally "secret date".

The reference word expansion unit 432 may generate the first candidate representative phrase by combining the extended representative reference words. For example, the reference word expansion unit 432 may combine “Hong Kong actress BBB” with “secret date” to generate “Hong Kong actress BBB and secret date” as the first candidate representative phrase. In this case, when the first candidate representative phrase is determined as the representative phrase, the representative phrase provider may combine the first candidate representative phrase “Hong Kong actress BBB and secret date” and the popular keyword “AAA” to expose the web page.

In the same manner, the first candidate representative phrase generator 430 may determine one or more reference words by analyzing the morphemes of all documents included in the cluster B. FIG. In one example, the reference word determiner 431 analyzes the morphemes of the documents LLL # 0 and MMM_news # 3 to determine the reference words "BBB", "concert", "image", "popular", "explosion", and "lover". Can be determined. At this time, the reference word determiner 431 counts the frequency in which the determined reference word is included in the documents LLL # 0 and MMM_news # 3, and represents two or more reference words in order of the highest frequency of the counted reference words. Can be determined by the reference word. Then, the reference word expansion unit 432 may calculate a conditional probability that the words consecutive to the determined representative reference word are included in the documents belonging to the cluster B.

The reference word expansion unit 432 may extend the representative reference word to a word corresponding to a probability having a large probability value if the probability of the calculated conditional probability is greater than or equal to a predetermined reference probability. In this case, the reference word expansion unit 432 may continuously expand the representative reference word until the conditional probability of the word to be extended becomes less than the reference probability. Finally, the reference word extension 432 may combine the extended representative reference words to generate a first candidate representative phrase for the documents included in the cluster B. FIG.

In this case, when the first candidate representative phrase is determined as the representative phrase, the representative phrase provider may combine the generated first candidate representative phrases for the cluster A and the cluster B with the popular keyword "AAA", respectively, to expose the web page. For example, the representative phrase provider may assign the exposure priority to the cluster A and the cluster B, and combine the first candidate representative phrase and the popular keyword of the cluster having the high exposure priority to expose the web page.

More specifically, in the case where the exposure priority of the cluster A is 80 and the exposure priority of the cluster B is 60, when the search for the popular keyword "AAA" is requested, the representative phrase provider may provide the representative phrase of the cluster A and the popular keyword "AAA." "To expose web pages. In this case, the representative phrase providing unit may determine the exposure priority using the weights of the documents included in the plurality of clusters. For example, the representative phrase provider may assign the exposure priority to the highest sum of the weights of the documents included in the cluster. Then, the representative phrase provider may select the cluster having the highest exposure priority among the plurality of clusters, and combine the first candidate representative phrase of the selected cluster with the popular keyword to expose the web page.

As another example, the representative phrase provider may combine the first candidate representative phrase of one of the plurality of clusters with popular keywords by using the size of the cluster to expose the web page. Here, the size of the cluster is the number of documents belonging to the cluster. That is, the representative phrase providing unit may combine the first candidate representative phrase of the cluster including a large number of documents among the cluster A and the cluster B with the popular keyword to expose the web page.

In this case, the representative phrase providing unit may calculate the size of the document by adding a weight according to the creation time of the document belonging to the cluster. Then, the representative phrase providing unit adds a higher weight to the recently written document, so that even if the number of documents included in the cluster A is smaller than the number of documents included in the cluster B, the number of recently created documents is larger than the cluster B, The size of cluster A may be larger than the size of cluster B. In this case, the representative phrase provider may combine and expose the first candidate representative phrase of the cluster A and the popular keyword.

Similarly, even if the number of documents included in Cluster B is less than the number of documents included in Cluster A, if Cluster B has more recent documents than the documents contained in Cluster A, then the size of Cluster B is the size of Cluster A. May be greater than Then, the representative phrase provider may combine and expose the first representative phrase of the cluster B and the popular keyword.

In this case, the representative phrase provider may not only expose all of the first candidate representative phrases corresponding to the popular keywords on the web page, but also popular keywords activated by a mouse, a touch, a pointer, etc. among the popular keywords exposed on the web page. The first representative phrase of may be exposed on a web page.

For example, as illustrated in FIG. 5, when the mouse points to the AAA keyword 510 in the list 500 including the real-time popular keywords, the representative phrase providing unit may include the first candidate representative phrase for the popular keyword AAA 510 and AAA (see FIG. 5). 521 can be exposed together on the web page.

In this case, the representative phrase provider may expose the representative phrase 521 for the popular keyword together with the photo 522 such as a still image of the popular keyword on the web page 520. To this end, the representative phrase provider may add weights to the documents according to the creation time of the collected documents. For example, the representative phrase providing unit may give a higher weight as the preparation time corresponds to the more recent. Then, the representative phrase provider may extract a document including a picture of a popular keyword from the documents included in the cluster. The representative phrase provider may provide a photo included in the document having the highest weight among the extracted documents in combination with the first candidate representative phrase and the popular keyword for the popular keyword.

In the above, the process of generating the first candidate representative phrase by expanding the representative reference word using the predetermined reference probability has been described, but the reference word expansion unit 432 calculates the reference probability using Equation 2 below. You can also calculate Then, the reference word expansion unit 432 may expand the representative reference word by comparing the calculated reference probability with the conditional probability.

In Equation 2, N (K_Cluster) is the number of documents included in the K cluster, and α is an arbitrary variable. As one example, real numbers such as 0.3, 0.4, 0.5, 0.7, and the like may be used.

으로 계산할 수 있다. According to Equation 2, the reference word expansion unit 432 may calculate a reference probability based on the number of documents belonging to the cluster. For example, when α = 0.5 is used, the reference word expansion unit 432 calculates a reference probability for the cluster A since the number of documents belonging to the cluster A is three.

Can be calculated as

6 is a flowchart provided to explain a method of generating a first candidate representative phrase according to an embodiment of the present invention.

First, in step 610, the document collection unit 410 may collect documents related to the real-time popular keywords to be exposed to the web page through the portal.

In operation 620, the cluster execution unit 420 may analyze the morphemes of the collected documents to perform the cluster.

In operation 630, the reference word determiner 431 may analyze the morphemes of the documents included in the cluster to determine the reference word for each document. In this case, the reference word determiner 431 may count the frequency in which the reference word determined through the morpheme analysis is included in the documents. For example, the frequency of the reference word and the reference word determined for the documents included in the cluster A may be as shown in Table 5 above.

In operation 640, the reference word determiner 431 may determine the representative reference word based on the frequency of the reference word. For example, the reference word determiner 431 may determine two reference words “Hong Kong” and “date” having a high frequency among the reference words determined according to Table 5 as the representative reference words.

Then, in operation 650, the reference word expansion unit 432 may calculate the conditional probability of the representative reference word determined using the bi-gram conditional probability. Here, the conditional probability means a probability that a representative reference word and a word before or after consecutive are included in documents included in the cluster.

In one example, the reference word extension 432 may calculate the conditional probability that the word "actress" continues after the representative reference word "Hong Kong" in the documents included in Cluster A to 2/3 and 1/3. In this case, the reference word expansion unit 432 may select a large conditional probability 2/3 of the calculated conditional probabilities.

In operation 660, the reference word expansion unit 432 may determine whether the selected conditional probability is greater than or equal to a preset reference probability. For example, when the reference probability is preset to 1/2, the reference word expansion unit 432 may determine whether the conditional probability 2/3 of the calculated “Hong Kong actress” is equal to or greater than the reference probability 1/2.

In this case, when the calculated conditional probability is determined to be greater than or equal to the reference probability, the reference word expansion unit 432 may expand the representative reference word into a word corresponding to the selected conditional probability. In operation 680, the reference word expansion unit 432 may continuously expand the extended representative reference word using the same method as that of the 650 to 660. In one example, the reference word extension 432 expands the representative reference word to "BBB" by calculating a conditional probability that "BBB" continues after "actress" and comparing the calculated conditional probability with the reference probability. Can be.

In this case, when the calculated conditional probability is determined to be less than the reference probability, in operation 670, the reference word expansion unit 432 may terminate expansion of the representative reference word.

In one example, the reference word expansion unit 432 may calculate a conditional probability that the consecutive words after the representative reference word "BBB" in the documents included in the cluster A will be included. In this case, since the calculated conditional probabilities are all 1/3 and less than the reference probability 1/2, the reference word expansion unit 432 may terminate the expansion of the representative reference word. The extended representative criterion word can then be "Hong Kong actress BBB".

Similarly, the reference word expansion unit 432 may perform expansion on the representative reference word "date" determined in step 640 in the same manner as "Hong Kong". Then, the reference word expansion unit 432 may expand the representative reference word to "secret date".

In operation 690, the reference word expansion unit 432 may generate the first candidate representative phrase by combining the extended representative reference words. In one example, the reference word extension 432 combines the "Hong Kong actress BBB" extended based on the representative reference word "Hong Kong" and the "Hong Kong actress" expanded based on the representative reference word "date". BBB and Secret Date "may be generated as the first candidate representative phrase.

In this case, when the first candidate representative phrase is determined as the representative phrase, the representative phrase provider may provide a combination of the generated first candidate representative phrase and the popular keyword.

For example, as shown in FIG. 5, the representative phrase provider may expose representative phrases for all popular keywords included in the popular keyword list 500 on a web page. In addition, the representative phrase provider may expose only the representative phrase for the popular keyword activated by the mouse among the popular keywords included in the popular keyword list 500 on the web page.

As another example, the representative phrase provider may expose the first candidate representative phrase 521 together with a picture 522 such as a still image of a popular keyword. At this time, the representative phrase providing unit) adds weights to the documents according to the collection time of the collected documents, and adds the photo included in the highest weighted document to the web page along with the first candidate representative phrases for the popular keywords. May be exposed. In addition, when a photograph is not included in the document having the highest weight, the representative phrase providing unit may use the photograph included in the document having the next highest weight. Here, the representative phrase providing unit may give a higher weight as the creation time of the document corresponds to the more recent.

In FIG. 6, the first candidate representative phrase for one cluster A is generated, and the first candidate representative phrase of the cluster A and the popular keyword are combined and provided. However, the method for providing the representative phrase includes a plurality of representative phrases. A first candidate representative phrase for the cluster of may be generated. When the first candidate representative phrase is determined as the representative phrase, the representative phrase provider may combine at least one of the generated first candidate representative phrases and a popular keyword to expose the web page.

For example, the representative phrase provider may combine the first candidate representative phrases for the generated cluster A and cluster B with the popular keyword "AAA" to expose the web page. In this case, the representative phrase provider may assign exposure priority to the cluster A and the cluster B, and combine the first candidate representative phrase and the popular keyword of the cluster having the high exposure priority to expose the web page.

More specifically, when the exposure priority of the cluster A is 80 and the exposure priority of the cluster B is 60, if a popular keyword "AAA" search is requested, the representative phrase providing unit is popular with the first candidate representative phrase of the cluster A. You can combine the keyword "AAA" to expose it on a web page. In this case, the representative phrase providing unit may determine the exposure priority using the weights of the documents included in the plurality of clusters. For example, the representative phrase provider may assign the exposure priority to the highest sum of the weights of the documents included in the cluster. Then, the representative phrase provider may select the cluster having the highest exposure priority among the plurality of clusters, and combine the first candidate representative phrase of the selected cluster with the popular keyword to expose the web page.

As another example, the representative phrase provider may combine the first candidate representative phrase of one of the plurality of clusters with popular keywords by using the size of the cluster to expose the web page. Here, the size of the cluster is the number of documents belonging to the cluster. That is, the representative phrase providing unit may combine the first candidate representative phrase of the cluster including a large number of documents among the cluster A and the cluster B with the popular keyword to expose the web page.

In this case, the representative phrase providing unit may calculate the size of the document by adding a weight according to the creation time of the document belonging to the cluster. Then, the representative phrase providing unit adds a higher weight to the recently written document, so that even if the number of documents included in the cluster A is smaller than the number of documents included in the cluster B, the number of recently created documents is larger than the cluster B, The size of cluster A may be larger than the size of cluster B. In this case, the representative phrase provider may combine and expose the first candidate representative phrase of the cluster A and the popular keyword.

Similarly, even if the number of documents included in Cluster B is less than the number of documents included in Cluster A, if Cluster B has more recent documents than the documents contained in Cluster A, then the size of Cluster B is the size of Cluster A. May be greater than Then, the representative phrase provider may combine and expose the first candidate representative phrase of the cluster B and the popular keyword.

Up to now, the configuration of generating the first candidate representative phrase and combining the first candidate representative phrase and the popular keyword as the first candidate representative phrase is determined as the representative phrase has been described with reference to FIGS. 4 to 6. . In this case, in FIGS. 4 to 6, two representative reference words are determined in order of high frequency among the plurality of reference words, and the representative reference words are expanded and representative phrases are generated based on the conditional probability. The reference word determiner 431 determines one representative reference word having the highest frequency among the plurality of reference words, and the reference word expander 432 expands the determined representative reference word to represent the representative phrase. You can also create

Hereinafter, referring to FIGS. 7 to 11, when the second candidate representative phrase is determined as the representative phrase, and when the second candidate representative phrase is determined as the representative phrase, the second candidate representative phrase and the popular keyword are combined to provide. This will be explained.

7 is a block diagram illustrating a detailed configuration of a second candidate representative phrase generation unit according to an embodiment of the present invention.

According to FIG. 7, the representative phrase providing system of FIG. 2 may further include a broadcast data collecting unit and a database.

The broadcast data collection unit 710 may collect broadcast data from a broadcast station. Here, the broadcasting station may include airwave TV broadcast, cable TV broadcast, radio broadcast, internet broadcast, and satellite TV broadcast. In this case, the broadcast data collection unit 710 may periodically collect broadcast data every predetermined time, such as every morning, every day lunch, every afternoon, or every Thursday.

The broadcast data collection unit 710 may store the collected broadcast data in the database 720 in the form of Table 6 below.

Broadcast stations Program name Start time End time Performer Name Name in the play Broadcast contents AAA AAA News 20100312 20:00 20100312 21:00 KKK Election ballot BBB2 XXX 20100312 22:00 20100312 23:20 OOO, PPP QQQ, RRR Novy, run away, general BBB1 YYY 20100312 22:00 20100312 23:00 JJJ Glasses secrets BBB2 ZZZ 20100312 21:00 20100312 22:00 VVV, PPP Rice planting

The second candidate representative phrase generator 730 may determine a broadcast program including a popular keyword in broadcast data. In this case, the second candidate representative phrase generator 730 may calculate a difference value as an absolute value of the difference between the exposure time of the popular keyword and the broadcast time of the determined broadcast program. Here, the broadcast time may include a broadcast start time and a broadcast end time.

For example, when the exposure time of the popular keyword is before the broadcast start time, the second candidate representative phrase generator 730 may calculate a difference value as an absolute value of the difference between the exposure time and the broadcast start time.

As another example, when the exposure time of the popular keyword is after the broadcast end time, the second candidate representative phrase generator 730 may calculate a difference value as an absolute value of the difference between the exposure time and the broadcast end time.

As another example, when the exposure time of the popular keyword is included between the broadcast start time and the broadcast end time, the second candidate representative phrase generator 730 may determine a difference between any one of the broadcast start time and the broadcast end time and the exposure time. The difference value can be calculated as an absolute value. In this case, since many searches for the broadcast program are performed 2 hours before the broadcast start time and 2 hours after the broadcast end time, the reference value may be preset to 2 hours.

Then, the second candidate representative phrase generation unit 730 may generate the representative phrase for the popular keyword by comparing the calculated difference value with a predetermined reference value. For example, if the calculated difference value is less than or equal to a predetermined reference value, the second candidate representative phrase generator 730 may generate the second candidate representative phrase based on the determined broadcast time of the broadcast program.

In more detail, the second candidate representative phrase generator 730 may include a broadcast program determiner 731, a matching score calculator 732, and a generator 733.

The broadcast program determiner 731 may match a popular keyword with the broadcast data to determine a broadcast program including the popular keyword in the broadcast data.

For example, when "XXX" is exposed as a popular keyword on a web page, the broadcast program determining unit 731 may determine a broadcast program including the popular keyword "XXX" in the broadcast data of Table 6.

In this case, when the exposure time is before the broadcast start time, the matching score calculator 732 may calculate a difference value as an absolute value of the difference between the broadcast start time and the exposure time of the determined broadcast program. Then, the generation unit 733 may determine that the popular keyword "XXX" has become a popular keyword in real time due to the broadcast program "XXX" when the calculated difference value is equal to or less than the preset reference value.

Accordingly, the generation unit 733 may generate the second candidate representative phrase for the popular keyword based on the broadcast program determined as the cause of the popular keyword.

For example, the generator 733 may generate “broadcast from 22:00” as the second candidate representative phrase based on the broadcast start time of the broadcast program “XXX”.

As another example, when "QQQ" is exposed as a popular keyword on a web page, the broadcast program determiner 731 may determine a broadcast program "XXX" including the popular keyword "QQQ" in the broadcast data of Table 6. In this case, when the exposure time is after the broadcast end time, the matching score calculator 732 may calculate a difference value as an absolute value of the difference between the broadcast end time and the exposure time of the determined broadcast program. Then, the generation unit 733 may determine that the popular keyword "QQQ" becomes a popular keyword in real time due to the broadcast program "XXX" when the calculated difference value is equal to or less than the predetermined reference value.

Accordingly, the generation unit 733 may generate the second candidate representative phrase based on the broadcast program determined as the cause of the popular keyword. For example, the generator 733 may generate “broadcast from 22:00 to 23:20” as the second candidate representative phrase based on the broadcast time of the broadcast program “XXX”.

Meanwhile, when the exposure time is included between the broadcast start time and the broadcast end time, that is, when the exposure time is after the broadcast start time and before the broadcast end time, the matching score calculator 732 starts broadcasting of the determined broadcast program. The difference value may be calculated as an absolute value of the difference between any one of the time and the broadcast end time and the exposure time. The generator 733 may generate the second candidate representative phrase by using the broadcast time of the broadcast program determined according to the difference value.

Hereinafter, referring to FIG. 7, a process of generating a second candidate representative phrase when the popular keyword has a structure in which several words are combined will be described.

First, when a popular keyword having a combination of several words is exposed, the broadcast program determiner 731 may form a plurality of words by morphologically analyzing the popular keyword. The broadcast program determiner 731 may match a word separated from the broadcast data to determine a broadcast program including the word separated from the broadcast data. In this case, the broadcast program determiner 731 may continue to separate the separated words until there is a broadcast program including the words separated from the broadcast data.

For example, when the popular keyword has a structure of "XXX PPP QQQ" in which several words are combined, the broadcast program determiner 731 may morph the popular keyword into "XXX", "PPP", and "QQQ". . The broadcast program determiner 731 may determine a broadcast program including words separated through matching with the broadcast data.

In this case, as an example of using PPP, the broadcast program determining unit 731 may determine "XXX" and "ZZZ" as a broadcast program including the word "PPP" separated from broadcast data. Then, the broadcast program determination unit 731 may end the continuous separation of the separated word "PPP" because there is a broadcast program including the separated words in the broadcast data. Here, when there is no broadcast program including the separated words in the broadcast data, the broadcast program determination unit 731 may continue to separate the separated word "PPP".

In this case, when there are a plurality of broadcast programs including the separated words, the broadcast program determiner 731 may select a broadcast program including the most separated words among the determined broadcast programs.

For example, when "XXX" and "ZZZ" are determined as a broadcast program including the word "PPP" separated from the broadcast data, the broadcast program determining unit 731 is divided into broadcast programs "XXX" and "ZZZ". The number of words can be counted. The broadcast program determining unit 731 then broadcast program "XXX" includes three separate words "XXX", "PPP", and "QQQ", and "ZZZ" represents one separate word "PPP". We can count by including. Accordingly, the broadcast program determiner 731 may select a broadcast program "XXX" that includes the most separated words.

As another example, when the broadcast program determining unit 731 includes a plurality of broadcast programs including the separated words, and the plurality of broadcast programs include the same number of separated words, the broadcast program determining unit 731 may determine the plurality of determined programs. Any one of a plurality of broadcast programs may be selected based on the broadcast time of the broadcast program. In this case, the broadcast program determiner 731 may compare the broadcast start time and the exposure time, the broadcast end time and the exposure time of the plurality of broadcast programs, and select a broadcast program corresponding to the broadcast time closest to the exposure time as a result of the comparison. .

Similarly, even when the popular keyword is composed of one word and there are a plurality of broadcast programs including the popular keyword, the broadcast program determining unit 731 selects a broadcast program that is closest to the exposure time among the broadcast times of the plurality of broadcast programs. Can be.

Then, the matching score calculator 732 may calculate a difference value as an absolute value of the difference between the broadcast time and the exposure time of the selected broadcast program. The generation unit 233 may generate the second candidate representative phrase using the broadcast time of the broadcast program determined based on the difference value. Here, the broadcast time may include a broadcast start time and a broadcast end time. In this case, the process of calculating the difference value is the same as the process of calculating the difference value in the matching score calculation unit 732 described above, if the popular keyword is composed of a single word will be omitted.

Meanwhile, when a plurality of broadcast programs including popular keywords are determined, the broadcast program determiner 731 may select one of the plurality of broadcast programs by using matching scores of the plurality of broadcast programs. Here, the popular keyword may be composed of one word or may be a combination of several words.

In this case, the matching score calculator 732 may add a matching score according to an item including a popular keyword among items constituting the broadcast data, and calculate a final matching score using the matching score and the weight.

First, the matching score calculator 732 adds a matching score for each broadcast program including the popular keyword according to whether the popular keyword among the items constituting the broadcast data is included in the program name, cast member, name in the play, and broadcast content. can do. For example, the matching score calculator 732 may include 100 when a popular keyword is included in a program name among items constituting broadcast data, 80 when included in a cast, 80 when included in a name in a play, and broadcast content. 50 may be added as a matching score.

In addition, the matching score calculator 732 may add a time score for each broadcast program using a time at which a popular keyword is exposed and a broadcast time of a broadcast program including the popular keyword. For example, when the time at which the popular keyword is exposed coincides with the broadcast time within a preset error range, the matching score calculator 732 may add a time score of 50. In this case, the matching score calculator 732 may add a time score by subtracting 50 to -5 points every 10 minutes for the broadcasting time of the broadcast program including the popular keyword from the time when the popular keyword is exposed.

In addition, the matching score calculator 732 may add broadcasting station weights for each broadcasting program according to the type of broadcasting to which the broadcasting program including the popular keyword belongs. In this case, the matching score calculator 732 may add high weights in the order of airwave broadcasting, cable broadcasting, and lion broadcasting. For example, the matching score calculator 732 may add weight 1 to over-the-air broadcast, weight 0.1 to cable broadcast, and 0.05 weight to lion broadcast.

Finally, the matching score calculator 732 may calculate the final matching score for each broadcast program by using a matching score, a time score, and a broadcasting station weight and an audience rating of the broadcast program added for each broadcast program including the popular keyword. For example, the matching score calculator 732 may calculate a final matching score for each broadcast program including a popular keyword by using Equation 3 below.

In Equation 3, P is a matching score added according to a program name, cast, name in the play, and broadcast content, T is time score, S is broadcasting station weight, and A is viewer rating.

According to Equation 3, the matching score calculator 732 may calculate the final matching score for each broadcast program by multiplying the sum of the matching score and the time score by the broadcasting station weight and multiplying the calculated product by the audience rating. For example, when the ratings of the broadcast programs "XXX" and "ZZZ" including the popular keyword PPP are 30% and 10%, the matching score calculation unit 732 multiplies (P + T) × S by 0.3 to give XXX. The final matching score may be calculated, and the final matching score of ZZZ may be calculated by multiplying (P + T) × S by 0.1.

Then, the broadcast program determiner 731 may select a broadcast program having the highest final matching score calculated among broadcast programs including popular keywords.

Subsequently, the generation unit 733 may generate the second candidate representative phrase for the popular keyword based on the broadcast time of the selected broadcast program. For example, as illustrated in FIG. 8, the generation unit 733 may generate “broadcast from 22:00” as the second candidate representative phrase 821 based on the broadcast time of the selected broadcast program “XXX”.

In this case, when the second candidate representative phrase is determined as the representative phrase, the representative phrase provider may provide a combination of the second candidate representative phrase 821 and the popular keyword 810. For example, the representative phrase provider may expose a photo 822 such as a still image of a popular keyword together with the second candidate representative phrase 821 on a web page. In this case, the representative phrase provider may expose the representative phrase and the picture of the popular keyword activated by the mouse, the pointer, etc. in the list 800 including the real-time popular keywords on the web page.

9 is a flowchart provided to explain a process of providing a second candidate representative phrase using broadcast time according to an embodiment of the present invention.

Referring to FIG. 9, in operation 910, the broadcast data collector 720 may periodically collect broadcast data from a broadcast station and store the broadcast data in the database 710. Here, the broadcasting station may include airwave TV broadcast, cable TV broadcast, radio broadcast, internet broadcast, and satellite TV broadcast.

In operation 920, the broadcast program determiner 731 may determine a broadcast program including a popular keyword in broadcast data. In one example, the broadcast program determiner 731 may determine a broadcast program "XXX" including the popular keyword "QQQ" exposed to the web page from the broadcast data having the configuration shown in Table 6.

In this case, in step 930, when the time at which the popular keyword is exposed is not included between the determined broadcast start time and the broadcast end time of the broadcast program, in step 940, the matching score calculator 732 may determine the broadcast time of the determined broadcast program. The difference value can be calculated as the absolute value of the difference between exposure times.

For example, when the exposure time is before the broadcast start time or after the broadcast end time, the matching score calculator 732 may determine the difference between the broadcast start time and the exposure time of the determined broadcast program, or the difference between the broadcast end time and the exposure time. The value can be calculated.

In operation 950, the matching score calculator 732 may compare whether the calculated difference value is equal to or less than a predetermined reference value. For example, since a keyword related to a broadcast program has a high probability of being searched as a popular keyword for two hours before the broadcast start time and two hours after the broadcast end time, the reference value may be preset to two hours.

In this case, when the difference value is less than or equal to the reference value, in step 960, the generation unit 733 may generate a second candidate representative phrase for the popular keyword based on the determined broadcast time of the broadcast program.

For example, when the exposure time corresponds to the broadcast start time of the determined broadcast program "XXX", the generation unit 733 uses the determined broadcast start time of the broadcast program "XXX" to perform "broadcast start from 22:00". The second candidate representative phrase may be generated.

As another example, when the exposure time corresponds to after the broadcast end time of the determined broadcast program "XXX", the generation unit 733 starts from "22:00" using the broadcast start time and broadcast end time of the determined broadcast program "XXX". Broadcast until 23:20 ”may be generated as the second candidate representative phrase.

In this case, when the second candidate representative phrase is determined as the representative phrase in the representative phrase determination unit, the representative phrase provider may provide a combination of the second candidate representative phrase and the popular keyword. For example, the representative phrase provider may combine the popular keyword "QQQ" with the second candidate representative phrase "start broadcasting from 22:00" to expose the web page. At this time, the representative phrase provider may combine the pictorial photo of the broadcast program "XXX" of the popular keyword "QQQ" with the popular keyword and the second candidate representative phrase to expose the web page.

In operation 960, when there are a plurality of determined broadcast programs, the broadcast program determiner 931 may select any one broadcast program based on the determined broadcast time of the broadcast program. For example, the broadcast start time and exposure time of the determined broadcast programs may be compared, and the broadcast end time and exposure time of the determined broadcast programs may be compared to select a broadcast program having a broadcast time closest to the exposure time. Then, the generation unit 733 may generate the second candidate representative phrase by using the broadcast time of the selected broadcast program.

FIG. 10 is a flowchart provided to explain a process of generating a second candidate representative phrase for a popular keyword having a combination of several words.

Referring to FIG. 10, in operation 1010, the broadcast program determiner 731 may separate a plurality of words that have been combined by morphologically analyzing a popular keyword having a combination of words.

For example, when the popular keyword has a structure of "XXX PPP QQQ" in which several words are combined, the broadcast program determiner 731 may morph the popular keyword into "XXX", "PPP", and "QQQ". .

In operation 1020, the broadcast program determiner 731 may determine a broadcast program including words separated from the broadcast data. For example, the broadcast program determiner 731 may determine "XXX" and "ZZZ" as a broadcast program including the word "PPP" separated from the broadcast data. Then, the broadcast program determiner 731 may count the number of separated words included in the broadcast programs "XXX" and "ZZZ". The determination unit 231 counts that the broadcast program "XXX" includes three separate words "XXX", "PPP", and "QQQ", and "ZZZ" includes one separate word "PPP". can do. Accordingly, the broadcast program determiner 731 may select a broadcast program "XXX" that includes the most separated words.

In this case, when there are a plurality of determined broadcast programs in step 1030, and there are a plurality of broadcast programs including a maximum of separated words in step 1040, in step 1050, the broadcast program determining unit 731 broadcasts among the determined broadcast programs. The broadcast program whose time is closest to the exposure time of the popular keyword can be selected. In this case, the broadcast program determiner 731 may select one of the broadcast programs closest to the exposure time by comparing the determined broadcast start time and broadcast end time of the plurality of broadcast programs with the exposure time.

In this case, in step 1040, if there are not a plurality of broadcast programs that include the separated words, in step 1060, the broadcast program determiner 731 may select a broadcast program that includes the separated words. For example, since the broadcast program "XXX" includes three separated words and the broadcast program "ZZZ" includes one separated word, the broadcast program determiner 731 may select "XXX".

In operation 1070, the matching score calculator 732 may calculate a difference between the broadcast time and the exposure time of the selected broadcast program. In this case, when the difference is less than or equal to the preset reference value, in operation 1090, the generation unit 733 may generate the second candidate representative phrase by using the broadcast time of the selected broadcast program. In addition, when the second candidate representative phrase is determined as the representative phrase in the representative phrase determination unit, the representative phrase provider may provide a combination of the second candidate representative phrase and the popular keyword. Here, the operations of steps 1070 to 1090 are overlapped with steps 940 to 960 described above, and thus a detailed description thereof will be omitted.

11 is a flowchart provided to explain a process of generating a second candidate representative phrase using a matching score according to an embodiment of the present invention.

According to FIG. 11, in operation 1110, the broadcast program determiner 731 may determine a broadcast program including a popular keyword from the collected broadcast data.

In operation 1020, the matching score calculator 732 may add a matching score for each of the determined broadcast programs. At this time, the matching score calculator 732 adds a matching score for each broadcast program including the popular keyword according to whether the popular keyword among the items constituting the broadcast data is included in the program name, the cast member, the name in the play, and the broadcast content. can do.

For example, the matching score calculator 732 includes 100 when a popular keyword is included in a program name among items constituting broadcast data, 80 when included in a cast, 80 when included in a name in a play, and broadcast content. 50 may be added as a matching score.

In operation 1130, the matching score calculator 732 may add a time score for each determined broadcast program.

For example, when the time at which the popular keyword is exposed coincides with the broadcast time of the determined broadcast program, the matching score calculator 732 may add a time score of the matched broadcast program to 50. At this time, even if the exposure time of the popular keyword does not exactly match the broadcast time, the matching score calculator 732 may add a time score of 50 if it matches within a preset error range.

When the broadcasting time of the broadcast program and the exposure time of the popular keyword do not match, the matching score calculator 732 may add a time score by subtracting -5 points from 50 every 10 minutes. For example, when the broadcast start time of the broadcast program and the exposure time of the popular keyword are 30 minutes apart, the matching score calculator 732 may add the time score to 35.

Subsequently, in operation 1140, the matching score calculator 732 may add broadcasting station weights for each broadcasting program. For example, the matching score calculator 732 may add weights differently according to whether the determined broadcast program is over-the-air broadcast, cable broadcast, or radio broadcast.

In operation 1150, the matching score calculator 732 may calculate a final matching score for each broadcast program. In this case, the matching score calculator 732 may calculate the final matching score for each broadcast program using the determined rating for each broadcast program as shown in Equation 3 above.

For example, the matching score calculator 732 may calculate a sum of the added matching score and the time score, and calculate a final matching score for each broadcast program determined by multiplying the calculated sum by a broadcasting station weight and an audience rating.

Then, in operation 1160, the broadcast program determiner 731 may select one of the broadcast programs determined based on the calculated final matching score. For example, the broadcast program determiner 731 may select a broadcast program having the highest calculated final matching score.

In operation 1170, the matching score calculator 732 may calculate a difference between the broadcast time and the exposure time of the selected broadcast program.

In operation 1180, when the difference is less than or equal to the preset reference value, in operation 1190, the generation unit 733 may generate the second candidate representative phrase by using the broadcast time of the selected broadcast program.

In this case, when the second candidate representative phrase is determined as the representative phrase, the representative phrase provider may provide a combination of the generated second candidate representative phrase and the popular keyword. Here, since operations of steps 1170 to 1190 overlap with steps 940 to 960 described above, a detailed description thereof will be omitted.

As described above, the representative phrase providing system according to an embodiment of the present invention may generate the first candidate representative phrase based on the document, and generate the second candidate representative phrase based on the broadcast data. Then, the final representative phrase for the popular keyword may be determined for either the first candidate representative phrase or the second candidate representative phrase based on the final popularity or the time score of the second candidate representative phrase. In addition, the representative phrase providing system may combine the determined representative phrase and the popular keyword and expose the web page.

In this case, when either one of the first candidate representative phrase and the second candidate representative phrase is not determined as the representative phrase for the popular keyword, the representative phrase providing system may match all of the first candidate representative phrase and the second candidate representative phrase with the popular keyword. Can be provided in combination. Also, the representative phrase providing system may randomly select any one of the first candidate representative phrase and the second candidate representative phrase to determine the final representative phrase for the popular keyword. In addition, the representative phrase providing system may expose only the popular keyword on the web page without determining all of the first candidate representative phrase and the second candidate representative phrase as the final representative phrase for the popular keyword. That is, in FIGS. 5 and 8, only the popular keyword list without the representative phrase may be exposed on the web page.

The methods according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

200: representative phrase providing system
210: first candidate representative phrase generation unit
220: second candidate representative phrase generation unit
230: popularity calculation unit
240: representative phrase determination unit
250: time score calculator
260: representative phrase provider

Claims (32)

In the representative phrase providing method performed by the representative phrase providing system comprising a first candidate representative phrase generation unit, a second candidate representative phrase generation unit and a representative phrase determination unit,
Generating, by the first candidate representative phrase generation unit, a first candidate representative phrase for a popular keyword based on a document;
Determining, by the second candidate representative phrase generation unit, a broadcast program including the popular keyword in broadcast data and generating a second candidate representative phrase for the popular keyword by using the determined broadcast time of the broadcast program; And
Determining, by the representative phrase determination unit, a final representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase.
Representative phrases providing method comprising a. The method of claim 1,
Determining the final representative phrase,
And the representative phrase determination unit determines that a higher degree of suitability for a popular keyword among the first candidate representative phrase and the second candidate representative phrase is the final representative phrase. The method of claim 1,
The representative phrase providing system further includes a popularity calculation unit,
Calculating, by the popularity calculator, the popularity of the document used to generate the first candidate representative phrase.
Representative phrases providing method further comprising. The method of claim 3,
Determining the representative phrase,
The representative phrase determining unit determines the first candidate representative phrase that is determined to have a higher suitability for the popular keyword by using the popularity of the document and a predetermined reference popularity as a final representative phrase. . The method of claim 2,
The representative phrase providing system further includes a time score calculator,
Calculating, by the time score calculator, a time score of the second candidate representative phrase by using a broadcast time of broadcast data and an exposure time of a popular keyword used to generate the second candidate representative phrase.
Representative phrases providing method further comprising. The method of claim 5,
Determining the representative phrase,
Wherein the representative phrase determination unit determines the second candidate representative phrase as the final representative phrase by using the time score and the preset reference time score of the second candidate representative phrase as being more suitable for the popular keyword. How to provide a representative phrase. The method of claim 1,
Generating the first candidate representative phrase,
Determining a reference word of documents including the popular keyword by the reference word determination unit of the first candidate representative phrase generation unit;
Determining, by the reference word determiner, the representative candidate word from the first candidate representative phrase generation unit;
Expanding, by the reference word expansion unit, the first candidate representative phrase generation unit by combining the representative reference word with a continuous word before or after the representative reference word; And
Generating the first candidate representative phrase by using the representative reference word expanded by the reference word extension of the first candidate representative phrase generation unit;
Representative phrases providing method comprising a. The method of claim 7, wherein
Expanding the representative reference word,
Calculating, by the reference word expansion unit, the first candidate representative phrase generation unit a conditional probability that consecutive words are included before or after the representative reference word in documents including the popular keyword; And
Expanding, by the reference word expansion unit, the representative candidate word based on the conditional probability by the first candidate representative phrase generation unit;
Representative phrases providing method comprising a. The method of claim 7, wherein
Determining the representative reference word,
Counting the frequency of the reference word by analyzing a morpheme of documents including the popular keyword by the reference word determination unit of the first candidate representative phrase generation unit; And
Determining the representative reference word based on the counted frequency of the reference word determination unit of the first candidate representative phrase generation unit;
Representative phrases providing method comprising a. The method of claim 1,
The representative phrase providing system further includes a document collecting unit and a cluster performing unit,
Collecting, by the document collecting unit, documents including the popular keyword; And
Performing a cluster on the collected documents by the cluster execution unit;
Further comprising:
Generating the first candidate representative phrase,
And the first candidate representative phrase generation unit generates the first candidate representative phrase by expanding a representative reference word based on documents belonging to the cluster. The method of claim 10,
Generating the first candidate representative phrase,
Adding, by the representative phrase providing unit, weights to the collected documents; And
Determining, by the representative phrase provider, an exposure priority of a cluster by using the weight;
Representative phrases providing method further comprising. delete The method of claim 1,
Generating the second candidate representative phrase,
When the broadcast program determining unit of the second candidate representative phrase generation unit has a plurality of the broadcast programs, one of the plurality of broadcast programs, the broadcast start time and the broadcast end time of the broadcast program are the ones closest to the time when the popular keyword is exposed. Steps to choose a program
Further comprising:
Generating the second candidate representative phrase using the broadcast time may include:
And a generation unit of the second candidate representative phrase generation unit generates the second candidate representative phrase using the broadcast time of the selected broadcast program. The method of claim 1,
The second representative phrase generation unit further includes a broadcast program determination unit and generation unit,
Generating the second candidate representative phrase,
When the broadcast program determining unit of the second candidate representative phrase generation unit is in the form of combining the popular keywords into several words, morphologically analyzing the popular keywords and separating the plurality of words into a plurality of words;
Selecting, by the broadcast program determiner, the second candidate representative phrase generation unit based on the plurality of words separated from the broadcast programs including the popular keyword; And
Generating a second candidate representative phrase by using the generation time of the selected broadcast program by the generation unit of the second candidate representative phrase generator;
Representative phrases providing method comprising a. The method of claim 1,
The second representative phrase generator further includes a matching score calculator, a broadcast program determiner, and a generator.
Generating the second candidate representative phrase,
Adding a matching score, a time score, and a broadcast station weight to a broadcast program including the popular keyword by the matching score calculator of the second representative phrase generation unit;
Calculating, by the matching score calculator of the second representative phrase generation unit, a final matching score for the broadcast program using at least one of the matching score, the time score, the broadcasting station weight, and the viewer rating of the broadcast program;
Selecting, by the broadcast program determination unit of the second representative phrase generation unit, any one broadcast program based on the final matching score from the broadcast programs including the popular keyword; And
Generating the second candidate representative phrase by using the generation time of the selected broadcast program by the generation unit of the second representative phrase generation unit;
Representative phrases providing method comprising a. The method of claim 1,
The representative phrase providing system further includes a representative phrase providing unit,
Providing the representative phrase providing unit combining the final representative phrase and the popular keyword
Representative phrases providing method further comprising. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1 to 11 and 13 to 16. A first candidate representative phrase generation unit generating a first candidate representative phrase for a popular keyword based on the document;
A second candidate representative phrase generation unit for determining a broadcast program including the popular keyword in broadcast data and generating a second candidate representative phrase for the popular keyword using the determined broadcast time of the broadcast program; And
A representative phrase determination unit that determines a final representative phrase for the popular keyword by using the first candidate representative phrase and the second candidate representative phrase.
Representative phrase providing system comprising a. 19. The method of claim 18,
The representative phrase determination unit,
Representative phrase providing system, characterized in that the final representative phrase is determined that the higher suitability of the popular keyword among the first candidate representative phrase and the second candidate representative phrase. 20. The method of claim 19,
Popularity calculation unit for calculating the popularity of the document used to generate the first candidate representative phrase
Representative phrase providing system further comprising. 21. The method of claim 20,
The representative phrase determination unit,
And a first candidate representative phrase determined as having a higher suitability for the popular keyword as the final representative phrase using the popularity of the document and a predetermined reference popularity. 19. The method of claim 18,
A time score calculator for calculating a time score of the second candidate representative phrase by using the broadcasting time of the broadcast data and the exposure time of the popular keyword used to generate the second candidate representative phrase
Representative phrase providing system further comprising. The method of claim 22,
The representative phrase determination unit,
Providing the representative phrase, characterized in that the second candidate representative phrase is determined to have a higher suitability for the popular keyword by using the time score and the predetermined reference time score of the second candidate representative phrase as the final representative phrase. system. 19. The method of claim 18,
The first candidate representative phrase generation unit,
A reference word determination unit determining a reference word by analyzing the morphemes of documents including the popular keyword and determining a representative reference word according to the determined frequency of the reference word; And
A reference word extension unit which expands the representative reference word by combining a continuous word before or after the representative reference word and the representative reference word, and generates the first candidate representative phrase using the extended representative reference word
Representative phrase providing system comprising a. 25. The method of claim 24,
The reference word expansion unit,
Calculating conditional probabilities including consecutive words before or after the representative criterion word in documents including the popular keyword, and extending the representative criterion word based on the conditional probability. system. 25. The method of claim 24,
The reference word determination unit,
And analyzing the morphemes of the documents including the popular keyword to count the frequency of the reference word and determine the representative reference word based on the counted frequency. 19. The method of claim 18,
A document collector configured to collect documents including the popular keyword; And
Cluster execution unit that performs a cluster on the collected documents
Further comprising:
The first candidate representative phrase generation unit,
And the first candidate representative phrase is generated by extending a representative reference word based on documents belonging to the cluster. delete 19. The method of claim 18,
The broadcast program determination unit,
When there are a plurality of broadcast programs, the broadcast start time and the broadcast end time of the broadcast program are selected from among the plurality of broadcast programs, the broadcast program being closest to the time when the popular keyword is exposed,
The second candidate representative phrase generation unit,
And the second candidate representative phrase is generated using the broadcast time of the selected broadcast program. 19. The method of claim 18,
The second candidate representative phrase generation unit,
When the popular keyword is in the form of a combination of words, the popular keyword is morphologically analyzed to be divided into a plurality of words, and any one broadcast is based on a plurality of words separated from the broadcast program including the popular keyword. A broadcast program determination unit for selecting a program; And
A generator configured to generate the second candidate representative phrase using the broadcast time of the selected broadcast program
Representative phrase providing system comprising a. 19. The method of claim 18,
The second candidate representative phrase generation unit,
A matching score, a time score, and a broadcast station weight are added to a broadcast program including the popular keyword, and a final matching score for the broadcast program is obtained by using at least one of the matching score, time score, broadcast station weight, and viewer rating of the broadcast program. Matching score calculator to calculate;
A broadcast program determiner configured to select one broadcast program based on the final matching score among broadcast programs including the popular keyword; And
A generator configured to generate the second candidate representative phrase using the broadcast time of the selected broadcast program
Representative phrase providing system comprising a. 19. The method of claim 18,
Representative phrase providing unit providing the final representative phrase and the popular keywords combined
Representative phrase providing system further comprising.

Images