KR101566616B1 - Advertisement decision supporting system using big data-processing and method thereof - Google Patents

Abstract

The present invention relates to a system for making an advertisement decision through big data processing. The system of the present invention comprises: a search and collection unit which searches for and collects online data including a target brand by crawling or screen-scraping online data including a web page; a filter unit which extracts time periods when keywords of the target brand, related keywords related to the keywords, reputation keywords which express reputation of the target brand, and online data including the reputation keywords are generated, from the data searched and collected by the search and collection unit; a recommendation management unit which recommends advertisement keywords for the target brand in consideration of a change in the number of the reputation keywords extracted according to the time period; and a database which saves the data searched and collected by the search and collection unit, the keywords extracted from the data by the filter unit, the related keywords, the reputation keywords and the time periods, and a reputation dictionary referred to extract the reputation keywords.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an advertising decision system,

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to an advertising decision system, and more particularly, to a system and method for advertising decision making through big data processing.

With the expansion of new types of multimedia content, social network services (SNS), and the dissemination and use of smart devices, the size of data generated and distributed on the web is increasing exponentially. A tremendous amount of data that exists and is still growing on the Web can be used to interpret the world. This is the Big Data. Big data means a vast amount of information digitized. It can filter out unnecessary data from big data and extract and analyze only useful information to read people's thoughts, opinions and trends and to predict their behavior in advance. Big Data is one of the next generation information technology (IT) technology that is not only in Korea but also in the world because of its usefulness.

The domestic big data market is expected to reach 300 billion won in 2015 and grow to 1 trillion won in 2020. The domestic market related to Big Data is growing by more than 28% every year. The use of Big Data is concentrated in the research and consulting fields, but it is likely to be used in commercial and advertising markets as well

Currently, the use of big data technology in the field of advertising marketing is actively conducted by some large companies, but it is not popularized because it is costly to apply to all companies. There is also a need for further discussion on how to effectively use Big Data in various areas where Big Data can be used.

Korean Patent Publication No. 10-2013-0051877 [Title: Big Data Extraction System and Method]

A first object of the present invention is to provide an advertising decision system through big data processing.

A second object of the present invention is to provide a method for making advertisement decision through big data processing.

According to an aspect of the present invention, there is provided an advertisement decision system using big data processing, which includes crawling or screen scraping data including a web page on-line, A search collection unit for searching and collecting online data including a target brand; A keyword associated with the keyword, an associated keyword associated with the keyword, a reputed keyword indicating a reputation for the target brand, and a time when online data including the reputable keyword is generated or an affirmative Trends, major issue keywords, etc .; A recommendation management unit for recommending an advertisement keyword for the target brand in consideration of a change in the number of the reputed keywords extracted according to a time; And a database for storing the data retrieved and collected by the search and collecting unit and a reputation dictionary referred to for extracting the keyword, the related keyword, the reputed keyword and the timing extracted by the filter unit from the data, and the reputed keyword, .

In addition, the filter unit assigns a rating for each of the reputed keywords or positive negative tendencies to each reputable keyword in consideration of the reputation level of the extracted reputed keyword, Recommending an advertisement keyword for a brand, and the database may further include a reputation step according to a reputed keyword.

In addition, the filter unit further extracts the kind of on-line channel on which the data is searched and collected from the data collected and collected by the search and collecting unit, and the recommendation management unit, when the keyword and the associated keyword are exposed The type of the channel to which the advertisement keyword is to be advertised is further recommended, and the database may further store the kind of on-line channel on which the keyword is searched and collected.

If the number of negative reputation keywords increases for the target brand, the recommendation management unit recommends a relative advertisement keyword of a recommended advertisement keyword with reference to the database, It is possible to store a relative advertisement keyword.

In addition, the recommendation management unit may recommend the advertisement keyword considering the contrast exposure index of the advertisement that divides the number of exposures on the online for a predetermined period of time for the advertisement keyword by the advertisement cost.

The recommendation management unit may change the recommended advertisement keyword if the variation amount of the number of exposures of the keyword, the related keyword, or the reputable keyword for a predetermined period exceeds a predetermined threshold range have.

In addition, the recommendation management unit may derive an issue keyword among the collected keywords including the keyword and the related keyword input by the user, and recommend that the advertisement is executed on the issue keyword.

In order to achieve the second object of the present invention, there is provided an advertisement decision method using big data processing according to one aspect of the present invention, including: , The online bulletin board, the bulletin of the online bulletin board, the web page on the online channel including the comment on the bulletin, and the information on the SNS (Twitter, Facebook) A first step of storing the data; A keyword of the target brand, an associated keyword associated with the keyword, a reputation keyword representing the reputation of the target brand referring to the reputation dictionary, and a type of online channel in which the data is searched and collected A second step of extracting and storing the time and the degree of involvement of the data; A third keyword for recommending an advertisement keyword for the target brand in consideration of at least one of a change amount of the keyword according to the extracted time series, a change amount of the related keyword, a change amount of the reputation keyword, Step < / RTI >

The second step further extracts the user's preference information on the web page, the number of comments on the web page, the preference information on the comment, and the number of comments on the comment from the data, Wherein the step of referring to the preference information of the web page of the user, the number of comments of the web page, the preference information of the comment, and the number of comments on the comment, , The related keyword, and the reputation keyword, to recommend the advertisement keyword. If the selection of the existing advertisement target keyword is based on the traffic information related to the advertisement target keyword such as the number of hits and the advertisement execution rate provided by the service provider such as the search engine, the present invention can collect and filter the reputation- Analyzing the online big data of the user who refers to the brand, extracting the issue keyword for the target brand of the target customer, and recommending that the advertisement is executed on the keyword targeted for the target of the target customer .

In the second step, the number of sentences including both the keyword of the target brand and the related keyword or the keyword of the target brand and the reputable keyword are included in the data, And the third step further differentiates the weights of the keyword, the related keyword and the reputed keyword of the target brand included in the sentence according to the value of the parameter, Recommended.

The second step may include the number of web pages in the web page spreader including the link information linking to the web page including the information about the target brand, the number of views of the web page, the number of comments of the web page, , The number of comments on comments, the number of other users or websites that have a relationship through the social network service, and the number of times that a web page having the same contents as the web page started by the web page spreader is exposed on-line The number of comments of the web page, the number of comments of the web page, the number of comments of the comment, the number of the comments of the comment, The number of other users or websites, the same number of web pages as the web page initiated by the web page spreader The web page spreader calculates a weight of the web page spreader in consideration of the number of times the web page for the web page is exposed on-line, and provides a notice to the user about the web page newly created by the web page spreader, The keyword of the target brand, the related keyword, and the weight of the reputable keyword included in the target brand may be differentiated to recommend the advertisement keyword.

In addition, the method may further include providing a notification to a user when a reputable keyword including negative reputation information increases, and may further include a reputation of a competitor's brand in addition to the own brand, To compare and analyze important competitors, and to analyze market share in the market.

In addition, by providing a guide to the user's reputation (positive / negative) about the brand keyword and the related keyword inputted by the customer, it is possible to grasp the negative tendency of the online reputation text in real time, Notification of the incidence, request for deletion, reputation risk management, and so on.

As described above, according to the embodiment of the present invention, the advertisement decision system and the advertisement decision method through the big data processing can be provided.

1 is a conceptual diagram illustrating a configuration of an advertisement decision system through a big data process according to an embodiment of the present invention.
2 is a diagram illustrating a change in the number of related keywords according to an embodiment of the present invention.
FIG. 3 and FIG. 4 are graphs showing changes in the moving average exposure count and the current day exposure count according to the embodiment of the present invention.
FIG. 5 is a diagram illustrating a change in flat information according to an exemplary embodiment of the present invention. Referring to FIG.
6 and 7 are diagrams illustrating a procedure of an advertisement decision method according to another embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the terms "comprises" or "having" and the like are used to specify that there is a feature, a number, a step, an operation, an element, a component or a combination thereof described in the specification, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Hereinafter, the same reference numerals will be used for the same constituent elements in the drawings, and redundant explanations for the same constituent elements will be omitted.

1 is a conceptual diagram illustrating a configuration of an advertisement decision system through a big data process according to an embodiment of the present invention.

1, the advertisement decision system 100 according to the present invention includes a search and collection unit 110, a filter unit 130, a recommendation management unit 150, and first to fourth databases 111, 131, 133, 151).

First, the search collecting unit 110 searches for and extracts a web page. A web page is a configuration in which a service provider or the like allows ordinary users to freely input text or to display contents according to their interests. For example, a web page is a web page, a news, a blog, a microblog, , Bulletin boards, comments, etc., and the SNS activity information. The search and collection unit searches and collects web pages about related brands for target brands or target brands stored in the online space. The web page may include a target brand and reputation information of a related brand. The target brand may be a brand that an advertiser desires to advertise, and the related brand may be a brand that is in competition with the target brand. Keywords related to the target brand and related brands may be exposed on the web page, and keywords that are similar or related to the keywords through analysis may be defined as related keywords. Also, through the analysis of the recommendation management section described below, it is possible to define a keyword to be recommended to an advertiser as an advertisement keyword for efficient advertisement.

The preferred online space is a social network service site such as a portal site where the target brand and related brand reputation information can be mainly uploaded, a user-designated site, a media site, a user meeting site on an online cafe, And the like. The web page including reputation information about the target brand or the related brand can be generally searched by using the keywords of the target brand and related brands, the similar keyword for the keyword, the related keyword for the keyword, and the like. On the other hand, the user can input and set the target brand or the related brand into the advertisement decision system according to the present invention in the form of a keyword or the like.

First, the search collecting unit can collect a new web page or an updated web page through a crawler by a web crawler. Web crawlers can be web robots, bots, and so on. They can crawl billions of web pages such as portal sites and custom sites by using a large number of computers to display web pages including reputation information about target brands and related brands in real time . In addition, the search and collecting unit can collect screen scraping on the web page. Screen scraping is a technique for extracting only necessary data from a plurality of web pages. When screen scraping is used, the user can automatically access the server of the web site and display only the necessary data. That is, it is possible to extract or extract necessary information from the contents displayed on the internet screen, and separately manipulate the contents. Thus, it is possible to search the necessary part of the huge amount of web pages including the target brand and reputation information about the related brands from time to time or periodically And can bring it. Since the search and collecting unit uses the screen scraping technique, even if the portal site in the online space, the social network service site such as Facebook, or the like, which is the object of collection, changes the form or logic of exposing the web page, You can search and extract. The data thus retrieved and extracted can be stored in the first database 111.

The first database 111 stores the search date and time of the web page searched and extracted by the search and collection unit, the type of the searched online channel, the title of the web page, the text, the publisher, the publication date, The type of portal site, the number of comments and the number of web pages, and the like. In addition, the search and collecting section collects web pages by various methods such as collecting the Internet access history list, analyzing the log file recorded on the web server, or collecting the internet access address by the program installed on the computer of the Internet user can do.

The filter unit 130 filters the web pages retrieved and extracted by the retrieval and collecting unit based on a predetermined criterion, the second database 131 stores information filtered by the filter unit, The third database 133 stores information that can be referred to when the web page is filtered by the filter unit.

First, the filter unit extracts the retrieved web page from the keyword frequency of the web page, the portal site that exposed the web page in the online space, the type of the channel such as the social network service, the target brand of the web page, The web page can be classified based on the web page. Meanwhile, the second database 131 may store the filtered data by the filter unit through interlocking with the first database. More specifically, the filter unit can perform keyword frequency analysis on the retrieved web page. For example, a keyword can be extracted from a plurality of Web site documents using a TFIDF (Term Frequency Inverse Document Frequency) method. TFIDF (Term Frequency Inverse Document Frequency) refers to the frequency of an electronic document including a specific term in a set of global documents, that is, a specific frequency of a document frequency (DF) Means the frequency of a term frequency in a document, that is, a term frequency (TF). This is because TFIDF is larger for keywords appearing in a specific document but less frequently in a general document, so using the TFIDF method is useful for extracting keywords from an electronic document. In addition, keywords can be extracted on a web page in various ways. The keyword may be related to a target brand or a related brand. For example, when the target brand is " A car ", the search collecting unit searches and extracts a web page related to " A car " You can derive model name, sales volume, promotion schedule, service shop, etc. of 'A car'. Likewise, keywords related to 'B car', which can be considered as a related brand of target brand, can be derived. In other words, the filter unit is, for example, a web page in which a related keyword a, b, c, d, etc. related to a target brand 'A car' is searched. A car - a, A car - b, A car - - d to construct a second database by constructing a keyword set. The related keywords such as a, b, c, d and the like can be recommended as advertisement keywords by assigning weights and the like through the recommendation management unit to be described below. In this way, the filter unit can generate an optimal related keyword and pair that can be recommended by reflecting trends appearing on a recent web page. In addition to the information provided by a service provider such as a portal site, You may.

In addition, a ranking structure may be provided between the related keywords for the target brand. In this case, the filter unit extracts one or more related keywords as representative related keywords through natural language processing and a text mining technique, sets a related keyword set, and allows the extracted extracted related keywords to be included in the set, Create a structure. For example, if the target brand is 'A school' and related keywords are 'Hong Gil Dong', 'Hong Gil Dong', 'Hong Gari', 'Hong Math', 'Hong Ssam', 'Hong Gil Dong Sam' And 'Gil - dong' are collected. In this case, the filter unit sets a related keyword 'Hong Kil-dong' as a representative related keyword through natural language processing, and the related keywords' Hong Kil-dong 'and' Hong Gari 'including' , 'Hong Matsu', 'Hongseum', 'Hong Gil Dong', 'Gil Dong', and 'Gil Dong Sam'. On the other hand, it is also possible to take into account the ranking structure among the related keywords stored in advance by the user in order to generate the ranking structure. For example, the related keywords of 'Hong Gil Dong', 'Hong Gil Dong', 'Hong Gari', 'Hong Math', 'Hong Ssam', 'Hong Gil Dong', 'Gil Dong', and ' Quot; Hong Gil-dong ", and the remainder may be preset to be included in the set represented by " Hong Gil-dong ". The data on the ranking structure may be stored in the third database 133 and referred to.

In addition, the filter unit can store the collected keywords by category. In other words, various keywords included in a plurality of web pages are converted into databases by category. In the above example, categories may be stored between related keywords a, b, c, and d associated with the target brand 'A automobile'. For example, if the related keywords a and b are related to the model name, c is related to the price, and d is related to the promotion schedule, the related keywords a and b are one category of the model name, It can be classified into categories. The category relationship to which the associated keyword belongs may be stored in advance in the third database and referred to.

In addition, the time series information regarding the keywords can be stored in the second database. For example, the frequency with which a vehicle's model name is exposed on a web page among keywords for the target brand 'A car' may be different from time to time. For example, storing information such as keyword exposures along with time-series information such as the spring season in which demand for automobiles is increasing and the season for not in the past can be analyzed to analyze changes in keyword exposure over time.

 Further, the filter unit may filter the target brand and the related brand in which channel on the online space is exposed, and store the information in the second database. The channel may be news, a social network service, a cafe, a blog, a general bulletin board, a personal homepage, a microblog, a video, a knowledge search, a cafe, a bulletin board, a comment, . For example, a social network service site can be divided into specific websites such as Facebook and Twitter to store exposure information of a target brand and related brands according to each channel. In addition, a source information web page including a generally generated content such as a news or a knowledge search can be distinguished from a spread web page circulating the source information page. For example, if the source information web page is a web page including news actually produced by a newspaper company or the like, the spread web page generally includes link information for accessing the source information web page such as Twitter or Facebook It can be called a web page. For example, a tweet retweeted among tweeters may correspond to the spread web page. This is because information that can be retrieved from a Twitter with a large number of followers can spread widely online. In this way, the filter unit can divide a web page into a source information web page and a spread web page, and store the filtered web page. In addition, the filter unit may store information on the number of web pages including link information of a specific source information web page retrieved and collected by the search collecting unit. The information and the like may be stored in the second database.

The channel information thus collected can be utilized for decision-making on the advertisement. The filter unit may filter the channel information by adding time series information to the channel information, and may filter the time series information by adding the time series information for each keyword.

In addition, the filter unit may filter the reputation information of the target brand and the related brand in the web page and store the filtered reputation information in the second database. The reputation information included on the web page may be filtered using a third database that stores a reputation dictionary containing information about the emotions. Negative, and neutral tendencies of the web page for the target brand and the related brand based on the reputation information. In addition, the flat panel information and the channel information may be filtered at the same time, and the information including the time series information may be filtered. In addition, the reputation dictionary can be classified not only in affirmative, negative, and neutral but also according to the reputation information, and stores the classified information in a stepped manner.

In this manner, the filter unit can classify and classify web pages in various angles through keywords, time series information on keyword exposure, channel information on which keywords are exposed, and reputation information on keywords on a web page.

It is needless to say that the third database may include a reputation dictionary including keywords related to affirmative, negative, and neutral, and a keyword directly input by the user.

In addition, the filter unit can check and filter sentences including target brand and related brands or words designated by the user, in order to collect the amount of text existing on the online space but to collect accurate user's interest. In other words, the sophistication of the filtering can be further enhanced through the association keyword designated by the user. For example, if the user is interested in the entrance examination institute, the user may pre-input a keyword for 'entrance examination', 'luckling', and 'video course' into the third database as reference keywords, , It is possible to filter the web page by checking whether the web page includes the reference keywords 'entrance', 'luck', and 'video lecture'.

In addition, the filter unit may filter only the web pages retrieved and extracted during the predetermined period of the preset web page. Accordingly, the advertisement decision system can confirm the user's interests and the like in a more detailed manner during a specific period. In addition, the filter unit may be set as a keyword for a specific field to identify the user's interest, and if the keyword is included in the web page, the web page is filtered so that the low relevance web page is not included in future calculations So that the web page can be more efficiently analyzed. Then, the filter unit can classify the words included in the web page into essential keywords corresponding to essential parts or selective keywords having a somewhat lower importance than the essential keywords. For example, the word 'A school for entrance examination' can be selected as a required keyword in the collected web page, and the word 'B instructor' in the web page is used for the additional function of 'A school for entrance examination' , And can be specified by an optional keyword. In particular, the mandatory keyword includes at least one word representing the field information to be advertised. Of course, this mandatory and optional keywords are user definable. In addition, the required keyword and the selected keyword may be stored in the third database.

The filter unit can analyze the association between the required keyword and the selected keyword based on the required keyword selected above. For example, in the filter section, the phrase 'A teacher in A school of entrance examination, which is searched on a web page, easily teaches mathematics from the basic course' B lecturer 'can be calculated.

The filter unit counts the number of sentences including both the required keywords and the selected keywords in the sentence based on the number of existing sentences in the web page, and then counts the number of sentences including only the required keywords. The reliability of the essential keyword and the selected keyword can be calculated by dividing the number of sentences including both the essential keyword and the selected keyword by the number of sentences including only the essential keyword as in the following expression.

Reliability = the number of sentences containing both mandatory and optional keywords / the number of sentences containing only mandatory keywords

In addition, by inputting in advance a homepage or site activation degree of a user who is gaining online credibility, for example, a power blogger, a power twitter, an active online cafe operator, and an influential user on the Internet, And the reliability of the information registered in the channel and the general information.

 The larger the computed reliability result, the more likely it is that the essential keywords and the selected keywords are on-line issues. On the contrary, the lower the computation result, the lower the importance of the sentences. Analysis of the interest of the user through the analysis of the sentence in the web page can be performed. On the other hand, a more specific analysis can be performed by searching the correlation rule between the required keyword and the selected keyword included in the adjacent sentence in the web page. On the other hand, the user can input and store the required keyword and the selected keyword in advance in the third database in advance, and the filter unit can divide the required keyword and the selected keyword by referring to the third database. The required keyword may be a target brand, and the selected keyword may be a related keyword for the target brand. That is, through the reliability between the required keyword and the selected keyword, it is possible to decide whether to recommend the related keyword as the advertisement keyword.

In addition, the filter unit may filter the preference information about the input web page and the number of comments on the web page, and store the filtered preference information in the second database. For example, recently, a general web page includes buttons such as 'recommend' and 'non-recommendation' for each web page so that the reader of the web page can input the preference. That is, the filter unit may extract information on the number of 'recommended' and 'non-recommendation' for the web page and the number of comments on the web page, and store the information in the second database. As another example, a comment on a web page may include preference information such as "recommend" or "deprecated" and comments on the comment. Comments on these comments may be repeated. As another example, for multiple answers to a particular question on a web page, such as a knowledge search, the user can select the most preferred answer. Such preference information can be extracted and differentiated into weights in the calculation of reliability and the like for each web page. That is, when the number of non-observations of the target brand is significantly larger than the recommended number or vice versa, the comment may be filtered and managed, and the weight of the comment may be differentiated from other comments.

In addition, the filter unit may extract time information when a specific web page is exposed to the user terminal. When a specific web page is exposed to a user terminal for a predetermined period of time, it can be considered that the user is interested in a particular web page. Such information can be used to weight each web page. Meanwhile, the exposure time information may be provided by analyzing the log of the server providing the web page. For example, it is possible to estimate the time taken to request the next web page after requesting the first web page from the user terminal as the exposure time.

Next, the recommendation management unit 150 recommends an advertisement keyword that can most effectively advertise a target brand using data filtered by the filter unit.

More specifically, the recommendation management unit can recommend an advertisement keyword and a channel of a target brand according to the time series information, analyze a correlation between the keyword and the reputation information to recommend an advertisement keyword or provide marketing information, A web page spreader that spreads the source information web page of the target brand, and deletes a malicious web page that hinders the reputation of the target brand.

In addition, the advertisement decision system according to the present invention may further include a fourth database 151 including keywords that the recommendation management unit can refer to in recommendation.

First, the recommendation management unit can recommend an advertisement keyword to an advertiser by using time series information of keywords of a target brand which is searched and filtered through a search collection unit and a filter unit. For example, referring to FIG. 2, if the target brand for advertisement is' A school for entrance examination ', the related keywords for' A school for entrance examination 'include' mathematics', 'English', ' , 'Teacher B', and 'teacher C'. The keyword 'teacher B' among the keywords related to 'A school for entrance examination' was the lowest in June when the exposure on the webpage gradually decreased after the highest concentration in January of the year, while the keyword 'teacher C' , The recommendation management department recommends the keywords for 'teacher B' and related keywords in December and January. In the case that the number of exposure is the lowest in December and the following January, , And in May and June, we recommend that the keywords for 'teacher B' and their associated keywords be reduced. In addition, the recommendation management department recommends advertisements for 'teacher C' in May and June, while recommending advertisements in December and January. Through the recommendation management unit, it is possible to overcome the problem of selecting a keyword depending on limited information such as the number of monthly views and cost per click for a keyword provided by a general portal site. The exposure information on the web page of the keywords can be analyzed according to a predetermined cycle such as a month or a day. For example, there are a seven-month period for professional baseball and a five-month period for off-season, a four-year cycle for the Olympic or soccer World Cup, and a one-month open cycle for the TOEIC test. It is possible to determine the keyword of the advertisement by analyzing the collected keywords. It goes without saying that the period may be adjusted according to the user's convenience.

In addition, the recommendation management unit can recommend the advertisement keyword more precisely using the moving average line of the keyword exposure. The moving average is the number of times a keyword is exposed on a webpage for a certain period of time and then divided by the number of days in that period. The moving average line is a line connecting the values continuously. That is, referring to the exposure moving average line of the keywords for different periods, it is possible to examine changes in the keywords in the short, medium, and long term.

 The daily average of the number of times that the keyword of the target brand is exposed to a web page during a predetermined period (7 days, 30 days, 180 days, 365 days, etc.) is calculated, and the average number of times of exposure between the daily average number of times of exposure and the number of times of exposure It is possible to select an advertisement keyword by calculating the departure ratio. Referring to FIG. 3, for example, if the difference between the average 7-day exposure count and the current day exposure count is less than 10%, it is recommended to continuously recommend the advertisement keyword, and the difference between the 7-day average exposure count and the current day exposure count If the value of the keyword is more than 10%, it is determined that the keyword exposure on the web page is rapidly changing, so that it is possible to review whether or not the keyword is recommended for the advertisement keyword. That is, another parameter is analyzed to determine whether to recommend the advertisement keyword continuously or stop the advertisement keyword recommendation. In addition, the recommendation strength can be varied according to the average period compared. For example, since the 30-day average is the data collected over a longer period than the 7-day average, the difference between the 7-day average and the current daily average is more important than the 30-day average and the current daily average, And a higher recommended strength can be given to the 180-day average as compared with the 30-day average.

Referring to FIG. 4, it is determined that the number of times of exposure of the keyword is shortened when the moving average line whose current number of times of exposure is increasing is downward (Feb 28 of FIG. 4) (FIG. 4, Aug. 15), it is expected that the number of times of keyword exposure will increase in the near future, and the advertisement keyword can be recommended.

From the foregoing and other points of view, if the 7-day average moving line on the graph passes through the 30-day average moving line, it is judged that there is a change in the 7-day average, and the keyword recommending strength can be changed by sensing the short-

As a result, it is possible to reduce the overall advertising cost by helping the advertiser to decide the advertisement of a specific keyword.

In addition, the recommendation management unit can recommend the channel to the advertiser by using the time series information of the channels on which the keywords are exposed. For example, if your target brand for advertising is "A school for admission," "A school for admission" and related keywords can be general websites, news, blogs, microblogs, videos, knowledge search, It can be exposed to various channels such as sites and comments. Recommendation management recommends the most effective ad keyword exposure channel by using the temporal information about the exposure level of target brand. For example, if the keyword 'A school for admission' is exposed to a large amount of information on social network service sites in the past while it has been exposed to a lot of knowledge search in the past, it is recommended to advertise on a social network service site channel. It is also possible to extract a change pattern of the extent to which a predetermined keyword is exposed to the channel and to change the recommendation of the channel according to the specific pattern.

In addition, the recommendation management unit can recommend the advertisement keyword based on the time-series change of the reputation information of the web page. For example, if the target brand for advertising is 'A school', the related keywords for 'A school' may be 'short-term' and 'cool English'. In this case, the related keyword related to the target brand extracted at this time shows that the reputation information indicating the negative emotion such as 'expensive', 'burdened', and 'not good' On the other hand, it is assumed that the web pages exposed to 'cool English' have shown that the reputation information showing positive feelings such as 'cheap', 'cheap', 'okay' At this time, the recommendation management unit may recommend that 'English as a keyword' is used as an advertisement keyword instead of 'English term'. In addition, the recommendation management unit may recommend a word having positive reputation information as an advertisement keyword, thereby making it possible to create a web page so as to include the keyword having the positive reputation information and the keyword " short-term " can do.

The fourth database may include advertisement keywords directly designated by the user and / or advertisement keywords to be automatically recommended. Also, a user can generate a pair by mapping a relative keyword for each of the advertisement keywords. With reference to the above example, the user designates advertisement keywords as 'English', 'mathematics', 'teacher B', and 'teacher C' for the target brand 'A school' The relative keyword of 'English' is 'mathematics', the relative keyword of 'mathematics' is 'English', the relative keyword of 'B teacher' is 'C teacher', the relative keyword of 'C teacher' To generate a pair. In this case, if the reputation information of 'English', which is a related keyword of the target brand 'A school', is judged to be negative on the web page by referring to the fourth database, the recommendation management unit searches the mathematical keyword ' Is recommended. Likewise, if it is judged that 'B' s reputation information is negative, 'C teacher' is recommended as an advertisement keyword. This can reduce negative reputation for the target brand as a whole while reducing unnecessary advertising expenditure on ad keywords having a negative reputation as a whole. On the other hand, the fourth database may store the relative keyword for the advertisement keyword in advance, in cooperation with the reputation dictionary of the third database and the like, instead of manually specifying the relative keyword by the user.

In addition, the recommendation management unit can advertise the advertisement keyword again as described above, and confirm the change of the online reputation information again, and recommend the advertisement keyword again. That is, a recommendation of the advertisement keyword is performed again by feeding back the change result of the reputation information according to the recommended advertisement keyword.

In addition, the recommendation management unit can delete a negative or malicious web page to a target brand. More specifically, the filter unit filters a web page including negative or malicious reputation information by referring to a flat dictionary or the like, and the recommendation management unit can report that a negative or malicious web page has occurred to the user. The report can be done through various means such as email, SMS, and comments. It is also possible to distinguish the warning step of the report on the basis of the degree of denial or maliciousness of the reputation information classified step by step. For example, if the reputation dictionary is mapped to the first step of 'not good', the 'low quality' is mapped to the fifth step, and the reputation information of 'not good' It is possible to raise the alert level in the case where the reputation information " is low quality " is included. On the other hand, the recommendation management unit can report the extracted reputation information to the user according to the period set by the user.

In addition, the recommendation management unit can manage a web page spreader that spreads the source information web page. For example, when a link of a web page including information on a target brand is spread through a web page spreader such as a Twitter, which is a social network service site, and the web page is repeatedly searched and extracted through the search and collecting unit, The manager can manage the Twitter account that spreads the link of the web page as the core web page spreader. On the other hand, the web page spreader is not limited to a social network service site, and any channel capable of producing and distributing a web page for a target brand such as a cafe, a blog, and a news website may be applicable. In addition, the recommendation management section refers to various parameters such as the number of views of the web page spreader, the time when the web page is exposed on the user terminal, the number of comments posted on each web page, the number of comments recommended or unavailable, It can be designated as a core web page spreader and managed. More specifically, the recommendation management unit determines the importance of each web page spreader based on the number of other users and / or other web sites that each web page spreader has a relationship with online, such as a social network service, The number of uploaded posts, and the number of responses of other users and / or other web sites to the posts. Data related to social network services may include retweets, comments, mentions, etc. for the post. In this way, more sophisticated analysis results for web pages can be obtained by referring to user behavior patterns, connections, influences, inclinations, and the like in online such as social network services. For example, the recommendation management unit may recommend a keyword to include a topic or a preferred word preferred by a community such as a specific on-line cafe or a Facebook group to be an advertisement through the analysis. Particularly, since data such as posts, comments, tweets, retweets, mentions, comments, etc. collected through social network services can contain information about objects that the user directly is interested in, unlike general web pages, The related keywords extracted from the data collected through the social network service can be recommended as advertisement keywords by applying different weights to the related keywords extracted from other channels.

On the other hand, among the data collected through the social network service, the data generated by the retweeting may be repeatedly generated by the limited users. Therefore, unlike the data generated by other social network services, a low weight is given Also,

 Also, the recommendation management unit may calculate the reputation rating of the target brand using the retrieved flat information in relation to the target brand.

On the other hand, the recommendation management unit can recommend the advertisement keywords by analyzing the keywords existing in the advertiser's web site. For example, an advertiser who wants to advertise 'A school' can usually select 'A school' as an advertisement keyword. In this case, users who already know 'A school' can easily access 'A school' website through search engine. However, for a user who does not know 'A school', the keyword 'A school' may hinder the effectiveness of the advertisement. In order to solve such a problem, the search collecting unit extracts a core keyword in a web page of an advertiser's web site, and the recommendation management unit applies a weight to the extracted keyword based on the following formula, It is possible to recommend a keyword advertisement keyword with the rank of a keyword.

Key ad keyword = tf * idf + weight

tf: term frequency, the total number of occurrences of a particular term in the entire document set

df: document frequency, the number of documents with a specific keyword in the entire document set

idf: inverse document frequency, inverse of df

weight: a weight for a given keyword in consideration of other parameters

In this way, considering the number of views on the ad keyword, the sales amount of the advertisement keyword in the previous month, the click through rate or the ad operation score of the online advertisement, and the time series trend of the related keyword and reputation information, Keywords can be provided. And the reference advertisement keywords other than the core advertisement keywords may be provided. Reference ad keywords are offered to advertisers who are not competitive in terms of cost, so that they are selectively screened to increase return on investment.

In addition, the recommendation management unit can select a core advertisement keyword by using a parameter called a contrast exposure index in consideration of the popularity of each advertisement keyword, the ad unit price, and the like. More specifically, the advertisement can compute the contrast exposure index by dividing the number of times of exposure of keywords for each search engine by the advertisement price for the keyword. The advertisement may apply the weight according to the search engine to the contrast exposure index parameter and finally select the core advertisement keyword considering the relationship with other parameters such as time series information and channel information.

Hereinafter, an embodiment according to the present invention will be mainly described.

According to the embodiment of the present invention, the search and collecting unit collects web pages related to 'A school', which is a target brand desired to be advertised, and displays keywords and associated keywords' A school ',' Video lecture, and math video lecture. Then, through the filter unit, the time series information of each keyword 'A school', 'Hong Gil-dong teacher', 'English video lecture', and 'math video lecture' is filtered on each web page. The recommendation management unit then uses the information that the 'mathematical video lecture' increases in a time-wise manner while the number of exposures of the 'English video lecture' keyword obtained by the filter unit decreases in a time-wise manner. We recommend the keyword as an ad keyword. In this case, the recommendation management unit may determine whether or not the keywords of the past 'mathematical video lecture' are recommended using a graph pattern of the number of exposures.

According to another embodiment of the present invention, the filter unit also extracts reputation information of the target brand 'A school'. On the web page, 'reputation', 'good', 'not bad', etc. are searched for 'English video lecture' by referring to reputation dictionary etc. about 'A school' In the case of searching for reputable information such as 'I'm a star', 'Bad', or 'Very bad', the recommendation management department stops referring to 'Math video lecture' while strengthening recommendation for 'English video lecture'. Of course, in this case, as in the previous embodiment, it may be recommended in consideration of the time-series characteristics of the flat panel information. Also, if the reputation information is expected to be bad for the 'mathematical video lecture' in a time-wise manner considering the temporal information obtained in the past periodically, it may not be recommended as the keyword for the 'mathematical video lecture' . In this case, the filter unit can be applied to the keywords such as 'good', 'good', and 'not bad', which are information on positive reputation. That is, the reputation information of 'good' is more positive than the reputation information of 'good', and the reputation information of 'good' is more positive than the reputation information of 'not bad'. Therefore, for example, if 100 reputation information 'good' is extracted from 'English video lecture' on web page and 50 'bad' is extracted, 'excellent' about 'mathematical video lecture' And 100 'bad' are extracted, recommendation management recommends 'English video lecture' as an advertisement keyword with priority over 'mathematical video lecture'.

According to another embodiment of the present invention, while the amount of the keyword for 'A school' is collected in the knowledge search among various channels for web pages, the amount of collected in the online cafe is reduced in a time-wise manner If there is a recommendation, the recommendation management recommends knowledge search as an advertisement channel for 'A school'. As in the previous example, the recommendation management unit can refer to the data on the number of times of exposure on each channel for the 'A school', and recommend the advertisement keyword of 'A school' to what channel.

According to another embodiment of the present invention, the recommendation management section can recommend using the relative keyword previously set in the fourth database by the user. The user can preliminarily store 'Mr. Cheol-Soo Kim' as the relative keyword of 'Mr. Hong Gil-dong' in the fourth database in preparation for the deterioration of the reputation information of 'Mr. Of course, the relative keyword for 'Mr. Hong Gil-dong' is not limited to 'Mr. Chul-soo Kim'. If the reputation information of 'Mr. Hong Gil-dong' filtered by the filter unit is increasing in the amount of the reputation information which is negative in a time series, the recommendation section stops recommending the advertisement keyword to 'Mr. Hong Gil-dong' I recommend 'Kim Cheol-soo,' the relative keyword for 'Mr. Hong Gil-dong', as an ad keyword.

According to another embodiment of the present invention, the system according to the present invention has 1,000 and 600 exposure times in the search engines of 'English video lecture' and 'math video lecture', respectively, and the total advertisement price is 1,000,000 won And 400,000 won from the external search engine, the recommendation management unit calculates the contrast exposures of the advertisement at 1 and 1.5 (exposure count / KRW 1,000) using the exposure count and the total advertisement price. Through the above calculations, it is confirmed that the advertisement keyword related to the 'math video lecture' is more effective. The recommendation management unit can finally recommend the 'math video lecture' as an advertisement keyword considering the contrast exposure index as one element have.

FIG. 5 is a view illustrating a change of flat information according to another embodiment of the present invention. According to another embodiment of the present invention, the number of times of positive reputation information such as' good ',' good ',' not bad ', etc. for' Mr. Hong, The number increases from 100 to 150. In this case, the change in the number of exposures is 1.69, from 150 cases to 50 cases / 59 days. This is because it is included in the range of 2, which is the allowable variation amount set by the user, so that the degree of exposure of the advertisement keyword and the related keyword to 'Mr. On the other hand, if the number of exposures from May 1 to May 31 increases from 200 to 2,000, the change is 2,000 to -200 cases / 31 days, which is 58.06, which is significantly out of the allowable variation range of 2 . For this reason, the recommendation management unit can change the recommendation level and the degree of advertisement of the advertisement keyword 'Mr. Hong Kil-Dong'. In other words, if the web page collected by the search and collecting department is not generated voluntarily by the advertiser, but the positive reputation information on the web page is increased, it is possible to reduce the advertisement cost for 'Mr. Hong Gil Dong' Or, conversely, to reinforce market dominance, we can reinforce the advertisement for "Mr. Hong Gil-dong". The recommendation management unit can notify that there is a critical change in a specific keyword.

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined in the appended claims. It will be possible.

100: Advertising decision system
110: Search collection section
130:
150: recommendation management section
111: first database
131: second database
133: Third database
151: fourth database

Claims (12)

delete A search collecting unit for searching and collecting online data including a target brand by crawling or screen scraping data including a web page on-line;
A keyword of the target brand, an associated keyword associated with the keyword, a reputed keyword indicating a reputation for the target brand, and a time when online data including the reputable keyword is generated are extracted from the data collected and collected by the search and collecting unit A filter portion;
A recommendation management unit for recommending an advertisement keyword for the target brand in consideration of a change in the number of the reputed keywords extracted according to a time; And
And a database for storing the keyword extracted from the data, the associated keyword, the reputed keyword, and the timing, and a reputation dictionary referred to for extracting the reputed keyword from the data, A method of advertising decision-making via big data processing,
The filter unit assigns a reputation level to each reputed keyword in consideration of the flatness level of the extracted flat keyword,
The recommendation management unit further recommends an advertisement keyword for the target brand considering the reputation level of each reputable keyword,
Wherein the database stores a reputation dictionary further comprising reputation steps according to reputation keywords. A search collecting unit for searching and collecting online data including a target brand by crawling or screen scraping data including a web page on-line;
A keyword of the target brand, an associated keyword associated with the keyword, a reputed keyword indicating a reputation for the target brand, and a time when online data including the reputable keyword is generated are extracted from the data collected and collected by the search and collecting unit A filter portion;
A recommendation management unit for recommending an advertisement keyword for the target brand in consideration of a change in the number of the reputed keywords extracted according to a time; And
And a database for storing the keyword extracted from the data, the associated keyword, the reputed keyword, and the timing, and a reputation dictionary referred to for extracting the reputed keyword from the data, A method of advertising decision-making via big data processing,
Wherein the filter unit further extracts a kind of on-line channel on which the data is retrieved and collected from the data collected by the search and collecting unit,
The recommendation management unit further recommends the type of the channel to which the advertisement keyword is to be advertised, considering the change in the number of times the keyword and the related keyword are exposed on each channel,
Wherein the database further stores a kind of on-line channel on which the keyword is searched and collected. A search collecting unit for searching and collecting online data including a target brand by crawling or screen scraping data including a web page on-line;
A keyword of the target brand, an associated keyword associated with the keyword, a reputed keyword indicating a reputation for the target brand, and a time when online data including the reputable keyword is generated are extracted from the data collected and collected by the search and collecting unit A filter portion;
A recommendation management unit for recommending an advertisement keyword for the target brand in consideration of a change in the number of the reputed keywords extracted according to a time; And
And a database for storing the keyword extracted from the data, the associated keyword, the reputed keyword, and the timing, and a reputation dictionary referred to for extracting the reputed keyword from the data, A method of advertising decision-making via big data processing,
The recommendation management unit recommends a relative advertisement keyword of a recommended advertisement keyword by referring to the database when the number of negative reputable keywords increases for the target brand,
Wherein the database stores relative advertisement keywords for the recommended advertisement keywords. 5. The method according to any one of claims 2 to 4,
Wherein the recommendation management unit recommends the advertisement keyword considering the contrast exposure index of the advertisement divided by the advertisement cost on the online for a predetermined period of time for the advertisement keyword. 5. The method according to any one of claims 2 to 4,
Wherein the recommendation management unit is configured to execute the big data processing for changing the recommended advertisement keyword when the variation amount of the number of exposures of any one of the keyword, the related keyword, and the reputed keyword for a predetermined period exceeds a predetermined threshold range Advertising decision system. delete Crawling or screen-scraping a web page on an online channel including a news, a blog, a knowledge search, an online cafe, an online bulletin board, a bulletin of the online bulletin board, and a comment on the bulletin, A first step of storing;
A keyword of the target brand, an associated keyword associated with the keyword, a reputation keyword representing the reputation of the target brand referring to the reputation dictionary, and a type of online channel in which the data is searched and collected A second step of extracting and storing the data and the generation time of the data; And
A third step of recommending an advertisement keyword for the target brand considering at least one of a change amount of the keyword according to the extracted time series, a change amount of the related keyword, a change amount of the reputation keyword, and a change amount of the kind of the on- A method for advertising decision through big data processing,
The second step further extracts the user's preference information on the web page, the number of comments on the web page, the preference information on the comment, and the number of comments on the comment,
The third step refers to the user's preference information about the web page, the number of comments on the web page, the preference information on the comment, and the number of comments on the comment, A method of advertising decision making through big data processing that differentiates the weights of keywords, related keywords, and reputed keywords of a brand to recommend an advertisement keyword. Crawling or screen-scraping a web page on an online channel including a news, a blog, a knowledge search, an online cafe, an online bulletin board, a bulletin of the online bulletin board, and a comment on the bulletin, A first step of storing;
A keyword of the target brand, an associated keyword associated with the keyword, a reputation keyword representing the reputation of the target brand referring to the reputation dictionary, and a type of online channel in which the data is searched and collected A second step of extracting and storing the data and the generation time of the data; And
A third step of recommending an advertisement keyword for the target brand considering at least one of a change amount of the keyword according to the extracted time series, a change amount of the related keyword, a change amount of the reputation keyword, and a change amount of the kind of the on- A method for advertising decision through big data processing,
Wherein the second step is a step of calculating the number of sentences including both the keyword of the target brand and the related keyword or both of the keywords of the target brand and the reputed keyword in the data by the number of sentences including only the keyword of the target brand Further, a divided value is extracted as a parameter,
Wherein the third step is for recommending the advertisement keyword by differentiating the weights of the keyword, the related keyword and the reputed keyword of the target brand included in the sentence according to the value of the parameter. Crawling or screen-scraping a web page on an online channel including a news, a blog, a knowledge search, an online cafe, an online bulletin board, a bulletin of the online bulletin board, and a comment on the bulletin, A first step of storing;
A keyword of the target brand, an associated keyword associated with the keyword, a reputation keyword representing the reputation of the target brand referring to the reputation dictionary, and a type of online channel in which the data is searched and collected A second step of extracting and storing the data and the generation time of the data; And
A third step of recommending an advertisement keyword for the target brand considering at least one of a change amount of the keyword according to the extracted time series, a change amount of the related keyword, a change amount of the reputation keyword, and a change amount of the kind of the on- A method for advertising decision through big data processing,
The second step may include the number of web pages in the web page spreader including the link information linking to the web page including the information about the target brand, the number of views of the web page, the number of comments of the web page, The number of comments, the number of other users or websites affiliated with each other via the social network service, and the number of times that a web page having the same content as the web page started by the web page spreader is exposed on the internet Extraction,
The third step may include the number of web pages in the web page spreader, the number of views of a web page, the number of comments on a web page, the number of comments on a comment, the number of comments on a comment, The weight of the web page spreader is calculated in consideration of the number of times a web page having the same content as the web page disclosed by the web page spreader is exposed on the online, A method of advertising decision making via big data processing that provides notifications to advertisers. Crawling or screen-scraping a web page on an online channel including a news, a blog, a knowledge search, an online cafe, an online bulletin board, a bulletin of the online bulletin board, and a comment on the bulletin, A first step of storing;
A keyword of the target brand, an associated keyword associated with the keyword, a reputation keyword representing the reputation of the target brand referring to the reputation dictionary, and a type of online channel in which the data is searched and collected A second step of extracting and storing the data and the generation time of the data; And
A third step of recommending an advertisement keyword for the target brand considering at least one of a change amount of the keyword according to the extracted time series, a change amount of the related keyword, a change amount of the reputation keyword, and a change amount of the kind of the on- A method for advertising decision through big data processing,
The second step may include the number of web pages in the web page spreader including the link information linking to the web page including the information about the target brand, the number of views of the web page, the number of comments of the web page, The number of comments, the number of other users or websites affiliated with each other via the social network service, and the number of times that a web page having the same content as the web page started by the web page spreader is exposed on the internet Extraction,
The third step may include the number of web pages in the web page spreader, the number of views of a web page, the number of comments on a web page, the number of comments on a comment, the number of comments on a comment, The number of times the webpage having the same content as that of the webpage disclosed by the webpage spreader is exposed on-line, the keyword of the target brand included in the webpage included in the webpage spreader, And a big data process for recommending an advertisement keyword by differentiating weights of the reputable keywords. The method according to any one of claims 8 to 11,
And providing a notification to a user when a reputable keyword including negative reputation information among the reputable keywords increases.

Images