KR102593996B1 - Apparatus and Method For Providing Recommendation Service Through Social Media Activity Analysis - Google Patents

Abstract

An apparatus for providing a recommendation service through social media activity analysis according to the present invention includes at least one memory; at least one database unit; and at least one processor; The at least one processor may include: (a) receiving and registering a first keyword of interest; (b) a process of crawling related social media content from a plurality of social media platforms using the first keyword of interest; (c) Analyzing the crawled social media content to extract user account information of at least one valid content highly related to the first keyword of interest, and corresponding the extracted user account information to the first keyword of interest A process of storing valid user account information in the database unit; (d) a process of crawling the active user's social media activity content based on the active user account information stored in the database unit; (e) extracting the crawled content into standardized text through morphological analysis, classifying the text into categories through machine learning, assigning category values, and storing them in the database; (f) Time-dependent action path information and movement of the effective user based on at least one of creation time information of the content, context before and after creation of the content, time information included in the content, and positional relationship of the text. A process of generating at least one piece of route information and storing it in the database unit; and (g) when a first keyword of interest is input from the user, the database unit extracts the action path information and the movement path information related to the first keyword of interest, and executes a recommendation algorithm to provide the location, action, and behavior to the user. Processes the process of recommending at least one of , travel course, and product.
As a result, the following actions, places, products, etc. suitable for social trends can be recommended to users who have entered a specific keyword of interest by executing an internal recommendation algorithm based on related action path information or movement path information.

Description

Apparatus and Method For Providing Recommendation Service Through Social Media Activity Analysis}

The present invention relates to an apparatus and method for providing recommendation services through social media activity analysis. More specifically, the present invention relates to a service that recommends places, actions, travel courses, and products for keywords of interest to users through social media activity analysis. It's about.

Recent developments in artificial intelligence and data analysis technology are having a significant impact on the online sales market and social media market. For example, personalized recommendation services that take into account the preferences of individuals consuming specific products or news are rapidly spreading.

These services form collaborations based on individual consumption methods, gender, age, etc. and recommend products or news similar to those consumed by users in a specific group.

However, since this personal recommendation service matches similar groups based on demographic analysis such as gender or age and is limited to simple areas of interest, it is difficult to provide services based on changes in interest patterns or interest paths of users over time. There are limits.

In order to solve the problems described above, in the present invention, effective users are extracted through analysis of social media activity content for each keyword of interest, and action path information or movement path information is extracted and stored by analyzing the social media activity content of the effective user. and analyzes social media activities to recommend the following places, actions, travel courses, and products based on action path information or movement path information related to the keyword of interest by executing an internal recommendation algorithm to the user who entered a specific keyword of interest. The purpose is to provide a device and method for providing recommendation services through .

The above object is to provide a recommendation service providing device through social media activity analysis, comprising: at least one memory; at least one database unit; and at least one processor; The at least one processor may include: (a) receiving and registering a first keyword of interest; (b) a process of crawling related social media content from a plurality of social media platforms using the first keyword of interest; (c) Analyzing the crawled social media content to extract user account information of at least one valid content highly related to the first keyword of interest, and corresponding the extracted user account information to the first keyword of interest A process of storing valid user account information in the database unit; (d) a process of crawling the active user's social media activity content based on the active user account information stored in the database unit; (e) extracting the crawled content into standardized text through morphological analysis, classifying the text into categories through machine learning, assigning category values, and storing them in the database; (f) Time-dependent action path information and movement of the effective user based on at least one of creation time information of the content, context before and after creation of the content, time information included in the content, and positional relationship of the text. A process of generating at least one piece of route information and storing it in the database unit; and (g) when a first keyword of interest is input from the user, the database unit extracts the action path information and the movement path information related to the first keyword of interest, and executes a recommendation algorithm to provide the location, action, and behavior to the user. This is achieved by a recommendation service providing device through social media activity analysis, characterized in that it processes a process of recommending at least one of, a moving course, and a product.

Here, the category may include at least one of location/place, food/product, brand, culture/leisure, nature/environment, person, issue, and emotion.

In addition, in step (e), the processor may further process extracting location information indicating a location where the user created the content or an artificially designated location from the crawled content.

In addition, in the process (f), the processor determines the chronological order based on the creation time information of the content, the before-and-after relationship of creation of the content, the time information included in the content, and the positional relationship of the text At least one of the action path information and the movement path information can be generated by assigning an order value and arranging the text to which the category value is assigned according to the time-series order value.

Furthermore, in the process (f), the processor includes creation time information of the content, a context before and after creation of the content, time information included in the content, location information included in the content, positional relationship of the text, and extracting the user's action or location based on the category value of the text, grouping the text in chronological order according to the action or location, and assigning an order value to at least one of the action path information and the movement path information. can be created.

Additionally, the processor may register location information included in the content and text to which the category value is assigned as a keyword.

Additionally, in step (e), the processor may further process a process of extracting text indicating satisfaction from the standardized text extracted from the crawled content and assigning weight to the corresponding text.

Here, the recommendation algorithm assigns a score according to the number of users or content with actions, places, and paths included in the extracted action path information and the movement path information, and selects a plurality of action paths based on the given score. It may include a process of calculating a score corresponding to the information and movement route information and recommending at least one of a place, action, movement course, and product to the user according to the calculated score.

Meanwhile, the above object is a method of providing a recommendation service through social media activity analysis, comprising: (a) receiving and registering a first keyword of interest; (b) crawling relevant social media content from a plurality of social media platforms using the first keyword of interest; (c) Analyzing the crawled social media content to extract user account information of at least one valid content highly related to the first keyword of interest, and corresponding the extracted user account information to the first keyword of interest Storing valid user account information in a database unit; (d) crawling the active user's social media activity content based on the active user account information stored in the database unit; (e) Extracting creation time information and location information of the content from the crawled content, extracting standardized text through morphological analysis, classifying the text into categories through machine learning, and assigning category values to the database unit Save to; (f) Time-dependent action path information and movement of the effective user based on at least one of creation time information of the content, context before and after creation of the content, time information included in the content, and positional relationship of the text. Generating at least one piece of path information and storing it in the database unit; and (g) when a first keyword of interest is input from the user, the database unit extracts the action path information and the movement path information related to the first keyword of interest, and executes a recommendation algorithm to provide the location, action, and behavior to the user. This can also be achieved by a method of providing a recommendation service through social media activity analysis, which includes the step of recommending at least one of , a moving course, and a product.

The step (f) determines the chronological order based on the creation time information of the content, the before-and-after relationship of creation of the content, the time information included in the content, and the positional relationship of the text, and assigns an order value; , At least one of the action path information and the movement path information can be generated by arranging the text to which the category value is assigned according to the time-series order value.

As described above, the apparatus and method for providing recommendation services through social media activity analysis according to the present invention executes an internal recommendation algorithm based on related action path information or movement path information to users who enter specific keywords of interest to provide social trends and information. The following places, activities, travel courses, and products suitable for your interests can be recommended.

1 is a schematic diagram illustrating an environment for providing recommendation services through social media activity analysis according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating an example of a process for providing a recommended service through analysis of social media activities in FIG. 1.
Figure 3 shows the detailed configuration of the data analysis unit and the database unit to explain the specific operation of the service server according to an embodiment of the present invention.
Figure 4 illustrates data before and after classification of advertising text by a user filter unit during data analysis in a service server according to an embodiment of the present invention.
Figure 5 illustrates data before and after being classified by an inference module during data analysis in a service server according to an embodiment of the present invention.
Figure 6 shows an example of social media content to explain a method of extracting action path information and/or movement path information by the path information extraction unit according to an embodiment of the present invention.
Figure 7 is a schematic diagram illustrating a recommendation algorithm according to an embodiment of the present invention.
Figures 8 and 9 show an example of recommendation information provided to a user terminal by a recommendation unit according to an embodiment of the present invention.
Figure 10 is a flowchart showing a method of generating action path information and/or movement path information in a service server according to an embodiment of the present invention.
Figure 11 is a flowchart explaining a method of generating and providing recommended route information by a service server according to an embodiment of the present invention.

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

1 is a schematic diagram illustrating an environment for providing recommendation services through social media activity analysis according to an embodiment of the present invention. The user terminal 10, SNS device 20, and service server 100 are connected by a network 30.

The user terminal 10 may include a mobile terminal such as a smartphone or a tablet PC, and a user terminal 10 such as a PC, and connects to the service server 100 through the network 30 to perform a service according to the present invention. Services can be provided. The user terminal 10 includes a communication module for transmitting and receiving data through a network such as the Internet.

The SNS device 20 refers to a server that provides social media platform services and may include, for example, devices that provide various SNS platform services such as Instagram, Facebook, Naver, Daum's blog, and cafe. Social media refers to online tools and platforms used to share opinions, thoughts, experiences, and perspectives. Content shared through social media takes various forms such as text, images, audio, and video, and includes blogs, social media, etc. Includes networks, instant message boards, podcasts, wikis, user-generated content (UCC), and more. In this specification, SNS (Social Network Service) and social media are used with the same meaning.

The service server 100 is a device for providing recommended services through social media activity analysis according to an embodiment of the invention, and includes a memory 110 that stores data and computer program codes, and a processor 120 that executes computer program codes. ), a database unit 130 in which a database is built, and a network interface 140 for network connection to the outside.

The database unit 130 includes a dictionary DB 131 used to analyze SNS activities and a relational database management system (RDBMS) 133 in which user information and SNS activity analysis information are stored. Here, the dictionary DB 131 and the RDBMS 133 may each be provided in a separate memory device or a separate DB server device. The network interface 130 includes an Internet communication module, etc.

When a keyword of interest is registered, the service server 100 connects to at least one SNS device 20, collects and analyzes related social media content, and extracts user account information of valid content highly related to the registered keyword of interest. It is stored as the information of the effective user, and the SNS activity content of the effective user is collected and analyzed using the information of the effective user to extract the user's action path information or movement path information and store it in the RDBMS (133). Here, keywords of interest are continuously accumulated, and user action path information or movement path information regarding the keywords of interest may also be updated periodically. Through this process, the RDBMS 133 builds a database containing the action path information and movement path information of a large number of users regarding a plurality of keywords of interest.

When a specific keyword of interest is input from the user terminal 10, the service server 100 extracts the action path information and movement path information of effective users related to the keyword of interest from the RDBMS 133, and uses an internal recommendation algorithm to At least one recommendation among places, actions, travel courses, and products is provided to the user terminal 10. Users can check the recommended information and decide on their next action. The service server 100 may include a web server that stores component files of a website for receiving service applications from users and providing recommended information services.

In FIG. 1, the data collection unit 121, data analysis unit 123, and recommendation unit 125 are implemented as a series of computer program codes that perform their respective functions, are stored in the memory 110, and are operated by the processor 120. It is executed through In Figure 1, for convenience of explanation, it is expressed as a configuration included in the processor 120.

The data collection unit 121 is used to collect related social media content by accessing the SNS device 20, and includes a crawler that crawls data.

The data analysis unit 123 extracts effective users by analyzing the collected data, and analyzes the contents of the effective users collected by the data collection unit 121 to provide user action path information and/or movement path information for keywords of interest. It is intended to extract and may include machine learning or deep learning inference engines for data extraction, classification, and prediction.

When a keyword of interest is input and searched from the user terminal 10, the recommendation unit 125 executes an internal recommendation algorithm based on the action path information and/or movement path information of active users stored in the RDBMS 133 to select the keyword of interest. It is intended to generate recommendation information for at least one of the user's next place, action, movement course, and product and provide it to the user terminal 10. Recommendation information such as place, action, movement course, and product is provided on a map, It may include a GUI creation program to display statistics, pictures, various techniques, etc.

FIG. 2 is a flowchart illustrating an example of a process for providing a recommended service through analysis of social media activities in FIG. 1.

The user can access the website served by the service server 100 through the user terminal 10 and use the recommendation service according to the present invention (S10). The website of the service server 100 according to the present invention may provide product sales services in addition to recommendation services. For example, a service that recommends products or places according to the present invention may be provided to members who subscribe to a website that provides product sales services.

When a keyword of interest is registered (S10), the service server 100 collects related social media content from a plurality of social media platforms using the keyword of interest (S11). The server administrator can register keywords of interest collectively or as needed, or search keywords entered by members can be automatically registered as keywords of interest.

The service server 100 analyzes the collected social media content and extracts the account information of the user whose content is highly related to the keyword of interest as the effective user information (S12). The extracted effective user information is stored in the database unit 130. By repeating this process, information on valid users corresponding to multiple keywords of interest is accumulated.

The service server 100 uses the user account information of the active user to collect social media content created by the user (S13). At this time, the collection target may be limited to content created over a certain period of time, in order to reflect the latest trends or trends.

The service server 100 analyzes the collected content, extracts the user's action path information and/or movement path information, and stores it in the database unit 130 (S14). While the conventional recommendation service generally extracts and analyzes only content highly related to the keyword of interest, the present invention collects content before and after content highly related to the keyword of interest based on the user, thereby determining the user's behavior regarding the keyword of interest. Path (pattern) information and/or even movement path information can be extracted. The above-described processes S10 to S14, S11 to S14, or S13 to S14 may be performed periodically or repeatedly as needed.

When the user accesses the service server 100 using the user terminal 10 and inputs a first keyword of interest, the service server 100 provides action path information and/or movement path regarding the first keyword of interest entered by the user. Information is extracted from the database unit 130 (S16), and an internal recommendation algorithm is executed to generate recommendation information that recommends at least one of place, action, travel course, and product to the user (S17). The service server 100 provides the generated recommendation information to the user terminal 10 (S18), and the user terminal 10 displays the recommended information (S19). As described above, recommended information may be provided in various formats that users can easily and intuitively understand. At this time, if the first keyword of interest entered by the user does not exist in the RDBMS 133, steps S10 to S18 are executed in real time to provide recommendation results.

Figure 3 shows the detailed configuration of the data analysis unit 123 and the database unit 130 to explain the specific operation of the service server 100 according to an embodiment of the present invention.

The dictionary DB 131 includes a morpheme dictionary DB 131a, a similar word dictionary DB 131b, a filter dictionary DB 131c, and an inference dictionary DB 131d. The morpheme dictionary DB (131a) is constructed by constructing text morpheme data as a DB, and the morpheme analysis module (123a) uses this to perform morpheme analysis. The similar word dictionary DB 131b refers to a dictionary DB about similar words with the same meaning, and the filter dictionary DB 113c refers to a DB about advertising morphemes. The inference dictionary DB 131d is for classifying processed morphemes by category field, and the category to which each morpheme belongs is matched. Here, the category field may be defined differently depending on the service content, and the number of categories may be set in various ways depending on the service. For example, the categories defined in the inference dictionary DB 131d according to the present invention include at least one of location/place, food/product, brand, culture/leisure, nature/environment, person, emotion/satisfaction, and issue. can do.

The data analysis unit 123 includes a morpheme analysis module 123a, a similar word matching module 123b, a user filter unit 123c, an inference module 123d, and a path information extraction unit 123e. The modules included in the data analysis unit 123 are software for data analysis divided by function, and may be prepared as separate programs or integrated. The data analysis unit may extract location information indicating the location where the user created the content or an artificially designated location from the crawled content and store it in the database unit 130 as location information for the content. Location information can be included in content as text, tags, latitude and longitude information, etc. indicating a region or place, and the data analysis unit can extract the location information included in the text or tags by converting it into latitude and longitude information. The data analysis unit may include an internal program that converts place names or area names into GPS information, that is, latitude and longitude information, or may be utilized as an external program. In this way, if the content contains text or location information indicating a place or location, the data analysis unit converts it and stores it in the RDBMS as location information for the content, and this location information is used when recommended content is displayed based on a map. It can be utilized. At this time, the location information may be hierarchically organized according to the details of the location information included in the content, for example, units in administrative districts such as city/county/gu/dong/street address.

The morpheme analysis module 123a is for extracting crawled content into standardized text through morpheme analysis, and can be implemented as a preprocessing module that tokenizes text by morpheme. The morpheme analysis module 123a separates text data constituting a sentence into morpheme units based on the morpheme dictionary DB 131a. The morpheme analysis module 123a can update the morpheme dictionary DB 131a for words that are not in the morpheme dictionary DB 131a.

The similar word matching module 123b is for processing different words with the same meaning into the same word, and processes morphemes with similar meaning based on the similar word dictionary DB 131b. For example, ‘Hana C.C., Hana CC, Hana Golf Course’ matches with ‘Hana Country Club.’ Meanwhile, the similar word matching module 123b may include a machine learning model, and for morphemes that are not in the similar word dictionary DB 131b, similar words are classified through the machine learning model and registered in the similar word dictionary DB 131b.

The user filter unit 123c is for filtering valid users and classifies advertising morphemes into separate groups based on the filter dictionary DB 113c. The user filter unit 123c classifies morpheme texts indicating advertising, spam, and illegality among the extracted morphemes into separate groups, and determines whether it is valid content by valid users based on the frequency of the texts classified into separate groups. After determining whether or not advertising content is valid, it is excluded from the analysis. Figure 4 illustrates data before and after classification of advertising text by the user filter unit 123c during data analysis in the service server 100 according to an embodiment of the present invention. Referring to Figure 4, it can be seen that sales, drive, test drive, quote, used car, dealer, etc. are classified as text indicating advertisement, spam, and illegal, respectively.

The user filter unit 123c extracts the user information of the content determined to be advertising content and registers it in the filter dictionary DB 113c to exclude the user from the valid users in subsequent content analysis. The user filter unit 123c and the filter dictionary DB 113c are used to exclude content for commercial purposes, such as advertising content, from the analysis by excluding it from valid content, and to exclude users of content for commercial purposes from valid users.

The inference module 123d is for classifying the morphemes of each word extracted from the content by category field, and matches and classifies the morphemes into corresponding categories based on the inference dictionary DB 131d. The inference module 123d may include an inference model through machine learning, and morphemes that are not in the inference dictionary DB 131d are classified into categories through machine learning. Here, the category field refers to a group name that represents the properties of each morpheme word. Categories, that is, properties of words, can be set in various ways in the dictionary according to the user's definition, and their number can also be set in various ways. It can be. For example, if the properties of a morpheme word are defined as location/place, food/product, brand, culture/leisure, nature/environment, person/family, issue, emotion/satisfaction, etc., the categories are location/place, food/satisfaction, etc. It can include products, brands, culture/leisure, nature/environment, people/family, issues, emotions/satisfaction, etc. Figure 5 illustrates data before and after being classified by the inference module 123d in the process of data analysis in the service server 100 according to an embodiment of the present invention, where (a) is the morpheme analysis module 123a ) and data extracted through the similar word matching module 123b, and (b) illustrates the category matched by the inference module 123d.

The RDBMS 133 maintains a database of data such as valid user account information corresponding to a plurality of keywords of interest, valid user content, content creation time, content location information, and text classified into categories and assigned category values. it means. The information extracted and analyzed by the data collection unit 121 and the data analysis unit 123 is stored in the RDBMS 133. Texts assigned location information and category values included in content stored in the RDMS can also be registered as keywords of interest. For example, the results of analysis of active users and their social media content regarding the first keyword of interest are stored in the RDBMS. This is not simply used for the first keyword of interest, but also other keywords of interest (text) extracted during the analysis process. It can also be used for .

The path information extraction unit 123e extracts the time of the effective user based on at least one of the creation time information of the effective user's content in the RDMS, the before-and-after relationship of the creation of the content, the time information included in the content, and the before-and-after position relationship of the text. Generate at least one of action path information and movement path information according to. For example, the path information extraction unit 123e determines the time-series order based on the creation time information for each content, the before-and-after relationship of creation of the content, the time information included in the content, and the forward-and-forward positional relationship between the extracted texts. At least one of action path information and the movement path information can be generated by assigning an order value between content and text, and arranging texts to which category values are assigned according to a time-series order value. As another example, the path information extraction unit 123e may include information on the creation time of the content, the before-and-after relationship of the creation of the content, the time information included in the content, the location information included in the content, the before-and-after positional relationship of the text in the content, and the positional relationship of the text in the content. At least one of action path information and movement path information can be generated by extracting the user's action or location based on the category value, grouping the text in a time-serial order according to the extracted action or location, and assigning an order value.

Figure 6 shows an example of social media content to explain the method of extracting action path information and/or movement path information by the path information extraction unit 123e according to an embodiment of the present invention. For example, if the keyword of interest is 'Gangneung travel', let's assume that a user with the ID Gildong was extracted as an effective user, and Figure 6 shows the scrolled data of the effective user. In Figure 6, for convenience of explanation, the colors are different for each category so that each text category can be distinguished. The green square represents the place/location category, the red square represents emotion/satisfaction, the yellow square represents food, and the black square represents text belonging to other categories.

In Figure 6, there are two posts, and the temporal relationship between the two can be determined by the writing order. Additionally, if the text given a category value representing the place/location is listed in the order of the text location, it becomes 'Gangneung - Gapyeong Rest Area - Gangneung - Gyeongpodae Beach - Jumunjin Unlimited Red Crab Refill - Delicious Coffee Shop'. Each place/position is assigned an order value. At this time, Gangneung is excluded because it is a higher level location information than other texts. Ultimately, the travel route is extracted as 'Gapyeong Rest Area - Gyeongpodae Beach - Jumunjin Red Crab Unlimited Refill - Delicious Coffee Shop'. At this time, text matching each place can be grouped. For example, it can be grouped into four groups: 'Gapyeong Rest Area - Broiler Farm, Gyeongpodae Beach - Wind, Jumunjin - Unlimited Red Crab Refill, Delicious Coffee Shop - Coffee - View'. At this time, grouping can be determined by considering the order of text in a sentence, punctuation marks such as commas and periods, and the distance between texts. This is not only a simple time-series order of places, but also the information included in the action path information and movement path information can be more detailed by grouping together texts with categories other than places. Additionally, if there are words expressing emotions or evaluations, weight can be given to text that matches (groups) them, for example, to the corresponding place or action. For example, in Figure 6, there are words expressing emotions or satisfaction, such as 'it was delicious', 'it was cool', 'it was really delicious', and 'it was so good'. In this case, weight can be given to 'Gyeongpodae Beach', 'Jumunjin Red Crab Unlimited Refill', and 'Delicious Coffee Shop' grouped with these words, respectively. Through the above process, the action route information (movement route information) extracted by the route information extraction unit 123e for the keyword of interest 'Gangneung travel' is 'Gapyeong rest area - Gyeongpo Beach - Jumunjin unlimited red crab refill - delicious coffee shop'. is stored in the RDBMS (133). Meanwhile, the texts of 'Gapyeong Rest Area, Gyeongpodae Beach, Jumunjin Red Crab Unlimited Refill, Delicious Coffee Shop' included in the above travel route information are not only used as keywords of interest for 'Gangneung Travel', but each of them is used as a separate keyword of interest. It can be registered and used for other keywords of interest.

When a keyword of interest is input and searched from the user terminal 10, the recommendation unit 125 extracts a plurality of action path information and movement path information corresponding to the relevant keyword of interest stored in the RDBMS 133 and executes an internal recommendation algorithm. Then, recommended information for at least one of the following locations, actions, travel courses, and products related to the keyword of interest is generated. As described above, the RDBMS 133 stores effective user information for each keyword of interest and a large number of action path information and movement path information corresponding thereto, so the recommendation unit 125 utilizes the data stored in the RDBMS 133. .

The recommendation algorithm assigns points to, for example, the number of users or content and the distance to the action path information of the effective user extracted from the RDBMS 133 and the actions, places, and paths included in the movement path information. Based on the score, a score corresponding to the number of cases of a plurality of action path information and movement path information is calculated, and at least one of a place, action, movement course, and product is recommended to the user according to the calculated score.

Figure 7 is a schematic diagram illustrating a recommendation algorithm according to an embodiment of the present invention. Let's assume that a keyword of interest called 'Gangneung travel' is input from the user terminal 10, and that there are three types of related action path information and movement path information within the standard area extracted from the recommendation unit 125 during the standard period as follows.

ㆍUser A, route information 'ㄱ -> ㄴ -> ㄷ -> ㄹ -> ㅁ -> ㅂ',

ㆍB user, route information 'ㄴ -> ㄹ -> ㅅ -> ㅁ -> ㅂ'

ㆍC user, route information ‘ㄱ -> ㄷ -> ㄴ -> ㅇ’

Here, ㄱ, ㄴ, ㄷ, ㄹ, ㅁ, ㅂ, ㅅ, ㅇ, etc. indicate place/position.

At this time, the starting location (reference location) can be determined by considering the current location of the user, location weight according to satisfaction, location with the most visits, geographical location, etc. In Figure 7, it is assumed that the reference position (starting position) is determined as 'ㄷ' in consideration of the current user's location.

In the recommendation algorithm, the value of each route starts from 0.9 according to the visit order of users who visited the place and decreases by 0.1 as the distance increases, so routes from the moment it reaches 0 are not considered, and different values are assigned to the same route. In this case, the higher value of the two is taken as the value.

The value of each place (location) is determined by the number of users who visited that place and has a value of 1.1 to 2.0. For example, if all 10 users visited location ‘A’, the value of location ‘A’ is 2.0, and if only 1 user out of 10 visited location ‘A’, the value of location A is 1.1. If points are given to each place (location) and route according to these criteria, points are given as shown in FIG. 7.

The recommendation algorithm reflects the scores given to the place and route to the recommended place/course and calculates the recommended route score by multiplying the place value by the route value for each place/course, and assigns the value to the route/place based on 'ㄷ'. Calculating is as follows.

ㆍㄷ→ㄱ, 0.9 * 1.1 = 0.99

ㆍㄷ→ㄴ→ㄱ, 0.9 * 1.2 * 0.8 * 1.1 = 0.9504

ㆍㄷ→ㄴ, 0.9 * 1.2 = 1.08

ㆍㄷ→ㄱ→ㄴ, 0.9 * 1.1 * 0.8 * 1.2 = 0.9504

ㆍㄷ→ㄹ, 0.9 * 1.1 = 0.99

ㆍㄷ→ㄹ→ㅁ, 0.9 * 1.1 * 0.8 * 1.1 = 0.8712

ㆍㄷ→ㄹ→ㅁ→ㅂ, 0.9 * 1.1 * 0.8 * 1.1 * 0.7 * 1.1 = 0.670824

ㆍㄷ→ㄴ→ㅇ, 0.9 * 1.2 * 0.8 * 1.0 = 0.8640

ㆍㄷ→ㄱ→ㄴ→ㅇ, 0.9 * 1.1 * 0.8 * 1.2 * 0.8 * 1.0 = 0.760320

Based on the above calculation results, recommendations are possible as follows depending on the depth (stage) of the path.

ㆍLocation recommendation: ㄴ

ㆍRecommended route information (2depth): ㄷ→ㄴ→ㄱ

ㆍRecommended route information (3depth): ㄷ→ㄱ→ㄴ→ㅇ

Figures 8 and 9 show an example of recommendation information provided to the user terminal 10 by the recommendation unit 125 according to an embodiment of the present invention. Referring to (a) of Figure 8, when the user inputs 'Busan' as a keyword of interest, area A is provided with a window for entering the keyword of interest (search keyword), and area B is a sub-interest of the entered keyword of interest. As an area where recommended keywords are displayed, in (a) of Figure 8, you can see that four sub-keywords of interest in Busan, 'coffee, pork, bakery, beef' are displayed. Area C displays keywords of interest on the map, and the locations where each keyword is mentioned are displayed in different colors for each keyword on the map, allowing you to intuitively know the distribution of keywords of interest by region. Area B contains icons for selecting and excluding recommended sub-keywords, which are selected or deselected when the user clicks on them. In this case, only the selected keywords have their distribution displayed in area C below.

Figure 8 (b) is a case where the user enters 'coffee' in the keyword input box in area A. Since there is no sub-keyword of interest for coffee, recommended information is not displayed in area B, and locations related to 'coffee' in the country are not displayed. is displayed on the map in area C. Looking at this, you can see at a glance which regions have the most interest in the keyword.

Figure 9 shows a screen when a user clicks on Gangneung on the map of area C in (b) of Figure 8 after entering 'coffee' in the keyword input window of area A. In this case, in area C, recommended route information related to coffee in Gangneung is displayed on the map, and in area D, recommended route information is displayed in text. Referring to Figure 9, each place name is connected by an arrow to indicate the overall recommended course, and a selection button is provided to view recommended post information corresponding to each place, so that when the user clicks the recommended post '>' button, the corresponding You can check the details of the user's post by linking to the post content.

Figure 10 is a flowchart showing a method of generating action path information and/or movement path information in the service server 100 according to an embodiment of the present invention. Referring to Figure 10, related social media content is collected from a plurality of social media platforms using registered keywords of interest, and effective users are extracted (S20). Specifically, crawl relevant social media content from multiple social media platforms using registered keywords of interest, analyze the crawled social media content, and obtain user account information of at least one valid content highly related to the keyword of interest. is extracted, and the extracted user's account information is stored in the RDBMS (133) as valid user account information corresponding to the keyword of interest. At this time, multiple valid user account information is registered for one keyword of interest.

Once account information about the active user is extracted, the active user's account information is used to access the SNS platform to crawl the user's social media activity content (S21), and the crawled content is extracted into standardized text through morphological analysis. Do it (S22). At this time, in addition to text classification, location information and creation time information included in the content are extracted and stored in the RDBMS (133). The service server 100 performs similar word matching on the standardized text extracted through the similar word matching module (S23). The service server 100 classifies text indicating advertisements, spam, and illegality among the standardized texts into separate groups through the user filter unit 123c, and if the text corresponding to the advertisement has a certain frequency or more, the user is classified as an effective user. Excluded from and excluded from valid content (S25, s26). The service server 100 classifies similar word-matched texts into predefined categories through the inference module 123d, assigns a category value to the classified text, and stores it in the RDBMS 133 (S27). Here, similar word matching and classification by category can be performed through a machine learning model.

Figure 11 is a flowchart explaining a method of generating and providing recommended route information by the service server 100 according to an embodiment of the present invention. Referring to FIG. 11, when a keyword of interest is input from the user terminal 10 and received by the service server 100 (S30), the service server 100 stores action path information corresponding to the keyword of interest stored in the RDBMS 133. and extract movement route information (S31). Then, the recommendation algorithm is run to assign a score to each place, action, and the path between each place or action included in the route information, extract the number of cases of various recommended candidate routes, and provide information about the place or action in each recommended candidate route. The score is multiplied by the route score to calculate the score for each recommended candidate route (S32). Then, the place, action, travel course, and product with the highest score are selected as the final recommended route, provided to the user terminal 10, and displayed on the map (S33).

In the above, even though all the components constituting the embodiment of the present invention have been described as being combined or operated in combination, the present invention is not necessarily limited to this embodiment. That is, as long as it is within the scope of the purpose of the present invention, all of the components may be operated by selectively combining one or more of them. In addition, although all of the components may be implemented as a single independent hardware, a program module in which some or all of the components are selectively combined to perform some or all of the functions of one or more pieces of hardware. It may also be implemented as a computer program having. The codes and code segments that make up the computer program can be easily deduced by a person skilled in the art of the present invention. Such a computer program can be stored in a computer-readable storage medium and read and executed by a computer, thereby implementing embodiments of the present invention. Storage media for computer programs may include magnetic recording media, optical recording media, etc.

In addition, terms such as “include,” “comprise,” or “have” described above mean that the corresponding component may be present, unless specifically stated to the contrary, and thus do not exclude other components. Rather, it should be interpreted as being able to include other components. All terms, including technical or scientific terms, unless otherwise defined, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention pertains. Commonly used terms, such as terms defined in a dictionary, should be interpreted as consistent with the contextual meaning of the related technology, and should not be interpreted in an idealized or overly formal sense unless explicitly defined in the present invention.

The above description is merely an illustrative explanation of the technical idea of the present invention, and various modifications and variations will be possible to those skilled in the art without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but are for illustrative purposes, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be construed as being included in the scope of rights of the present invention.

10: user terminal 20: SNS device
100: service server 110: memory
120: Processor 121: Data collection unit
123: Data analysis unit 125: Recommendation unit
130: Database unit 131: Dictionary DB
133: RDBMS 140: Network interface unit

Claims (10)

In a device for providing recommendation services through social media activity analysis,
at least one memory;
at least one database unit; and
includes at least one processor;
The at least one processor,
(a) Process of receiving and registering a first keyword of interest;
(b) a process of crawling related social media content from a plurality of social media platforms using the first keyword of interest;
(c) Analyzing the crawled social media content to extract user account information of at least one valid content in the order of highest relevance to the first interest keyword, and extracting the extracted user's account information into the first interest keyword A process of storing effective user account information corresponding to a keyword in the database unit;
(d) a process of crawling the active user's social media activity content based on the active user account information stored in the database unit;
(e) extracting the crawled content into standardized text through morphological analysis, classifying the text into categories through machine learning, assigning category values, and storing them in the database;
(f) Time-dependent action path information and movement of the effective user based on at least one of creation time information of the content, context before and after creation of the content, time information included in the content, and positional relationship of the text. A process of generating at least one piece of route information and storing it in the database unit; and
(g) When a first keyword of interest is input from the user, the database unit extracts the action path information and the movement path information related to the first keyword of interest, and executes a recommendation algorithm to provide the user with location, behavior, Processes a process of recommending at least one of a travel course and a product;
In process (f), the processor
Based on the creation time information of the content, the context before and after the creation of the content, the time information included in the content, and the positional relationship of the text, a chronological order is determined and an order value is assigned, and the category value is assigned. A recommendation service providing device through social media activity analysis, characterized in that it generates at least one of the action path information and the movement path information by arranging text according to the time series order value.
According to paragraph 1,
The category includes at least one of location/place, food/product, brand, culture/leisure, nature/environment, person/family, issue, and emotion/satisfaction. A device for providing recommendation services through social media activity analysis. .
According to paragraph 2,
In process (e), the processor
A recommendation service providing device through social media activity analysis, characterized in that it further processes a process of extracting location information indicating a location where the user created the content or an artificially designated location from the crawled content.
delete According to paragraph 3,
In process (f), the processor
User behavior based on the creation time information of the content, the context of creation of the content, the time information included in the content, the location information included in the content, the positional relationship of the text, and the category value of the text, or Extracting a place, grouping the text in a time-serial order according to the action or place, and assigning an order value to generate at least one of the action path information and the movement path information through social media activity analysis Recommendation service provision device.
According to paragraph 3,
The processor,
A recommendation service providing device through social media activity analysis, characterized in that the location information included in the content and the text given the category value are registered as keywords.
According to paragraph 1,
In process (e), the processor,
A recommendation service providing device through social media activity analysis, characterized in that the process of extracting text indicating satisfaction from the standardized text extracted from the crawled content and assigning weight to the corresponding text is further processed.
According to clause 6,
The recommendation algorithm assigns a score according to the number of users or contents with actions, places, and paths included in the extracted action path information and the movement path information, and based on the given score, a plurality of action path information and Providing a recommendation service through analysis of social media activities, including the process of calculating a score corresponding to the movement path information and recommending at least one of a place, action, movement course, and product to the user according to the calculated score. Device.
In a method of providing a recommendation service through analysis of social media activities performed by at least one processor,
(a) receiving and registering a first keyword of interest;
(b) crawling relevant social media content from a plurality of social media platforms using the first keyword of interest;
(c) Analyzing the crawled social media content to extract user account information of at least one valid content in the order of highest relevance to the first interest keyword, and extracting the extracted user's account information into the first interest keyword Storing effective user account information corresponding to the keyword in a database unit;
(d) crawling the active user's social media activity content based on the active user account information stored in the database unit;
(e) Extracting creation time information and location information of the content from the crawled content, extracting standardized text through morphological analysis, classifying the text into categories through machine learning, and assigning category values to the database unit Save to;
(f) Time-dependent action path information and movement of the effective user based on at least one of creation time information of the content, context before and after creation of the content, time information included in the content, and positional relationship of the text. Generating at least one piece of path information and storing it in the database unit; and
(g) When a first keyword of interest is input from the user, the database unit extracts the action path information and the movement path information related to the first keyword of interest, and executes a recommendation algorithm to provide the user with location, behavior, Comprising a step of recommending at least one of a travel course and a product;
The step (f) determines the chronological order based on the creation time information of the content, the before-and-after relationship of creation of the content, the time information included in the content, and the positional relationship of the text, and assigns an order value; , A method of providing a recommendation service through social media activity analysis, characterized in that at least one of the action path information and the movement path information is generated by arranging the text to which the category value is assigned according to the time-series order value.
delete

Images