KR20100070661A - Apparatus and method for offering item recommendation - Google Patents

Abstract

PURPOSE: An apparatus and a method for offering a recommendation item are provided to supply a recommendation item of high quality using preference information and context information of a user. CONSTITUTION: A filtering object item extracting unit(110) extracts a filtering object item using context information of the first user. A recommendation item list extracting unit(120) extracts a recommendation item list among the filtering object items by a collaborative filtering type. A recommendation item providing unit(130) provides the extracted recommendation item list. A context tag setup unit(140) sets a context tag in the first item according to a context when the second user accesses the first item.

Description

Apparatus and Method for Offering Item Recommendation}

The present invention relates to an apparatus and a method for providing a recommendation item.

Today's world is brimming with information, and the rate of increase of information is increasing, and the cycle to double the total amount of information is getting faster.

As a result, today's users inevitably have a necessity to be provided with only the information that is really necessary for them.

Researches are being conducted in academia and industry to provide users with the necessary information, and accordingly, recommendation systems such as providing customized advertisements, recommending customized products at shopping malls, recommending movies at movie ticket homepages, and recommending music at music sales homepages. This is actually used.

However, it is still difficult to provide information that the user feels really useful because it is merely using information of other users' preferences (such as movie ratings) or simply using personal information of the user.

In particular, if only the preference information of another user is used, the current user's taste is not considered. If only the personal information of the current user is used, preference information of another user or preference information of another item of the current user may be reflected. It was difficult to provide a high quality referral system.

In addition, the ubiquitous environment is expected to become popular in the near future, and the user's context information can be collected and used more easily and in detail, thereby providing a recommendation system that utilizes the context information more effectively. Is being requested.

The present invention has been proposed to solve the above-described problem, and an object thereof is to provide an apparatus and method for providing a high quality recommendation item using both user preference information and context information.

In addition, an object of the present invention is to provide a recommendation item providing apparatus and method for enhancing the user's satisfaction by providing an item actually preferred by the user as a recommendation item and efficiently providing a product or a service.

Another object of the present invention is to provide an apparatus and method for providing a recommendation item effectively using context information of a user in a ubiquitous environment.

According to an aspect of the present invention, there is provided an apparatus for providing a recommendation item.

The recommendation item providing apparatus for providing a recommendation item to a first user according to an embodiment of the present invention may include a filtering target item extracting unit and a collaborative filtering method for extracting a filtering target item using context information of the first user. It may include a recommendation item list extraction unit for extracting a list of recommended items of the filtering target item by a collaborative filtering method and a recommendation item list providing unit for providing the extracted recommendation item list.

According to another aspect of the present invention, a method for providing a recommendation item is provided.

In the method of providing a recommendation item for providing a recommendation item to a first user according to an embodiment of the present disclosure, the method may further include extracting a filtering target item using context information of the first user. The method may include extracting a list of items and providing the extracted list of recommended items.

According to another aspect of the present invention, a recording medium having a program recorded thereon is provided.

A program of instructions that can be executed by an electronic device is tangibly implemented to provide a recommendation item to a first user, and a recording medium that records a program that can be read by the electronic device can read context information of the first user. Extracting a filtering target item by using the method, extracting a recommendation item list among the filtering target items by a collaborative filtering method, and providing the extracted recommendation item list.

According to the present invention, there is an effect of providing an apparatus and method for providing a high quality recommendation item using both user preference information and context information.

In addition, the present invention has the effect of increasing the user's satisfaction by providing an item that the user actually prefers as a recommendation item so that the product or service can be efficiently provided.

In addition, the present invention has the effect of providing an apparatus and method for providing a recommendation item effectively using context information of a user in a ubiquitous environment.

Hereinafter, an embodiment of an apparatus and a method for providing a recommended item according to the present invention will be described in detail with reference to the accompanying drawings, and in the following description with reference to the accompanying drawings, the same or corresponding components are given the same reference numbers. Duplicate explanations will be omitted.

1A is a block diagram of an apparatus 100 for providing recommendation items according to an exemplary embodiment.

Referring to FIG. 1A, the recommendation item providing device 100 according to an exemplary embodiment may include a filtering target item extracting unit 110, a recommendation item list extracting unit 120, a recommendation item list providing unit 130, and the like. The context tag setting unit 140 may be included.

The recommendation item providing apparatus 100 of FIG. 1A is a device for providing a recommendation item to a first user. Here, the recommended item may be various goods, services, information, or the like according to an embodiment. For example, a movie may be recommended to a mobile terminal user who wants to reserve a movie, and a product (fund, deposit, loan, etc.) suitable for a mobile terminal user who wants to view a financial business may be recommended. You can also recommend news stories that are relevant to your time and place. A variety of sensor technologies, such as Radio Frequency Identification (RFID) systems, can be used to recommend products that are appropriate for the user at large discount stores or department stores.

For convenience, the embodiment of FIG. 1A assumes that a book is recommended.

In the description of the embodiment of FIG. 1A, a first user refers to a user who is currently visiting a shopping mall or a bank homepage, and recommends an appropriate item to the first user by using context information of the first user. .

The context tag setting unit 140 sets a context tag to the first user according to context information of the user.

The context information of the user refers to information about the situation of the user. That is, for example, it may include all information that can be collected, such as personal information of the user, current time information, location information, and the like.

The personal information of the user may include, for example, one or more of information such as gender, age, education, occupation, income, housing classification, hobbies, and the like. The timing information may include, for example, one or more of information such as Christmas season, holiday, vacation, season, Olympics, and the like. The location information may be the current user's country and city information, and may include specific location information in a large discount store, for example, one or more of location information having an error of 1 m range or nearest product corner information.

Such user's context information may be collected by a user's direct input, may be collected by an administrator (such as subscription to an Internet homepage), or may be collected by an automated program. For example, when the current date is a date between December 1 and December 25, the user's context information may include 'Christmas season' information.

When the application of the ubiquitous environment spreads, context information of a user may be automatically collected using an RFID system.

The collected context information of the user is set as a tag to the first user.

For example, if user A is 35 years old and female and the context information of the current Christmas season has been collected, three tags can be set for user A (mid-30s, men, Christmas). If you later want to extract the tags for user A, the above three tags can be provided.

The context tag setting unit 140 may also set a tag for each item. For example, a tag set for a book may be set by analyzing a title or set by a publisher in charge or a book shopping mall manager.

According to another embodiment, a context may be set for an item according to a user context when a user accesses a specific item. It is using a kind of folksonomy.

That is, the context tag of the user who is interested in the item (purchase or inquiry) may be analyzed to set a context tag on the item. For example, if 98% of the users who purchased item A are 23-26 years old, you can set a tag of 'mid 20's for item A. Or, you could set a 'Christmas' tag on a book that contains 'Christmas' in the title. The real estate item in Yeoksam-dong, Gangnam-gu, Seoul can be tagged 'Yeoksam-dong, Gangnam-gu, Seoul', and the 'Jeju-do' tag can be set for newspaper articles mainly read by users in Jeju.

In this way, context tag setting using folksonomy can share the user's context (context) information for recommending items for other users, easily access and search contents, and automatically recommend system without inconvenience. There is an advantage to improve the reliability of.

The filtering target item extractor 110 extracts the filtering target item by using context information of the first user.

For example, in the case of an internet book shopping mall, hundreds of thousands to hundreds of millions of book products may exist in a database. Among them, books that match user context information may be extracted as items to be filtered.

To this end, the context tag setting unit 140 may utilize a tag set for the first user. If the tag of the first user (35 years old, male, Christmas) is set, this tag can be utilized. Even without the tag setting method, context information of the first user may be collected and used.

The filtering target item extractor 110 may extract, as a filtering target item, an item in which a tag matching the tag of the second or more limit number among the context tags set in the first user is set.

The second limit number may be set by a user or an administrator.

For example, suppose that the first user is a woman in her mid 30s and the season is Christmas season. If the second limit number is 3, the book items in which all the tags corresponding to 'mid-30s', 'women', and 'Christmas' are set among the book items existing in the database may be extracted as the filtering target items.

When the second limit number is 1, a book item in which at least one of tags corresponding to 'mid-30s', 'women', and 'Christmas' among book items existing in the database is set may be extracted as a filtering target item.

In some embodiments, the tags are not only mechanically identical but also logically identical, i.e. synonyms, synonyms (eg eggs and eggs), or expressions in other languages with the same meaning (eg eggs, eggs and egg), Alternatively, the tags may correspond to the expressions (water sports and wind surfing) that are in a vertical relationship with each other.

For example, when the 'water sports' tag is set for the first user and the 'wind surfing' tag is set for the first item, the 'water sports' tag and the 'wind surfing' tag may correspond.

The recommended item list extractor 120 extracts a list of recommended items from the filtering target items by collaborative filtering.

Collaborative filtering is a method that automatically predicts users' interests based on taste information obtained from many users. The fundamental assumption of the collaborative filtering approach is the premise that users' past trends will remain the same in the future. For example, a collaborative filtering or recommendation system for music uses a partial list of users' likes (likes or dislikes) to predict their preferences for the music. The system is not limited to specific user information, but uses information collected from many users. This is different from simply treating the average rating of each item's interest based on the number of votes. In other words, it is a technique to identify customers with similar patterns in preference and interest based on their preferences and expressions of interest. Products that have not yet purchased each other to customers with similar tastes are used to provide services in the form of cross-recommending or recommending related products according to the tastes or lifestyles of the classified customers.

Types of collaborative filtering include item based filtering and user based filtering.

Cooperative filtering will be described later in detail with reference to FIGS. 2A and 2B.

The recommended item list providing unit 130 provides a list of recommended items to the first user.

For example, a list of links to a purchase page of a recommended book may be displayed on a portion of a terminal screen of a mobile terminal user who wants to purchase a book in a bookstore.

1B is an interface for providing a recommendation item list according to an embodiment of the present invention.

In the user field 150, an identifier of the current user is described, and in the current context information field 160, 35 years old, a woman, and a Christmas are described as the context of the current user. According to this information, the recommended book list 170 is provided. In the embodiment of FIG. 1B, only the title of the book is presented, but information such as a cover photo, author, ISBN, publisher, and price of the book may also be provided.

2A is a block diagram of a recommendation item list extractor 120 according to an embodiment of the present invention.

In the embodiment of FIG. 2A, the recommendation item list extractor 120 uses an item-based filtering technique among the collaborative filtering techniques. That is, the expected preference for the target item of the first user is extracted based on the preference of other items (items other than the target item) of the first user (current user).

The recommendation item list extractor 120 according to the embodiment of FIG. 2A includes a similarity table generator 210a, a preference extractor 220a, an expected preference extractor 230a, a list extractor 240, and a sorter between items. 250 may be included.

The similarity table generation unit 210a between items generates a database in which similarity between items is calculated.

Table 1 is a similarity table between items according to an embodiment of the present invention.

A B C D E F A One 0.023 0.3 0.128 0.103 0.148 B 0.023 One 0.045 0.222 0.091 0.4 C 0.3 0.045 One 0.105 0.065 0.182 D 0.128 0.222 0.105 One 0.12 0.073 E 0.103 0.091 0.065 0.12 One 0.2 F 0.148 0.4 0.182 0.073 0.2 One

The value calculated by assuming that the similarity between items that are completely identical is 1 and the similarity between items that are not related at all is 0.

Referring to Table 1, A and D have a similarity of 0.128, and B and F have a similarity of 0.4. It can be seen that B is relatively similar to F (0.4) and less similar to E (similarity 0.091) or D (0.222).

Item information may be used to generate the similarity table. For example, if the information of the publisher, author, etc. is the same, the similarity will be increased, and the similarity of the number of pages, the genre, and the title will be the basis for calculating the similarity. Several techniques may be used for the similarity extraction, but since this is a known technique, a detailed description thereof will be omitted.

The preference extracting unit 220a extracts a preference for the first evaluation item of the first user. That is, a preference is extracted for items that the first user has already purchased or viewed and rated (such as a book rating).

It is assumed that the first user has already evaluated items D, E, and F in Table 1 above.

It is assumed that the first user has a preference of 4.5, 4.0, and 1.0 for D, E, and F, respectively.

Such a preference may be measured by user input, administrator input, or user behavior pattern analysis.

That is, the first user may directly input the preferences for the D, E, and F items, or the administrator may input the user's evaluation opinion. Alternatively, the user may give high preference to items that are frequently accessed or items that have been viewed for a long time.

The expected preference extracting unit 230a extracts an expected preference for the filtering item of the first user based on a preference of the first evaluation item of the first user and a similarity table between the items.

The expected preference extractor 230a may include a weighted preference extractor 231a and a regular preference extractor 232a.

The weight preference extractor 231a extracts a first sum of a value obtained by multiplying the similarity between the first evaluation item and the first evaluation item by the first user's preference for each evaluation item.

Table 2 is a table for explaining an expected preference extraction process according to an embodiment of the present invention.

Item preference Similarity with A (A analogy) x (preference) Similarity with B (B Similarity) x (Preference) Similarity with C (C similarity) x (preference) D 4.5 0.128 0.576 0.222 0.999 0.105 0.473 E 4.0 0.103 0.312 0.091 0.364 0.065 0.26 F 1.0 0.148 0.148 0.4 0.4 0.182 0.182 total 0.379 1.036 0.713 1.763 0.352 0.915 Normal value 2.734 2.473 2.599

The filtering target item extractor 110 extracts an item including D, E, and F as a filtering target item, and assumes that the first user inputs a preference for A, B, and C. See Table 1.

The 'Preferences' column indicates a preference for each evaluation item of the first user. The first user may know that the items D, E, and F have a preference of 4.5, 4.0, and 1, respectively.

The column Similarity to A shows the similarity between each item and item A. A and D have a similarity of 0.128, A and E have a similarity of 0.103, and A and F have a similarity of 0.148.

In the (A similarity) x (preference) column, the value of the first user's preference for each item multiplied by the similarity between each item and A is described.

4.5 * 0.128 = 0.576 for D, 4.0 * 0.103 = 0.312 for E, 1.0 * 0.148 = 0.148 for F. This is a value considering the weight according to the similarity.

Accordingly, the sum (first sum) of the values for the D, E, and F items of the (A similarity) x (preference) column becomes 1.036.

The regular preference extractor 232a divides the first sum into a sum of similarities and extracts an expected preference. That is, the regular preference 2.734 may be extracted by dividing the first total (1.036) by the total similarity (0.379) for each item of A.

The reason for dividing the first total by the sum of the similarities is that the weighted values are multiplied by the similarities. This is for estimating the user's expected preference regardless.

In this way, the expected preferences of 2.784, 2.473, and 2.599 can be extracted for the A, B, and C items.

In the above example, the expected preference is extracted for only three items, but the expected preference may be extracted for a larger number of items in actual application.

The list extracting unit 240 extracts a list of recommended items including a first limited number of items from the filtering items based on the expected preference.

The first limit number may be set by a user or an administrator, or determined by a user interface (such as a screen size of a web browser or a terminal).

For example, if the first limit number is set to 2, items A and C having high expected preference among items A, B, and C of Table 2 may be included in the recommendation item list.

According to another embodiment, only an item having an expected preference above the first limited preference may be included in the recommended item list. For example, only items having a preference of 2.5 or more may be included in the recommendation item list. In this case, only A and C are included in the recommendation item list.

The sorter 250 sorts the list of recommended items based on the expected preference. If all of the items A, B, and C of Table 2 are included in the recommendation item list, the items may be sorted in the order of A, C, and B, which are high expected preferences, and provided to the user.

2B is a block diagram of a recommendation item list extractor 120 according to another embodiment of the present invention.

In the embodiment of FIG. 2B, the recommendation item list extractor 120 uses a user-based filtering technique among the collaborative filtering techniques. The expected preference of the target item of the first user (current user) is extracted based on the preference of the target item of another user (a user other than the first user).

The user-based filtering technique is very similar to the item-based filtering technique, but differs in that the basis for extracting the expected preference is similarity between user preference information and users other than the current user.

Referring to FIG. 2B, the recommendation item list extractor 120 includes a similarity table generator 210b, a preference extractor 220b, an expected preference extractor 230b, a list extractor 240, and a sorter between users. 250 may be included.

The similarity table generation unit 210b between users generates a database that calculates the similarity between each user.

Table 3 is a similarity table between users according to an embodiment of the present invention.

P Q R S T U P One 0.023 0.3 0.128 0.103 0.148 Q 0.023 One 0.045 0.222 0.091 0.4 R 0.3 0.045 One 0.105 0.065 0.182 S 0.128 0.222 0.105 One 0.12 0.073 T 0.103 0.091 0.065 0.12 One 0.2 U 0.148 0.4 0.182 0.073 0.2 One

The value calculated by assuming that the similarity between completely identical users is 1 and that the similarity between completely dissimilar users is 0.

Referring to Table 3, user P and user S have a similarity of 0.128, and user Q and user U have a similarity of 0.4. It can be seen that Q is relatively similar to U (0.4) and less similar to T (similarity 0.091) or S (0.222).

User information may be used to generate the similarity table. For example, if the genders are the same, the similarity may be increased, and the greater the age difference, the lower the similarity. In addition, similarity may be increased when occupation groups are similar, and similarity may be increased when hobbies are the same. Several techniques may be used for the similarity extraction, but since this is a known technique, a detailed description thereof will be omitted.

The preference extracting unit 220b extracts a preference for the filtering target item of the pre-evaluation user.

Assume that the item to be filtered is A. It is assumed that preference information is stored by S, T, and U users (previous evaluation users) for item A.

In other words, S, T, and U users have already purchased or viewed items A and rated them (such as book ratings), and extract their preferences.

For example, assume that S user has a preference of 4.5, T user has 4.0, and U user has a preference of 1.0.

Such a preference may be measured by user input, administrator input, or user behavior pattern analysis.

That is, the S, T, and U users may directly input the preference for the item A, or the administrator may input the user's evaluation opinion. Alternatively, the user may give high preference to items that are frequently accessed or items that have been viewed for a long time.

An expected preference extractor 230b extracts an expected preference for the filtering item of the first user based on the preference of the first evaluation user and the similarity table between the users.

The predicted preference extractor 230b may include a weighted preference extractor 231b and a normal preference extractor 232b.

The weight preference extracting unit 231b extracts a first sum of a value obtained by multiplying the similarity between the respective evaluation user and the first user by the preference for the first item of each evaluation user.

Table 4 is a table for explaining an expected preference extraction process according to another embodiment of the present invention.

user preference Similarity with P (P Similarity) x (Preference) Similarity with Q (Q Similarity) x (Preference) Similarity with R (R similarity) x (Preference) S 4.5 0.128 0.576 0.222 0.999 0.105 0.473 T 4.0 0.103 0.312 0.091 0.364 0.065 0.26 U 1.0 0.148 0.148 0.4 0.4 0.182 0.182 total 0.379 1.036 0.713 1.763 0.352 0.915 Normal value 2.734 2.473 2.599

It is assumed that the filtering target item extracting unit 110 extracts an item including the A item as a filtering target item, and that S, T, and U users input the preference for the A item. Reference was made.

The 'Preferences' column shows the preferences for items A, S, T, and U. For items A, S, T, and U users can see that they have a preference of 4.5, 4.0, and 1, respectively.

The similarity with P column indicates the similarity between each user and the P user. User S and user P have a similarity of 0.128, user P and user T have a similarity of 0.103, and user P and user U have a similarity of 0.148.

The (P Similarity) x (Preference) column describes a value obtained by multiplying the preference for each item of the first user by the similarity between A and each item.

4.5 * 0.128 = 0.576 for S, 4.0 * 0.103 = 0.312 for T, 1.0 * 0.148 = 0.148 for U. This is a value considering the weight according to the similarity.

Accordingly, the sum (first sum) of the values for the S, T, and U users of the (P similarity) x (preference) column is 1.036.

The regular preference extractor 232b divides the first total by the sum of the similarities and extracts the expected preference. That is, the regular preference 2.734 may be extracted by dividing the first total (1.036) by the total similarity (0.379) for each user of P.

The reason for dividing the first total by the sum of the similarities is that since the weighted value is multiplied by the similarity, the items that have received the preference ratings from users who are similar to the first user may receive a relatively high preference rating. This is to estimate the expected preference of the user regardless of who received it.

In this way, the expected preferences of 2.784, 2.473, and 2.599 can be extracted for P, Q, and R users. That is, it can be expected that the P user has a preference of 2.784, the Q user of 2.478, and the R user of 2.599.

In the same way as above, it is possible to extract the expected preference of P for all the filtering target items.

The list extractor 240 and the sorter 250 sort and provide a list of recommended items based on the predicted preference for each item extracted by the predicted preference extractor 230b. This has been described above with reference to FIG. 2A.

3 is a flowchart illustrating a recommendation item providing process according to an embodiment of the present invention.

In operation S310, the recommendation item providing apparatus 100 sets a context tag to the first user according to the context information of the first user (current user). Since this has been described in detail in the description of the context tag setting unit 140 of FIG. 1A, a detailed description thereof will be omitted.

In operation S320, the recommendation item providing apparatus 100 sets a context tag to the first item according to a context when a second user accesses the first item. Here, the second user refers to a user except the first user (current user).

Since this has been described in detail in the description of the context tag setting unit 140 of FIG. 1, a detailed description thereof will be omitted.

In operation S330, the filtering target item is extracted using the context information of the first user. In other words, an item suitable for this is extracted as the filtering target item using context information such as the gender, age, and time of the current user. Since this has been described in detail in the description of the filtering target item extractor 110 of FIG. 1A, a detailed description thereof will be omitted.

In step S340, a list of recommended items from the filtering target items is extracted by the collaborative filtering method. This has been described in detail with reference to the recommendation item list extractor 120 of FIG. 1A and FIGS. 2A and 2B. In addition, the recommended item list extraction process will be described with reference to FIGS. 4A to 4D.

A list of recommended items extracted in step S350 is provided. Since it has been described in detail in the description of the recommendation item list providing unit 130 of FIG. 1A and the description of FIG. 1B, a detailed description thereof will be omitted.

4A is a detailed flowchart of step S340 of FIG. 3 according to an embodiment of the present invention.

The process of FIG. 4A uses an item-based collaborative filtering technique.

In step S410a, the recommended item list extractor 120 generates a similarity table between items. This has been described in detail in the description of the similarity table generator 210a between the items of FIG. 2A.

In operation S420a, the recommended item list extractor 120 extracts a preference for the first evaluation item of the first user. This has been described in detail in the description of the preference extracting unit 220a of FIG. 2A.

In operation S430a, the recommendation item list extractor 120 extracts an expected preference for the filtering item of the first user based on a preference for the first evaluation item of the first user and a similarity table between the items.

4B is a detailed flowchart of step S430a of FIG. 4A.

In operation S431a, the recommendation item list extractor 120 extracts a first sum of a value obtained by multiplying the similarity of each evaluation item and the first filtering item by a preference of the first user for each evaluation item. This has been described in detail in the description of the weight preference extractor 231a of FIG. 2A.

In step S432a, the recommended item list extractor 120 extracts an expected preference by dividing the first total by the total of the similarities. This is described in detail in the description of the normal preference extractor 232a of FIG. 2A.

In operation S440, the recommendation item list extractor 120 extracts a recommendation item list including the first limited number of items from the filtering items based on the expected preference. This has been described in detail in the description of the list extractor 240 of FIG. 2A.

In operation S450, the recommended item list extractor 120 sorts the recommended item list based on the expected preference. This has been described in detail in the description of the alignment unit 250 of FIG. 2A.

The list of recommended items extracted through the above process is provided to the user through the interface as shown in FIG. 1B.

4C is a detailed flowchart of step S340 of FIG. 3 according to an embodiment of the present invention.

The process of Figure 4c uses a user-based collaborative filtering technique.

In step S410b, the recommended item list extractor 120 generates a similarity table between users. This is described in detail in the description of the similarity table generation unit 210b of FIG. 2B.

In operation S420b, the recommended item list extractor 120 extracts a preference for the filtering target item of the pre-evaluation user. This has been described in detail in the description of the preference extracting unit 220b of FIG. 2B.

In operation S430b, the recommendation item list extractor 120 extracts an expected preference for the filtering item of the first user based on the preference for the first item of the pre-evaluation user and the similarity table between the users.

4D is a detailed flowchart of step S430b of FIG. 4C.

In operation S431b, the recommendation item list extractor 120 extracts a first sum of a value obtained by multiplying the similarity between the respective evaluation users and the first user by the preferences of the first evaluation users of the respective evaluation users. This has been described in detail in the description of the weight preference extracting unit 231b of FIG. 2B.

In operation S432b, the recommended item list extractor 120 extracts an expected preference by dividing the first total by the total of the similarities. This is described in detail in the description of the normal preference extractor 232b of FIG. 2B.

In operation S440, the recommendation item list extractor 120 extracts a recommendation item list including the first limited number of items from the filtering items based on the expected preference. This has been described in detail in the description of the list extractor 240 of FIG. 2A.

In operation S450, the recommended item list extractor 120 sorts the recommended item list based on the expected preference. This has been described in detail in the description of the alignment unit 250 of FIG. 2A.

The list of recommended items extracted through the above process is provided to the user through the interface as shown in FIG. 1B.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise.

In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

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.

An embodiment of the present invention may include a computer readable medium including program instructions for performing various computer-implemented operations. The computer readable medium may include program instructions, local data files, local data structures, or the like, alone or in combination. The media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts.

In addition, the components illustrated in FIGS. 1A, 2A, and 2B do not necessarily have a hardware configuration, and some components may be implemented in the form of an application program that performs the same or similar functions. In addition, each component may be combined or separated within the scope of the invention.

So far I looked at the center of the embodiment for the present invention. Many embodiments other than the above-described embodiments are within the claims of the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. The disclosed embodiments should, therefore, be considered in an illustrative rather than a restrictive sense. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

1A is a block diagram of an apparatus 100 for providing recommendation items according to an exemplary embodiment.

1B is an interface for providing a recommendation item list according to an embodiment of the present invention.

2A is a block diagram of a recommendation item list extractor 120 according to an embodiment of the present invention.

2B is a block diagram of a recommendation item list extractor 120 according to another embodiment of the present invention.

3 is a flowchart illustrating a recommendation item providing process according to an embodiment of the present invention.

4A is a detailed flowchart of step S340 of FIG. 3 according to an embodiment of the present invention.

4B is a detailed flowchart of step S430a of FIG. 4A.

4C is a detailed flowchart of step S340 of FIG. 3 according to an embodiment of the present invention.

4D is a detailed flowchart of step S430b of FIG. 4C.

Claims (17)

In the recommendation item providing apparatus for providing a recommendation item to a first user, A filtering target item extracting unit which extracts a filtering target item by using context information of the first user; A recommended item list extracting unit which extracts a list of recommended items from the filtering target items by a collaborative filtering method; And And a recommendation item list providing unit for providing the extracted recommendation item list. The method of claim 1, The recommended item list extraction unit, An item similarity table generating unit generating an item similarity table; A preference extracting unit which extracts a preference of the first evaluation item of the first user; An expected preference extracting unit configured to extract an expected preference for the filtering item of the first user based on a preference table of the first evaluation item of the first user and a similarity table between the items; And And a list extracting unit which extracts a list of recommended items including a first limited number of items from the filtering items based on the expected preference. The method of claim 1, The recommended item list extraction unit, A similarity table generator between users for generating a similarity table between users; A preference extracting unit which extracts a preference of the filtering target item of a pre-evaluation user; An expected preference extracting unit configured to extract an expected preference for the filtering item of the first user based on a preference table of the first evaluating user and a similarity table between the users; And And a list extracting unit which extracts a list of recommended items including a first limited number of items from the filtering items based on the expected preference. The method according to claim 2 or 3, The recommended item list extraction unit, And a sorting unit for sorting the list of recommended items based on the expected preferences. The method of claim 2, The expected preference extracting unit, A weight preference extracting unit configured to extract a first total of a value obtained by multiplying the similarity between the respective evaluation items and the first filtering item by the preferences of the first user for each evaluation item; And a normal preference extracting unit which extracts an expected preference by dividing the first total by the total of the similarities. The method of claim 3, The expected preference extracting unit, A weight preference extracting unit for extracting a first total of a value obtained by multiplying the similarity between the respective evaluation user and the first user by the preference for the first item of each evaluation user; And a normal preference extracting unit which extracts an expected preference by dividing the first total by the total of the similarities. The method according to any one of claims 1 to 3, The recommended item providing device, The apparatus may further include a context tag setting unit configured to set a context tag to the first user according to the context information of the first user. The filtering item extraction unit, The apparatus for providing a recommendation item, which extracts, as a filtering target item, an item in which a tag matching a tag equal to or greater than a second limit number among the context tags set in the first user is set. The method according to any one of claims 1 to 3, And a context tag setting unit configured to set a context tag in the first item according to a context of a second user's access to the first item. In the recommendation item providing method for providing a recommendation item to the first user, Extracting an item to be filtered using the context information of the first user; Extracting a list of recommended items from the filtering target items by a collaborative filtering method; And And providing a list of the extracted recommended items. 10. The method of claim 9, Extracting a list of recommended items of the filtering target items by the collaborative filtering method, Generating a similarity table between items; Extracting a preference for the first evaluation item of the first user; Extracting an expected preference for the filtering item of the first user based on a preference table of the first evaluation item of the first user and a similarity table between the items; And And extracting a recommendation item list including a first limited number of items from the filtering items based on the expected preference. 10. The method of claim 9, Extracting a list of recommended items of the filtering target items by the collaborative filtering method, Generating a similarity table between users; Extracting a preference for the filtering target item of a pre-evaluation user; Extracting an expected preference for the filtering item of the first user based on a preference table of the first evaluating user and the similarity table between the users; And And extracting a recommendation item list including a first limited number of items from the filtering items based on the expected preference. The method according to claim 10 or 11, wherein Extracting a list of recommended items of the filtering target items by the collaborative filtering method, And sorting the list of recommended items based on the estimated preferences. The method of claim 10, Extracting an expected preference for the filtering item of the first user based on the preference table of the first evaluation item of the first user and the similarity table between the items, Extracting a first total of a value obtained by multiplying a similarity between the respective evaluation items and the first filtering item by the preferences of the first user for each evaluation item; And extracting an expected preference by dividing the first sum by the sum of the similarities. The method of claim 11, Extracting an expected preference for the filtering item of the first user based on the preference of the first evaluation user of the first evaluation user and the similarity table between the users, Extracting a first sum of a value obtained by multiplying a similarity between the respective evaluation user and the first user by a preference of the respective evaluation user with the first item; And extracting an expected preference by dividing the first sum by the sum of the similarities. The method according to any one of claims 9 to 11, The recommended item providing method, Setting a context tag on the first user according to the context information of the first user, Extracting the filtering target item using the context information of the first user, And extracting, as a filtering target item, an item in which a tag matching the tag of the second or more limit number among the context tags set in the first user is set as the filtering target item. The method of claim 15, The recommended item providing method, And setting a context tag in the first item according to a context of a second user's access to the first item. In the recording medium on which a program of instructions that can be executed by an electronic device for providing a recommendation item to a first user is tangibly implemented, and which records a program that can be read by the electronic device, Extracting an item to be filtered using the context information of the first user; Extracting a list of recommended items from the filtering target items by a collaborative filtering method; And And a recording medium for recording a program for executing the step of providing the extracted recommended item list.

Images