JP5738785B2 - Recommendation program, apparatus and method for recommending critics that meet user preferences - Google Patents

Description

  The present invention relates to a recommendation service technology for recommending items that meet user preferences.

  2. Description of the Related Art Conventionally, there is a recommendation service technology that recommends an item based on a user's preference using the user's implicit action history information. Here, the “user” refers to an entity that performs a selection action based on a preference such as purchase / viewing an item. Further, the “item” may be a product / service, and includes products such as home appliances, contents such as electronic books, and viewing contents such as web pages and televisions. Examples of the action history information include an item purchase history and a usage browsing history for the user. As an algorithm for estimating items to be recommended, there is typically “Collaborative Filtering” (for example, see Non-Patent Documents 1 and 2). According to this algorithm, it is estimated that an item with a high degree of relevance should be recommended to the user based on the action history for the user's item and the action history for another user's item. As a result, the item recommended to the user is highly likely to meet the user's preference.

  FIG. 1 is a system configuration diagram including an EC site server and an SNS site server in the prior art.

  According to FIG. 1, an EC (Electronic Commerce) site server 3 is connected to the Internet. The “EC site server” is a site that sells items to users, and specifically may be a site that downloads an application to a smartphone, for example. According to the EC site, a web beacon is embedded in an item purchase completion page. For example, the display history of the purchase completion page displayed when the user 1 purchases the item 3 is accumulated in the database as the purchase history.

  In recent years, according to the EC site, an unspecified number of users who are actually purchasing or considering an item can post a review comment (review) on the product. A user who has posted a critical comment in this way is generally referred to as a “reviewer”. The criticism comment is disclosed to an unspecified number of third parties, and a user who is considering purchasing the product can freely view the criticism comment and use it as a judgment material for purchase.

  Also, according to FIG. 1, an SNS (Social Networking Service) site server 5 is connected to the Internet. The “SNS site server” publishes posted text (which can be regarded as a critical comment for a specific item) posted by one user in a group of a plurality of users. For example, there are facebook (registered trademark), twitter (registered trademark), google + (registered trademark), and mixi (registered trademark), which are also generally referred to as miniblog sites. The critical comments include, for example, tweets about the app on the SNS site server and articles summarizing how to use the app on the blog site server. In the following description, the SNS site server is described. Of course, it may be a communication site server, and may be a blog server, a bulletin board site server, or a chat server.

  For example, according to the EC site server 3, for example, a “Like” button of facebook is displayed on the display part for each comment comment posted by a critic. This button simply means that the user has an interest card. General readers who have a good appreciation for the comment can click the “Like” button. Here, a general reader who performs evaluation (stage evaluation, rating) on a critical comment is hereinafter referred to as a “critical reviewer (rater)”. When the button is clicked by a critique evaluator, information indicating that the critique evaluator has received a critique of the critique comment is collected in the SNS site server 5. Then, the information is disclosed to other users belonging to the group of the SNS site server site. These other users will be able to see critical comments for which the critical reviewer has a critique.

  As can be seen from FIG. 1, according to the prior art, a case where the recommendation device 1 is provided on the Internet can be assumed. Such a recommendation device 1 recommends an item to each user in cooperation with, for example, the EC site server 3. That is, the recommendation device 1 functions completely different from the SNS site server 5.

  2. Description of the Related Art Conventionally, there is a technique for recommending a conversation partner who can communicate smoothly using history information during synchronous communication such as conversation or chat (see, for example, Patent Document 1). According to this technique, the attribute of the request source is compared with the attribute of the information provider, and the information provider who will provide information according to the preference of the request source can be selected. Thereafter, the activity of information exchange between them can be further collected, and the attribute of the information provider can be further updated according to the activity.

  In addition, the client sends a search request for an individual requesting a criticism comment to the server, and the server detects the document linked to the criticism comment by the similarity of the contents, and identifies the critic who created the document. There is a technique for extracting (see, for example, Patent Document 2). According to this technique, it is possible to calculate an index value of criticism experience in the same knowledge field and rank candidate individuals.

  Further, for an item selection support device that recommends a plurality of items that are suitable for the user, a similar representative recommender similar to the representative recommender is extracted, and the recommended item of the similar representative recommender is also recommended to the user. (For example, refer to Patent Document 3). The user is displayed so that recommended items can be selected.

JP 2011-013786 A JP 2010-061183 A JP 2007-200099 A

P. Resnick, N. Iacovou, M. Suchak, P. Bergstrom, and J. Riedl. “GroupLens: Open Architecture for Collaborative Filtering of Netnews” In Conference on Computer Supported Cooperative Work, pp. 175-186 (1994) Badrul Sarwar, George Karypis, Joseph Konstan, and John Reidl. 2001.Item-based collaborative filtering recommendation algorithms.In Proceedings ofthe 10th international conference on World 'Wide Web (WWW '01) .ACM, New York, NY, USA, 285 -295.

  However, according to the conventional recommendation service technology, only a preference item for the user is inferred, and there is an inorganic impression, which often does not lead to the user's purchase (use) willingness (marketing). In particular, the recommendation information itself has a strong image as a commercial (advertisement information, advertisement information), and is often less interesting for users with a low willingness to purchase. In addition, there are many cases where some users do not desire distribution of recommendation information such as advertisement information.

  In addition, if a critic's “Like” is given to a critic's critique comment through the SNS site server, other users will be forced to view it. It does not necessarily lead to purchase incentives. This is because the preferences of the members (followers and friends) of the group belonging to the SNS site server are not necessarily the same.

  As for the evaluation of the critic, the level of the critic's comment is graded (like simple evaluation! 5 stars (★★★ ☆☆)), Bad / Some of them are expressed as “Good”, and the information is certainly useful for users who are interested in the item. In addition, it is possible to carefully select useful criticisms such as displaying / not displaying criticism comments according to the classification of the critique evaluation levels by critics, and it is easy for the user to browse. However, as described above, if there is no preference for the item itself, it is not always useful for the user to view the comment even if the comment is recommended by the “Like” button. Absent.

  According to the technique described in Patent Document 1, only an appropriate one-to-one communication (conversation) partner is found. Moreover, according to the technique described in Patent Document 2, the user himself / herself searches consciously, and does not recommend implicit information. In addition, critics' evaluations are expressed in terms of activity, such as the number of document creation experiences, and are not based on the user who receives the review. Furthermore, according to the technique described in Patent Document 3, the user actively selects an item, and automatically detects a critic who describes a criticism comment based on the user's preference. Absent.

  Here, the inventors thought it beneficial to find a critic who met the user's preference. That is, it was considered preferable to recommend not only items (items) that meet the user's preference but also critics that met the user's preference. The critical comment for each item in the reviewer is assumed to match the user's preference.

  Therefore, the present invention provides a recommendation program, an apparatus, and a method capable of recommending to a user a reviewer who meets the user's preference while a review comment is posted by a large number of reviewers. The purpose is to provide.

According to the present invention, in the recommendation program for causing the computer mounted on the apparatus to function to recommend other items based on the item usage history information in the user,
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, collaborative filtering means for calculating an estimated preference degree of each item by collaborative filtering for user preference information based on the usage history or evaluation history of each item;
For each user identifier , user preference storage means for storing the estimated preference of each item;
For each reviewer identifier (critic), a reviewer rating calculation means for calculating an estimated rating by averaging, accumulating, normalizing or maximizing the evaluation level of each item of each rater identifier (critic evaluator);
For each reviewer identifier (critic), a reviewer evaluation degree accumulation means for accumulating an estimated evaluation degree of each item,
Reviewer usefulness calculating means for calculating , for each item, a sum of multiplication values of the estimated preference level in the user identifier to be recommended and the estimated evaluation level in the reviewer identifier as a “usefulness level” ;
Reviewer selecting means for selecting a reviewer identifier (critic) based on at least the highest usefulness for the user identifier;
The computer is made to function as a recommendation means for transmitting the selected reviewer identifier (critic) to a user-operated terminal.

According to another embodiment of the recommendation program of the present invention,
The reviewer evaluation degree storage means further stores a reviewer identifier and a critical comment corresponding to the item,
It is also preferable that the recommending means obtains a criticism comment criticized for each item by the reviewer based on the selected reviewer identifier from the reviewer evaluation degree accumulating means, and transmits it to the terminal operated by the user.

According to another embodiment of the recommendation program of the present invention,
Recommendation means are
From the user preference storage means, an item whose estimated preference is equal to or greater than a predetermined threshold for the user is selected and transmitted to the user operation terminal.
It is also preferable that a critic based on the selected reviewer identifier obtains a critical comment criticized on the selected item from the reviewer evaluation degree storage means or another server, and transmits it to the terminal operated by the user.

According to another embodiment of the recommendation program of the present invention,
For user preference information storage means, the item usage history stores the presence / absence of purchase / use of merchandise service as an item, or installation on an owned terminal using an application as an item, based on the user identifier It is also preferred to have the computer further function to be.

According to another embodiment of the recommendation program of the present invention,
It is also preferable that the computer further function so that the evaluation level determined by the reviewer in the reviewer evaluation level storage means stores one or more levels representing the proof of the review comment.

According to another embodiment of the recommendation program of the present invention,
Critic commentary based on the reviewer identifier is a comment text posted on an SNS (Social Networking Service) site or a blog (Web log) site,
The evaluation level judged by critics is “Like!”, “+1”, “Retweet”, “Bookmark”, “Good / bad rating”, “Not helpful / not helpful”, “Favorite”. Alternatively, it is also preferable to make the computer function so as to be at the “5-level evaluation” level.

According to the present invention, in the recommendation device for recommending other items based on the usage history information of items in the user,
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, collaborative filtering means for calculating an estimated preference degree of each item by collaborative filtering for user preference information based on the usage history or evaluation history of each item;
For each user identifier , user preference storage means for storing the estimated preference of each item;
For each reviewer identifier (critic), a reviewer rating calculation means for calculating an estimated rating by averaging, accumulating, normalizing or maximizing the evaluation level of each item of each rater identifier (critic evaluator);
For each reviewer identifier (critic), a reviewer evaluation degree accumulation means for accumulating an estimated evaluation degree of each item,
Reviewer usefulness calculating means for calculating , for each item, a sum of multiplication values of the estimated preference level in the user identifier to be recommended and the estimated evaluation level in the reviewer identifier as a “usefulness level” ;
Reviewer selecting means for selecting a reviewer identifier (critic) based on at least the highest usefulness for the user identifier;
And recommending means for transmitting the selected reviewer identifier (a critic) to the user-operated terminal.

According to the present invention, in the recommendation method for recommending other items based on the usage history of items in the user, using the device,
The device
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, a user preference storage unit that calculates an estimated preference level of each item by collaborative filtering on user preference information based on the usage history or evaluation history of each item, and stores the estimated preference level of each item;
Reviewers identifier (critic's) for each, averaging the evaluation level of each item of each rater identifier (critic evaluator), the cumulative reduction, to calculate the estimated evaluation degree normalized or maximum value of the estimated evaluation of each item A reviewer evaluation degree accumulation unit that accumulates
The device
A first step of calculating , for each item, a sum of multiplication values of an estimated preference level in a user identifier to be recommended and an estimated evaluation level in a reviewer identifier as a “usefulness level” ;
A second step of selecting a reviewer identifier (a critic) based on at least the highest usefulness for the user identifier;
And a third step of transmitting the selected reviewer identifier (critic) to a user-operated terminal.

  According to the recommendation program, the apparatus, and the method of the present invention, a reviewer with a high evaluation that meets the user's preference is recommended to the user while a comment is posted by a large number of reviewers on the item. can do.

It is a system block diagram containing EC site server and SNS site server in a prior art. It is a system configuration diagram in the present invention. It is explanatory drawing showing user preference information. It is explanatory drawing showing a reviewer evaluation level. It is a functional block diagram of the recommendation apparatus in this invention. It is a sequence diagram in the present invention. It is explanatory drawing showing the presumed preference degree computed by collaborative filtering. It is explanatory drawing showing the reviewer evaluation degree calculated by the reviewer evaluation degree calculation part. It is explanatory drawing showing the usefulness for every critic with respect to each user. It is a 1st sequence diagram which collects a critical comment and an evaluation level. It is a 2nd sequence diagram which collects a critical comment and an evaluation level. It is a 3rd sequence diagram which collects a critical comment and an evaluation level.

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

  FIG. 2 is a system configuration diagram according to the present invention.

  According to FIG. 2, a recommendation device (server) 1 connected to the Internet acquires user preference information from, for example, the EC site server 3 or a terminal 2 based on a user operation. “User preference information” is, for example, a usage history of each item represented by binary (0, 1) for each user identifier. Of course, the user preference information may be expressed in a stepwise level by an action such as “favorite”, “good / bad evaluation”, “bookmark”, and “five-step evaluation” of an item by the user. In addition, the “item usage history” stores, for example, whether or not the user based on the user identifier has purchased or used the product service as an item, or whether or not the application is installed on the owning terminal as an item. Is. Of course, the item is not limited to the product service, and may be a link (URL) to a web page describing the details of the product / service or a hash tag related to the product / service (to the user). , Recommend URLs and hashtags).

  FIG. 3 is an explanatory diagram showing user preference information.

  According to FIG. 3, for each user identifier (column), “1” is assigned to items (rows) purchased or used by the user in the past. Conversely, items that have not been purchased or used in the past are given “0”. According to FIG. 3, for example, the user 1 can understand that the applications 1, 2, 3, and 5 are already installed in his / her smartphone.

  Moreover, the recommendation apparatus 1 acquires the evaluation level for every reviewer identifier from the SNS site server 5, for example. The “evaluation level” is a value that is evaluated by each reviewer for each reviewer identifier and a review comment that the reviewer has reviewed for each item. The critic's comment may be a comment published on the same product detail page of the EC site, or a comment text posted on an external SNS site or blog site. When using a critical comment from an external SNS site or the like, it is also desirable to determine which item is a comment by using a commonly defined item identifier or by using a hash tag associated with the item.

  Here, the “item evaluation level” refers to one or more levels set by a critique evaluator and representing a remark on a critique comment critiqued by the critique of the item. Specifically, “Like!” On facebook, “+1” on google +, “Retweet” on Twitter, “Bookmark”, other “good / bad rating”, “not helpful / not helpful” "Favorite" or "5-level evaluation" level. For example, the number of times “Like” or “+1” is clicked or the number of “Retweet” may be expressed in a stepwise level.

  FIG. 4 is an explanatory diagram showing the reviewer evaluation level.

  According to FIG. 4, each reviewer identifier has a table of rater identifiers (columns) and items (rows). The rater identifier is an identifier assigned to each critic. The critique evaluator assigns an evaluation level to the critique comment for each item of the critic. According to FIG. 4, the critic evaluator 1 gives the evaluation level “4” to the criticism comment criticized by the critic 1 against the application 1. Also, the critic evaluator 3 assigns the evaluation level “2” to the critique comment criticized by the critic 1 against the app 1.

  The recommendation device 1 acquires user preference information and an evaluation level from the EC site server 3 and the SNS site server 5 via an API (Application Programming Interface). The API is a rule interface for using the function of the application service. This API is prepared differently for each server.

  And the recommendation apparatus 1 of this invention accumulate | stores user preference information and an evaluation level, The criticism comment is posted by many critics with respect to goods etc. (item), The said user's preference is met. Recommend the appropriate critic to the user.

FIG. 5 is a functional configuration diagram of the recommendation device according to the present invention.
FIG. 6 is a sequence diagram in the present invention.
In the following, each functional component of FIG. 5 will be described with reference to the sequence of FIG.

  According to FIG. 5, the recommendation device 1 includes a communication interface 10 connected to the Internet, a user preference storage unit 11, a reviewer evaluation storage unit 12, a reviewer usefulness calculation unit 13, and a reviewer selection unit 14. And a recommendation unit 15. These functional components excluding the communication interface 10 are realized by executing a recommendation program that causes a computer installed in the apparatus to function.

[User preference storage unit 11]
The user preference storage unit 11 stores the estimated preference of each item for each user identifier. Here, according to FIG. 5, the recommendation device 1 includes a user preference information storage unit 111 and a collaborative filtering unit 112.

  The user preference information accumulation unit 111 receives and accumulates user preference information (see FIG. 3 described above) for each user identifier, for example, from the EC site server 3 or the terminal 2 based on a user operation (see S61 in FIG. 6).

  The collaborative filtering unit 112 calculates the estimated preference degree of each item for each user identifier using collaborative filtering from the user preference information of the user preference information storage unit 111 (see S62 in FIG. 6). The calculated estimated preference level is accumulated by the user preference level accumulation unit 11 (see S63 in FIG. 6).

Collaborative filtering is performed in the following two steps using user preference information. However, the following steps are examples of collaborative filtering, and are not limited to this method.
(S1) The relevance between items is analyzed, for example, by counting the number of combinations of items purchased by users at the same time. For example, according to FIG. 3, three users have purchased the apps 1 and 3. That is, the second item purchased together with the first item is cross tabulated.
(S2) Next, an estimated ratio to be purchased as a set is calculated from the sum of the combinations. This estimated ratio is defined as an estimated preference level.

  FIG. 7 is an explanatory diagram showing the estimated preference degree calculated by collaborative filtering.

According to FIG. 7, for example, the user 1 has purchased the applications 1, 2, 3, and 5 in the past. On the other hand, the estimated preference degree that the user 1 purchases the application 4 is 0.73. In addition, the estimated preference for the user 1 to purchase the application 6 is 0.99. Here, it is estimated that the user 1 has a high probability of purchasing the application 6. The collaborative filtering algorithm is implemented in advance as a standard function in, for example, spreadsheet software.
Pref (U, A): Estimated preference level for the app A with respect to the user U However, the estimated preference level for an app that has already been purchased gives the maximum value defined for collaborative filtering. For example, if the preference degree of the collaborative filtering is defined as 0 to 1, Pref (U, A) = 1.

[Reviewer accumulation unit 12]
The reviewer evaluation degree accumulation unit 12 stores the estimated evaluation degree of the critical comment for each item for each reviewer identifier. Here, according to FIG. 5, the recommendation apparatus 1 further includes a reviewer evaluation level accumulation unit 121 and a reviewer evaluation degree calculation unit 122.

  The reviewer evaluation level accumulating unit 121 accumulates the evaluation level of each critic evaluator with respect to the critique comment of each item by the critic (see FIG. 4 and S64 in FIG. 6). Of course, the evaluation level may have a negative evaluation.

The reviewer evaluation degree calculation unit 122 calculates an estimated evaluation degree obtained by averaging (or accumulating, normalizing, and maximum value) the evaluation level of each item for each reviewer identifier (see S65 in FIG. 6).
Rev (R, A): Evaluation degree of app A by the critic R The normalization is a value between 0 and 1 (primary value-minimum value) / (maximum value-minimum value). Say. It is also preferable to calculate an estimated evaluation degree obtained by multiplying the evaluation degree for each critic by a weight α that is different for each critic.

  The reviewer evaluation degree accumulating unit 12 preferably accumulates a review comment acquired in advance from the SNS site server 5 and a review comment corresponding to the item. It is desirable that critical comments can also be sent to the user.

  FIG. 8 is an explanatory diagram showing the reviewer evaluation degree calculated by the reviewer evaluation degree calculation unit.

  According to FIG. 8, the estimated evaluation degree which averaged the evaluation level from each critique evaluator with respect to each item for each reviewer identifier is represented. This estimated evaluation degree is accumulated by the reviewer evaluation degree accumulation unit 12 (see S66 in FIG. 6).

[Reviewer usefulness calculator 13]
The reviewer usefulness calculation unit 13 calculates the usefulness Point (U, R) of the critic R for the user U (see S67 in FIG. 6).

  FIG. 9 is an explanatory diagram showing the usefulness of each critic for each user.

The reviewer usefulness calculation unit 13 sets the estimated evaluation degree Rev (R, A) of each item A in the reviewer identifier R to the estimated preference degree Pref (U, A) of each item in the user identifier U of the user to be recommended. Multiply for each item, and add the multiplication value of each item. For example, it is calculated by the following two exemplary formulas.
Point (U, R) = Σ (Rev (R, A) * Pref (U, A))
Point (U, R) = Σ (Rev (R, A) * Pref (U, A)) / Num (R)
Num (R): Number of criticism comments left by critic R

[Reviewer selection unit 14]
The reviewer selection unit 14 selects the reviewer identifier having the highest usefulness Point (U, R) for the user identifier (see S68 in FIG. 6). Alternatively, one or more reviewer identifiers whose usefulness points (U, R) are equal to or greater than a predetermined threshold may be selected.

[Recommendation 15]
The recommendation unit 15 transmits the selected reviewer identifier to the user-operated terminal (see S69 in FIG. 6).

  Here, it is also preferable that the recommendation unit 15 obtains a criticism comment criticized for each item by the reviewer based on the selected reviewer identifier from the reviewer evaluation degree accumulation unit 12 and transmits it to the user's operation terminal. . As a result, the user can view the critics' criticism comments that suit their preferences.

  Moreover, it is also preferable that the recommendation part 15 selects the item from which the estimated preference degree becomes more than a predetermined threshold about the said user who should be recommended from the user preference accumulation | storage part 11, and transmits to the terminal 2 of user operation. At this time, it is also preferable that a reviewer based on the selected reviewer identifier obtains a review comment reviewed for the selected item from the reviewer evaluation degree storage unit 12 and transmits it to the terminal 2 operated by the user. As a result, the user can know the recommended item and can also view the critic's critique comment that suits his / her preference in the item.

  Finally, a technique for collecting critical comments and reviewer evaluation levels is described. That is, how to accumulate the reviewer evaluation level in S64 of FIG. 6 will be described in detail.

  FIG. 10 is a first sequence diagram for collecting critical comments and evaluation levels.

  According to FIG. 10, the EC site server 3 has a function of collecting evaluation levels. First, a reviewer uses the terminal 4 to click a “comment” button on a desired item on the item page of the EC site server 3 to describe a comment. The critic's terminal 4 transmits a criticism comment associated with the item to the EC site server 3.

  Thereafter, the critique evaluator can use the terminal 6 to view the item in the EC site server 3 and visually recognize the critique comment. At this time, it is assumed that the critique evaluator assigns, for example, an evaluation level 4 for the critique comment. The terminal 6 of the critic evaluator transmits this evaluation level to the EC site server 3.

  Then, the EC site server 3 transmits the collected evaluation level to the recommendation device 1 based on the present invention.

  FIG. 11 is a second sequence diagram for collecting critical comments and evaluation levels.

  According to FIG. 11, the evaluation level is collected by the SNS site server 5 via the EC site server 3. First, a reviewer uses the terminal 4 to click a “comment” button on a desired item on the item page of the EC site server 3 to describe a comment. The critic's terminal 4 transmits a criticism comment associated with the item to the EC site server 3.

  Thereafter, the critique evaluator can use the terminal 6 to view the item in the EC site server 3 and visually recognize the critique comment. At this time, it is assumed that the critique evaluator clicks, for example, a “Like” button on the critique comment. The terminal 6 of the critic evaluator transmits this evaluation level to the EC site server 3.

  The EC site server 3 transfers the critical comment to the SNS site server 5 of, for example, facebook (registered trademark) corresponding to the “Like!” Button. As a result, the SNS site server 5 publishes the transferred critical comment in the group.

  Here, the EC site server 3 counts the number of critical comments transferred for the item, and sets the number as the evaluation level. The EC site server 3 transmits the collected evaluation level to the recommendation device 1 according to the present invention.

  FIG. 12 is a third sequence diagram for collecting critical comments and evaluation levels.

  According to FIG. 12, the evaluation level is collected by the SNS site server 5. First, the critic uses the terminal 4 and clicks the “Comment SNS” button on a desired item on the item page of the EC site server 3 to write a criticism comment. At this time, the critic describes a hash tag (for example, # ... in twitter (registered trademark)) in addition to the criticism comment. It is also desirable that this hash tag is automatically assigned to the posting form in advance when the button is clicked. The critic's terminal 4 transmits a criticism comment associated with the item to the EC site server 3.

  In response to this, the EC site server 3 transfers the critical comment to which the hash tag is assigned to the SNS site server 5. Thereby, the SNS site server 5 publishes the critical comment together with the hash tag in the group.

  Thereafter, the critique evaluator can browse the critique comments on the SNS site server 5 using the terminal 6. At this time, it is assumed that the critique evaluator clicks, for example, a “Retweet” button (or “favorite”) for the critique comment. The terminal 6 of the critique evaluator transmits information of this “Retweet” button to the SNS site server 5.

  If the evaluation level of “Retweet” is “+1”, the SNS site server 5 counts the number. And the SNS site server 5 transmits an evaluation level to the recommendation apparatus 1 based on this invention. In addition, about the recommendation apparatus 1, the comment comment of the SNS site server 5 can be searched using a hash tag, the number of search hits can be counted, and evaluation levels can also be collected.

  As described above, the recommendation apparatus 1 can receive the evaluation levels collected by various sequences as shown in FIGS.

  As described above in detail, according to the recommendation program, the apparatus, and the method of the present invention, while a comment is posted by a large number of reviewers on a product or the like (item), the user likes it. It is possible to recommend to the user a critic who has left or commented on an item that is presumed to be preferred or who has a preference, and has obtained a high evaluation in the criticism comment. Thereby, the user can know a critic who has a high evaluation and suits his / her preference. According to the present invention, ternary information of user preference information, a critic's critique comment on an item, and a critique evaluator's evaluation level for the critique comment is used. Accordingly, it is possible to recommend a critic who meets the user's preference instead of simply recommending an item to the user.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Recommendation apparatus 10 Communication interface 11 User preference storage part 111 User preference information storage part 112 Collaborative filtering part 12 Reviewer evaluation degree storage part 121 Reviewer evaluation level storage part 122 Reviewer evaluation degree calculation part 13 Reviewer usefulness calculation part 14 Reviewer selection part 15 Recommendation section 2 User terminal 3 EC site server 4 Critic reviewer terminal 5 SNS site server 6 Critic reviewer terminal

Claims (8)

In a recommendation program that causes a computer mounted on a device to function to recommend other items based on item usage history information in a user,
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, collaborative filtering means for calculating an estimated preference degree of each item by collaborative filtering for user preference information based on the usage history or evaluation history of each item;
For each user identifier, and user preference storage means that accumulates the estimated degree of preference for each item,
For each reviewer identifier (critic), a reviewer rating calculation means for calculating an estimated rating by averaging, accumulating, normalizing or maximizing the evaluation level of each item of each rater identifier (critic evaluator);
The reviewers identifier (critic's) each, and reviewers evaluation degree storing means stores the previous SL estimation evaluation of each item,
Reviewer usefulness calculating means for calculating , for each item, a sum of multiplication values of the estimated preference level in the user identifier to be recommended and the estimated evaluation level in the reviewer identifier as a “usefulness level” ;
Reviewer selecting means for selecting a reviewer identifier (critic) based on at least the highest usefulness for the user identifier;
A recommendation program for causing a computer to function as recommendation means for transmitting a selected reviewer identifier (a critic) to a user-operated terminal. The reviewer evaluation degree accumulating means further accumulates the review comment corresponding to the reviewer identifier and the item,
The recommendation means acquires a criticism comment criticized for each item by the reviewer based on the selected reviewer identifier from the reviewer evaluation degree accumulation means, and transmits it to the user's operation terminal. The recommendation program according to claim 1. The recommendation means is:
From the user preference storage means, an item whose estimated preference is equal to or greater than a predetermined threshold for the user is selected and transmitted to a user-operated terminal.
A reviewer based on the selected reviewer identifier acquires a review comment reviewed for the selected item from the reviewer rating storage means or another server, and transmits the comment to a user operation terminal. The recommendation program according to claim 2. Regarding the user preference information storage means, the usage history of the item is based on whether or not the user based on the user identifier has purchased or used a product service as an item, or has been installed on an owned terminal using an application as an item The recommendation program according to any one of claims 1 to 3 , wherein the computer is further made to function so as to be stored. Valuation determines critique's in the reviewers valuation storage means according to claim 4, characterized in that the computer is further caused to function as is obtained by storing one or more stages of level representing the impression for criticism Comments Recommended program. The critic's critique comment based on the reviewer identifier is a comment text posted on an SNS (Social Networking Service) site or a blog (Web log) site,
The evaluation level determined by the critics is “Like!”, “+1”, “Retweet”, “Bookmark”, “Good / bad evaluation”, “Not helpful / not helpful” for the comment text. The recommendation program according to any one of claims 1 to 5 , wherein the computer is caused to function at a level of "favorite" or "five-level evaluation". In the recommendation device that recommends other items based on the usage history information of items in the user,
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, collaborative filtering means for calculating an estimated preference degree of each item by collaborative filtering for user preference information based on the usage history or evaluation history of each item;
For each user identifier, and user preference storage means that accumulates the estimated degree of preference for each item,
For each reviewer identifier (critic), a reviewer rating calculation means for calculating an estimated rating by averaging, accumulating, normalizing or maximizing the evaluation level of each item of each rater identifier (critic evaluator);
The reviewer identifier (critic's) each, and reviewers evaluation degree storing means stores the previous SL estimation evaluation of each item,
Reviewer usefulness calculating means for calculating , for each item, a sum of multiplication values of the estimated preference level in the user identifier to be recommended and the estimated evaluation level in the reviewer identifier as a “usefulness level” ;
Reviewer selecting means for selecting a reviewer identifier (critic) based on at least the highest usefulness for the user identifier;
A recommendation device comprising: a recommendation means for transmitting a selected reviewer identifier (a critic) to a user-operated terminal. In the recommendation method of recommending other items based on the usage history of items in the user using the device,
The device is
The reviewer identifier is the identifier of the “critic” who posted a critical comment on each item,
A rater identifier is an identifier of a “critic evaluator” that assigns an evaluation level to the critique comment of each item criticized by the reviewer (reviewer identifier).
For each user identifier, by collaborative filtering for user preference information based on the usage log or evaluation history of each item, a user preference degree storage unit that calculates an estimated degree of preference for each item, and storing said estimated degree of preference for each item,
Reviewers identifier (critic's) for each, averaging the evaluation level of each item of each rater identifier (critic evaluator), calculated cumulative reduction, normalization or the maximum value of the estimated evaluation degree, before Symbol estimation of each item A reviewer evaluation degree accumulating unit for accumulating the evaluation degree;
The device is
A first step of calculating , for each item, a sum of multiplication values of an estimated preference level in a user identifier to be recommended and an estimated evaluation level in a reviewer identifier as a “usefulness level” ;
A second step of selecting a reviewer identifier (a critic) based on at least the highest usefulness for the user identifier;
A recommendation method comprising: a third step of transmitting the selected reviewer identifier (critic) to a user-operated terminal.

Images