JP4699337B2 - RECOMMENDED INFORMATION DETECTION PROVIDING DEVICE, RECOMMENDED INFORMATION DETECTION PROVIDING METHOD, AND PROGRAM THEREOF - Google Patents

Description

  The present invention provides recommended information detection that can provide, as recommended information, information that is highly likely to be of interest to the user and that belongs to a different concept other than the known concept (this is innovative information). The present invention relates to a device, a recommendation information detection providing method, and a program thereof.

Non-Patent Document 1 discloses a technique for profiling information indicating user interest and extracting a user set of close interest using the correlation of the user interest. In addition, based on the assumption that “users with close interests each take the same action”, for example, information (item information, etc.) that is interesting to other users who have a high degree of approximation with the profile of user A is used. Non-Patent Document 2 discloses a collaborative filtering technique that extracts information useful for the user A and recommends the user A information that the other user is interested in.
Makoto Nakajo, 2 others, "Blog community creation support based on user interest ontology generation", IPSJ Research Report, Information Processing Society of Japan, Vol.2006, No.59 p.77-84, May 2006 30-31 days issue Toshiyuki Masui, “City Corner Collaborative Filtering of Interface”, ASCII Co., Ltd., UNIX MAGAZINE, Vol.20 # 12, p.175-180, issued December 1, 2005

  Here, the collaborative filtering technique as described in Non-Patent Document 2 is applied to many recommendation (information recommendation) services. That is, such a collaborative filtering technique is a convenient technique in that information that does not belong to the user's interest profile can be recommended from the interest profiles of other users. However, since the above collaborative filtering technology simply makes recommendations using only information (instance information) belonging to other users' interest profiles, information on what concept the information belongs to (for example, There is no genre to which the product belongs, and in the end, information that belongs to a known concept that the user A already knows tends to recommend.

  For example, in the collaborative filtering of Non-Patent Document 2, for 6 news articles, 4 (Ken, Lee, Meg, Nan) users give 5 grades as shown in FIG. (Some articles do not have an evaluation score because they do not read all articles.) The technique for recommending the sixth article to a user named Ken is described. However, there is a problem that it is not known what concept the article No. 6 belongs to. Therefore, there is a high possibility of recommending to the user Ken that the article 6 that may belong to the same concept as the article 2 and the article 5 that Ken has already given high evaluation. It is not suitable to use technology. In another example, for example, by using collaborative filtering technology, an artist “Kinks” can be recommended to a user who is interested in the artist “Who”, for example. However, since “Who” and “Kinks” are both UK rock bands of the late 60s, the concept (concept = class) information (eg, genre) that is the background of the artist belongs to the same concept, Is information on a known concept, and “Kinks” itself belonging to the known concept is highly likely to be information known to the user to whom the recommendation information is provided (sometimes referred to as an artist: alias instance). Considering the purpose of greatly expanding the user's interest, the user is highly likely to be interested, and the user may recommend information belonging to a different concept other than a known concept (this is referred to as innovative information) Although it is considered desirable, the collaborative filtering technique of Non-Patent Document 2 often recommends information of the same concept, which is not suitable for a technique aiming at greatly expanding user interest. The technique of Non-Patent Document 1 is only a technique until the degree of approximation between ontology is calculated, and is a technique that does not lead to a user recommending information belonging to a different concept other than a known concept.

  Therefore, the present invention provides a recommended information detection / providing device that can recommend information <which is referred to as innovative information> that belongs to a different concept other than a known concept, and that the user is likely to be interested in. It is an object of the present invention to provide a recommended information detection providing method and a program thereof.

  To achieve the above object, the present invention provides an interest ontology generating means for generating an interest ontology of each user who created blog information based on a template ontology, an interest ontology of a user to whom recommended information is provided, and other Approximation degree calculating means for calculating the degree of approximation with the user's interest ontology using an approximation method between ontologies, and specifying a plurality of interest ontologies having a high value of the degree of approximation, and corresponding to the interest ontology The user set extraction means for extracting users, the extracted user's interest ontology, and the recommendation information providing target user's interest ontology are compared, and the recommended information providing target user's interest ontology belongs to First, when the user A evaluates an instance belonging to the extracted user's interest ontology, the evaluation is scheduled. Predicted value calculation means for calculating a value using a predicted value calculation formula, instance specifying means for specifying an instance from which the predicted value higher than a specified value is obtained, and the specified instance in the extracted user's interest ontology The number of classes retroactively from the hierarchical position of the class to which the class belongs to the class that goes back in the root direction in the extracted interest interest ontology hierarchy of the user and also belongs to the interest ontology of the recommended information provision target user. An recommended information detecting and providing apparatus comprising: an innovative degree calculating means for calculating an innovative degree; and a recommendation information output means for outputting an instance having a high innovative degree as recommendation information to the recommended information providing user. is there.

  The present invention is also a recommendation information detection and provision method in a recommendation information detection and provision apparatus, wherein the interest ontology generation means of the recommendation information detection and provision apparatus creates an interest ontology of each user who has created blog information based on a template ontology. And the degree-of-approximation calculation means of the recommendation information detection and provision apparatus calculates the degree of approximation between the interest ontology of the user to whom the recommendation information is to be provided and the interest ontology of other users using an approximation measurement method between the ontology. Then, the user set extraction means of the recommended information detection / providing device identifies a plurality of interest ontology having a high value of the approximation, extracts the user corresponding to the interest ontology, and predicts the recommended information detecting / providing device The value calculation means compares the extracted user's interest ontology with the recommendation information providing target user's interest ontology, and When the user A evaluates an instance that does not belong to the interest ontology of the information provision target user and belongs to the extracted user interest ontology, a predicted value of the evaluation is calculated using a predicted value calculation formula. The instance specifying means of the recommended information detection / providing device specifies an instance where the predicted value higher than a prescribed value is obtained, and the innovative degree calculating means of the recommended information detection / providing device is configured to extract the user's interest ontology From the hierarchical position of the class to which the identified instance belongs to the class that goes back in the root direction in the extracted user interest ontology hierarchy and belongs to the user interest ontology of the recommended information provision target The retroactive stage number of classes is calculated as an innovative degree, and the recommended information detection / providing device is calculated. Recommendation information output means is a recommendation information detection providing method and outputs to the recommendation information providing target user high the innovative degree instance as recommendation information.

  Further, the present invention is a program to be executed by a computer of a recommendation information detection and provision apparatus, an interest ontology generation process for generating an interest ontology for each user who created blog information based on a template ontology, and a recommended information provision target The degree of approximation between the user's interest ontology and another user's interest ontology using an approximation measurement method between the ontology, and a plurality of interest ontology with a high value of the degree of approximation The user set extraction process for extracting the user corresponding to the interest ontology, the extracted user interest ontology, and the recommendation information providing target user interest ontology are compared, and the recommended information providing target user is compared. Instances that do not belong to the interest ontology of, but belong to the extracted user interest ontology A predicted value calculation process for calculating a predicted value of the evaluation when the user A evaluates using a predicted value calculation formula; an instance specifying process for specifying an instance from which the predicted value higher than a specified value is obtained; In the extracted user's interest ontology, from the hierarchical position of the class to which the identified instance belongs, the class goes back to the root direction in the extracted user's interest ontology hierarchy, and the recommendation information providing target user's interest ontology Innovative degree calculation processing for calculating the retroactive level of the class up to the class to which it belongs as an innovative degree, and recommended information output processing for outputting the instance having the high innovative degree as recommendation information to the recommended information providing user. A program to be executed by a computer.

  According to the present invention, it is possible to provide innovative information that has a high degree of innovativeness and is likely to be of interest to each user. Therefore, when the user accepts such innovative information, the range of interest of the user jumps dramatically. Based on this interest, the user may purchase information belonging to a new concept (for example, a product of a music artist). In other words, it is possible to give unexpected “awareness” to users who tend to stick to one concept of interest. For example, if the recommended information is a product, the opportunity for purchasing the product obtained from the user will be dramatically expanded. The possibility of being able to do so will be obtained. In addition, by extracting such innovative information with a high degree of innovation and recommending it to the user, the user's interest can be naturally expanded and applied to services such as community formation.

  Hereinafter, an innovative information detection providing device (recommended information detection providing device) according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the innovative information detection and provision apparatus according to the embodiment. As shown in the figure, reference numeral 1 denotes an innovative information detection / providing device. The innovative information detection and provision apparatus 1 includes a communication processing unit 11, a control unit 12, an interest ontology generation unit 13, a user set extraction unit 14, a difference class / instance information extraction unit 15, an innovative degree calculation unit 16, and a recommendation processing unit 17. I have. The innovative information detection and provision apparatus 1 performs the following process in order to recommend innovative information belonging to a different concept other than a known concept that is likely to be of interest to the user.

  First, the interest ontology generation unit 13 of the innovative information detection providing device 1 acquires blog information (blog entries) of a plurality of users from each blog server. The blog information is stored in a database or the like in association with the user ID of the user who created the blog. Here, the blog information is described in a markup language such as HTML. Then, the interest ontology generation unit 13 uses all the blog information stored in the database as a template ontology (information indicating the hierarchical relationship between the class and the instance to which the blog information belongs and the template). An ontology for each blog, that is, an interest ontology for each user is generated. Here, the method for generating the interest ontology is determined to classify the description word in the blog information as Ci if the name attribute (word) of a certain class Ci in the template ontology exists in the description in the blog information. If there is a name attribute of the instance Ii (εCi) belonging to Ci, the words described in the blog information are classified into class Ci-instance Ii. The blog information may be classified into a plurality of classes.

  Next, the interest ontology generation unit 13 calculates the number of users considered to be interested in the name attribute (or word attribute) of the instance belonging to the lowest layer class Cr forming the template ontology. When calculating the number of users who are interested in instances belonging to the class Cr, the number of users is calculated as 1, even if the same user describes the same instance in a plurality of blog information. Next, the interest ontology generation unit 13 similarly calculates the number of users who are interested in the name attribute of the lowermost class Cr. The calculation of the number of users who are interested in the name attribute of the lowermost class Cr is performed by calculating the number of users interested in the name attribute of all instances belonging to the lowermost class Cr and the number of users interested in the name attribute of the lowermost class Cr. And the sum of At this time, the number of users is set to 1 even when the same user is interested in a plurality of instances or is interested in the lowermost class and instances belonging to the lowermost class at the same time.

  The number of users who are interested in the name attribute of the instance and the name attribute of the class is recursively increased to the root class in the hierarchical relationship. Thereby, the number of users who are interested in the whole (domain) of the hierarchical relationship and its distribution can be derived. That is, the interest ontology generation unit 13 counts the number of users who are interested in the overall hierarchy of the template ontology, the number of users interested in the name attributes of each class and instance in the hierarchy, and the name attributes of each class and instance. The ID of the user who is interested in is identified and recorded in a server or the like.

  Next, the interest ontology generation unit 13 uses the user ID of the number of users who are interested in the entire hierarchical relationship (domain) and the distribution information thereof, and uses the user ID to identify the class and instance that the user of the user ID is interested in. Generate hierarchical relationship information, that is, an interest ontology for one user. Then, an interest ontology for each user is generated using all user IDs registered in the server or the like, and stored in a database or the like.

FIG. 2 is a diagram showing an overview of the user A's interest ontology and the user B's interest ontology. As shown in this figure, the interest ontology of User A has the word class “Rcok” as the root, and the word classes “Alternative”, “Madchester”, and “Shoegaze” at the lower layer of the root, and “Madchester” An instance of the words “New Order” and “Happy Mondays” belongs to the class, and an instance of the words “Coldplay”, “My Bloody Valentine” and “Galaxie500” belongs to the class “Shoegaze”.
User B's interest ontology is rooted in the word class “Rcok”, and “Madchester” and “Shoegaze” classes belong to the “Alternative” class below that route, and “US Inside” below the route. "Athens" and "Elephant 6" classes belong to the class, and the instances of the words "Happy Mondays" and "Stone Roses" belong to the "Madchester" class, and "Coldplay", " An instance of the word “My Bloody Valentine” belongs, an instance of the word “REM” belongs to the class “Athens”, and an instance of the words “Elf Power” and “Olivia Tremor Control” belongs to the class “Elephant 6” Ontology. It is assumed that the words of each class and instance shown in FIG. 2 are music artist names and music words, and these words are described in a blog created by user A or user B.

Next, the user set extraction unit 14 has an interest similar to that of the user A by using an interest ontology of one user A, an interest ontology of another user, and an approximation measurement method between the ontology. A user set Gu composed of X users is calculated. Here, the interest ontology of the user A and O _A, other users, for example, the interest ontology of the user B and O _B. In the hierarchical relationship between classes and instances of interest ontology, the topology consisting of parent class and small class is T ₁ (one-level hierarchy of parent class and child class), and the topology consisting of class and instances belonging to that class is T ₂ . Further common classes of both interests ontology interest ontology O _A and interest ontology O _B is defined as C _i, the common instance is defined as I _i ∈I. In particular, a class set belonging to topology T ₁ is C (T ₁ ), and a class set belonging to topology T ₂ is C (T ₂ ). Further, the interest coincidence for the common class is I (C _i ), the interest coincidence for the common instance is I (I _i ), and the interest coincidence under the common class is I _t (C _i ).

Specifically, the user set Gu is calculated by first analyzing the common class of the interest ontology O _A and the interest ontology O _{B by} the user set extracting unit 14, the common class forming the topology T ₁ , and the topology extracting a common class of forming a T _2. Here, there is a common class C ₁ that forms both T ₁ and T ₂ topologies. SUMMARY view of similarity calculation between ontology 3, the common class a1, b1, b2 of interest ontology _{O A} and interest ontology _{O B} forms a topology _{T 1} in interest ontology _{O A} and interest ontology _{O B,} b2 , b3, c4 forms a topology _{T 2} in interest ontology _{O a} and interest ontology _{O B.}

Next, assuming that the common class C _i has a common instance I _i between both interest ontologies, the user set extraction unit 14 selects the smaller value of the interest degrees for the instances I _i of both ontologies and The degree of interest I (I _i ) for the instance I _i is calculated. For example, in FIG. 3, the degree of interest coincidence with respect to the common instance a belonging to the common class c3 is 2. Similarly, the smaller value of the degrees of interest of both the ontology O _A and O _B with respect to the class C _i is calculated as the interest coincidence degree I (C _i ) regarding the common class in both the ontology O _A and O _B. For example, the interest coincidence for the common class C3 is zero.

Next, the user set extraction unit 14, the intersection of the child class belonging to a common class _{C i} in _{O A} and _{O B} and N _{(C i),} also the sum of the child class which belong to a common class _{C i} in OA and OB With the set as U (C _i ), the interest coincidence I _t (C _i ) under the class C _i ∈ C (T ₁ ) is

Calculated by For example, the interest coincidence for the subordinate to the common class b2 is (9 + 3) / 2 = 6. Then, the value obtained by adding the interest coincidence I _t (C _i ) to all common classes forming the topology T ₁

Is calculated as the interest coincidence S (T ₁ ) for the common class set forming the topology T ₁ . On the other hand, for the common class C _i ∈ C (T ₂ ), the instance set under C _i in O _A is I _A (C _i ), and the instance set under C _i in O _B is I _B (C _i ). , C _i interest level I _t (C _i )

Calculate with For example, the interest coincidence with the common class c3 is ((2 + 0 + 3 + 0) / 4) = 5/4. The interest matching score I _{t (C i)} yelling added in all common classes of forming a topology T ₂ value

Is the interest matching degree S (T ₂ ) for the common class set forming the topology T ₂ . Then, S _{(T 1)} and S _(T 2), using the evaluation function f (X) corresponding to the importance for both topology, between two interest ontology _O A, the degree of approximation So of _{O B} winding,

Calculated by Then, the degree of approximation So between the above-mentioned user A and other users' interest ontology is calculated for every combination of all users A and other users, and a user set Gu composed of X users with similar degrees of approximation So is calculated. . X is an experientially set variable for recommending innovative information to the user A. In addition, the above-described method for measuring the degree of approximation between interest ontology is not limited to the above-described method, but may be replaced with a method for selecting users with close interest using the correlation coefficient in the collaborative filtering technique shown in Non-Patent Document 2. It may be.

  However, since the method of the above-described approximation degree measurement method calculates the approximate user in consideration of the conceptual information that is the background of the instance called the class, the recommendation accuracy is increased. Also, as described above, by extracting X users who are interested in user A first, it is possible to calculate a highly accurate value in the next extraction process of difference class / instance information. For example, in the collaborative filtering technique of Non-Patent Document 2, when another user gives an extremely high evaluation value to an article, the high evaluation value of the user who has not evaluated the article is high. Pulled by the evaluation of the user who gave the evaluation value, there is a problem that the evaluation value of the user who has not evaluated the article is highly estimated, this apologizes to the user who has not evaluated the article There is a possibility of providing recommended information. Therefore, such a problem can be avoided by narrowing down the parameter when collaborative filtering is calculated for X users who are interested in advance.

  Next, the difference class / instance information extraction unit 15 compares the information of the user A's interest ontology with the information of the user's interest ontology belonging to the user set Gu, and extracts the information of the difference class and the difference instance. And accumulate in the database. Specifically, first, the user A previously gives an evaluation value to a class or instance in his / her interest ontology. Each of the X users included in the user set Gu also gives evaluation values to classes and instances in his / her interest ontology. Assume that these evaluation values are registered in a server or the like.

Here, a correlation coefficient between evaluation value vectors based on evaluation values given to each class or instance of the interest ontology is used as the similarity between users. The correlation coefficient can be calculated by dividing the covariance of two evaluation value vectors by the product of the standard deviation of each vector. For example, the evaluation value vectors of the respective users are determined from the evaluation values of the users A and B. These correlation coefficients γ _AB can be calculated as shown in Equation (6). Here, σ _A and σ _B are standard deviations, A ′ and B ′ are averages of evaluation values given to each class and instance of interest ontology by user A and user B, respectively, and Cov (A, B) is covariance. .

  Further, when a user A evaluates a class or an instance that does not belong to the user A's interest ontology and belongs to the user B's interest ontology, the predicted value of the evaluation is calculated by the equation (7).

  Note that an approximation measurement function between ontology may be applied to the correlation coefficient. However, the use of the approximation measurement function between the ontology has an advantage that the recommendation accuracy is higher because the approximate user can be selected in consideration of the concept information that is the background of the instance called the class.

  And the innovative degree calculation part 16 calculates the distance of an instance of a predicted value higher than a certain threshold among the calculated predicted values and the interest ontology of the user A as an innovative degree. Specifically, the method of calculating the degree of innovativeness is based on the interest ontology of the user B from the hierarchical position of the class to which the instance having a high predicted value belongs among the instances when the user A evaluates in the interest ontology of the user B. The retroactive number of classes up to the class that goes back in the root direction in the hierarchy and belongs to the user A's interest ontology is calculated as the innovative degree.

  In other words, as shown in FIG. 2, user A's interest ontology has "Rock" as the root, and there is an "Alternative" class below it, and user B's interest ontology has "Rock" as the root. Below that there are “Alternative” and “US Indie” classes. And if the predicted value of user A's evaluation for the instance “Elf Power” under the class “Alternative” → “US Indie” → “Athens” → “Elephant6” in User B ’s interest ontology is high, The innovative level of the instance “Elf Power” is a class in the root direction in the hierarchy of the user B ’s interest ontology from the class “Elephant6” to which the instance “Elf Power” belongs, and also belongs to the user ’s interest ontology. Since the number of retroactive classes up to a certain class is “Elephant6” → “Athens” → “US Indie” → “Alternative”, it is “3”.

  As described above, the innovative degree calculation unit 16 calculates the innovative degree of the instance having a high predicted value among the instances when the user A evaluates in the interest ontology of the user B. Moreover, the innovative degree calculation part 16 calculates the innovative degree of the instance with a high predicted value among the instances when the user A evaluates in the interest ontology of each of the X users included in the user set Gu.

Here, the degree of innovativeness increases as the number of hierarchical levels from the position of the same class as the class included in the user A's interest ontology to the class to which the instance with the high predicted value belongs is deeper in the leaf node direction of the hierarchy. . Therefore, the higher the value of the innovative degree, the farther from the concept that the user A is known. However, since the instance has a high predicted value when the user A evaluates, the instance is far from the known concept of the user A, but the user A is likely to be interested. Accordingly, the recommendation processing unit 17 selects and extracts the number of instances having a high value of the innovative level from among the innovative levels calculated by the innovative level calculation unit 16 in advance, and extracts the innovative information for the user A. as (recommendation information) G _I, outputs with innovative degree of that instance. By recommending an instance having a high degree of innovativeness as innovative information, the user A can greatly expand his / her own interest range based on the information on the instance. Conversely, when an instance having a low innovative degree is recommended as innovative information, the user A can incorporate the instance information into his own interest relatively easily.

  As described above, the embodiment of the present invention has been described. However, since it is possible to provide innovative information with a high degree of innovativeness that may be of interest to each user, when the user accepts such innovative information, the user's interest range As a result, the user may purchase information belonging to a new concept (for example, a product of a music artist). In other words, it is possible to give unexpected “awareness” to users who tend to stick to one concept of interest. For example, if the recommended information is a product, the opportunity for purchasing the product obtained from the user will be dramatically expanded. The possibility of being able to do so will be obtained. In addition, by extracting such innovative information with a high degree of innovation and recommending it to the user, the user's interest can be naturally expanded and applied to services such as community formation.

  The above-described innovative information detection / providing device has a computer system therein. The process described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by the computer reading and executing this program. Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

  The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

It is a block diagram which shows the structure of an innovative information detection provision apparatus. It is a figure which shows the outline | summary of the interest ontology of the user A and the user B. FIG. It is a schematic diagram of the approximation degree calculation between ontologies. It is a figure which shows the example which provides the recommendation information using the conventional collaborative filtering.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Innovative information detection provision apparatus 11 ... Communication processing part 12 ... Control part 13 ... Interest ontology production | generation part 14 ... User set extraction part 15 ... Difference class and instance information extraction part 16 ... Innovative degree calculation unit 17 ... Recommendation processing unit

Claims (3)

An interest ontology generating means for generating an interest ontology of each user who created the blog information based on the template ontology;
An approximation calculating means for calculating the degree of approximation between the ontology of interest of the user to whom the recommendation information is provided and the interest ontology of other users using an approximation measurement method between the ontology;
User set extraction means for identifying a plurality of interest ontologies having a high value of the approximation and extracting the users corresponding to the interest ontologies;
Comparing the extracted user's interest ontology with the recommendation information providing target user's interest ontology, the extracted user's interest ontology does not belong to the recommended information providing target user's interest ontology. A predicted value calculation unit that calculates a predicted value of the evaluation when the user A evaluates an instance to which the user A belongs using a predicted value calculation formula;
An instance specifying means for specifying an instance where the predicted value higher than a specified value is obtained;
In the extracted user's interest ontology, from the hierarchical position of the class to which the identified instance belongs, the class goes back to the root direction in the extracted user's interest ontology hierarchy, and the recommendation information providing target user's interest ontology An innovative degree calculating means for calculating the retroactive number of classes up to the class to which it belongs as an innovative degree;
A recommendation information output means for outputting an instance having a high degree of innovativeness as recommendation information to the user of the recommendation information provision target;
An apparatus for providing and providing recommended information, comprising: A recommended information detection / providing method in a recommended information detection / providing device,
The interest ontology generating means of the recommended information detection providing device generates an interest ontology for each user who created the blog information based on the template ontology,
The approximation degree calculation means of the recommended information detection and provision apparatus calculates the degree of approximation between the interest ontology of the user to whom the recommendation information is to be provided and the interest ontology of other users by using an approximation measurement method between ontologies,
The user set extraction means of the recommended information detection providing device specifies a plurality of interest ontology having a high value of the approximation, and extracts the user corresponding to the interest ontology,
The predicted value calculation means of the recommended information detection / providing device compares the extracted user's interest ontology with the interest information ontology of the recommended information providing target user, When a user A evaluates an instance that does not belong and belongs to the extracted user's interest ontology, a predicted value of the evaluation is calculated using a predicted value calculation formula,
The instance specifying means of the recommended information detection providing device specifies an instance where the predicted value higher than a specified value is obtained,
Innovative degree calculation means of the recommended information detection providing device is a class that goes back in the root direction in the hierarchy of the extracted user interest ontology from the hierarchical position of the class to which the identified instance belongs in the extracted user interest ontology. And calculating the retroactive number of classes up to the class belonging to the interest ontology of the user to whom the recommendation information is provided as the innovative degree,
A recommendation information detection and provision method, wherein the recommendation information output means of the recommendation information detection and provision apparatus outputs the instance having a high degree of innovativeness as recommendation information to the user who is the provision target of the recommendation information. A program to be executed by the computer of the recommended information detection / providing device,
An interest ontology generation process for generating an interest ontology for each user who created the blog information based on the template ontology;
Approximation degree calculation processing for calculating the degree of approximation between an interest ontology of a user to whom recommendation information is provided and an interest ontology of another user using an approximation measurement method between ontologies;
A user set extraction process for identifying a plurality of interest ontologies having a high value of the approximation and extracting the users corresponding to the interest ontologies;
Comparing the extracted user's interest ontology with the recommendation information providing target user's interest ontology, the extracted user's interest ontology does not belong to the recommended information providing target user's interest ontology. A predicted value calculation process for calculating a predicted value of the evaluation when the user A evaluates an instance to which the user A belongs using a predicted value calculation formula;
An instance identification process for identifying an instance for which the predicted value higher than a specified value is obtained;
In the extracted user's interest ontology, from the hierarchical position of the class to which the identified instance belongs, the class goes back to the root direction in the extracted user's interest ontology hierarchy, and the recommendation information providing target user's interest ontology An innovative degree calculation process for calculating the retroactive number of classes up to the class to which the
A recommendation information output process for outputting an instance having a high degree of innovativeness as recommendation information to the user of the recommendation information provision target;
A program that causes a computer to execute.

Images