U.S. patent application number 14/155155 was filed with the patent office on 2014-05-08 for item recommendation system, item recommendation method and program.
This patent application is currently assigned to Sony Corporation. The applicant listed for this patent is Sony Corporation. Invention is credited to Takehiro HAGIWARA, Naoki KAMIMAEDA, Katsuyoshi KANEMOTO, Masanori MIYAHARA.
Application Number | 20140129376 14/155155 |
Document ID | / |
Family ID | 45096990 |
Filed Date | 2014-05-08 |
United States Patent
Application |
20140129376 |
Kind Code |
A1 |
KAMIMAEDA; Naoki ; et
al. |
May 8, 2014 |
ITEM RECOMMENDATION SYSTEM, ITEM RECOMMENDATION METHOD AND
PROGRAM
Abstract
An item recommendation system includes: a feedback receiving
unit receiving feedbacks with respect to items by users; a feedback
history storage unit storing information of feedbacks received at
present or in the past by the feedback receiving unit in units of
combinations of users and items; a relevance calculation unit
calculating relevance between items base on first information
indicating the degree in which the same user gives feedbacks to
respective items and second information indicating the degree in
which many users give feedbacks by using feedback information
stored by the feedback history storage unit when receiving the
feedback by the feedback receiving unit; and a recommendation unit
immediately recommending items having higher relevance with respect
to the item to the user who has given the feedback to the item by
the feedback receiving unit based on the relevance calculated by
the relevance calculation unit.
Inventors: |
KAMIMAEDA; Naoki; (Kanagawa,
JP) ; KANEMOTO; Katsuyoshi; (Chiba, JP) ;
MIYAHARA; Masanori; (Tokyo, JP) ; HAGIWARA;
Takehiro; (Tokyo, JP) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Sony Corporation |
Tokyo |
|
JP |
|
|
Assignee: |
Sony Corporation
Tokyo
JP
|
Family ID: |
45096990 |
Appl. No.: |
14/155155 |
Filed: |
January 14, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13154305 |
Jun 6, 2011 |
8666834 |
|
|
14155155 |
|
|
|
|
Current U.S.
Class: |
705/26.7 |
Current CPC
Class: |
G06Q 30/0631 20130101;
G06Q 30/02 20130101 |
Class at
Publication: |
705/26.7 |
International
Class: |
G06Q 30/06 20060101
G06Q030/06 |
Foreign Application Data
Date |
Code |
Application Number |
Jun 15, 2010 |
JP |
2010-135655 |
Claims
1.-8. (canceled)
9. An item recommendation method comprising: receiving, by one or
more processors, feedback for a plurality of items from a plurality
of users; recommending, by the one or more processors, at least one
item to a user based on first information and second information,
the first information indicating feedback given by the user for at
least one item and second information indicating past feedback
given by at one or more other users to a plurality of items.
10. The recommendation method of claim 9, wherein feedback includes
at least one of purchase history, evaluation values, or access
history of detailed information about an item.
11. The recommendation method of claim 9, further comprising:
determining, by the one or more processors, a user taste by
analyzing the feedback from the user; extracting, by the one or
more processors, an item group comprising a subset of the plurality
of items relevant to the feedback received for the at least one
item; and generating, by the one or more processors, a
recommendation list based on the user taste comprising one or more
items from the item group, wherein the at least one item
recommended in the recommending step is selected from the
recommendation list.
12. The recommendation method of claim 11, wherein determining user
taste further comprises assigning scores to items for which
feedbacks for one or more items are received, the scores based on
purchase frequency of the one or more items, evaluation value for
the one or more items, and access frequency of detailed information
about the one or more items.
13. An item recommendation system comprising: a memory unit having
instructions stored thereon; one or more processors configured to
execute the stored instructions, which cause the one or more
processors to: receive feedback for a plurality of items from a
plurality of users; recommend at least one item to a user based on
first information and second information, the first information
indicating feedback given by the user for at least one item and
second information indicating past feedback given by at least one
another user to a plurality of items.
14. The recommendation system of claim 13, wherein feedback
includes at least one of purchase history, evaluation values, or
access history of detailed information about an item.
15. The recommendation system of claim 13, wherein the one or more
processors are further configured to: determine a user taste by
analyzing the feedback from the user; extract an item group
comprising a subset of the plurality of items relevant to the
feedback received for the at least one item; and generate a
recommendation list based on the user taste comprising one or more
items from the item group, wherein the at least one recommended
item is selected from the recommendation list.
16. The recommendation system of claim 15, wherein the one or more
processors determine user taste by assigning scores to items for
which feedbacks for one or more items are received, the scores
based on purchase frequency of the one or more items, evaluation
value for the one or more items, and access frequency of detailed
information about the one or more items.
17. A non-transitory computer readable medium having stored thereon
instructions which when executed cause one or more processors to
perform a recommendation method comprising: receiving feedback for
a plurality of items from a plurality of users; recommending at
least one item to a user based on first information and second
information, the first information indicating feedback given by the
user for at least one item and second information indicating past
feedback given by at least one another user to a plurality of
items.
18. The recommendation method of claim 17, wherein feedback
includes at least one of purchase history, evaluation values, or
access history of detailed information about an item.
19. The recommendation method of claim 17, further comprising:
determining, by the one or more processors, a user taste by
analyzing the feedback from the user; extracting, by the one or
more processors, an item group comprising a subset of the plurality
of items relevant to the feedback received for the at least one
item; and generating, by the one or more processors, a
recommendation list based on the user taste comprising one or more
items from the item group, wherein the at least one item
recommended in the recommending step is selected from the
recommendation list.
20. The recommendation method of claim 19, wherein determining user
taste further comprises assigning scores to items for which
feedbacks for one or more items are received, the scores based on
purchase frequency of the one or more items, evaluation value for
the one or more items, and access frequency of detailed information
about the one or more items.
Description
FIELD
[0001] The present disclosure relates to an item recommendation
system, an item recommendation method and a program.
BACKGROUND
[0002] In recent years, businesses using networks are rapidly
expanded. For example, a system of purchasing items at on-line
stores on the network is widely used. In most of such on-line
stores, a system of recommending items to users is provided. For
example, when the user accesses to detailed information of a
certain item, information of items relating to the item are
presented as related items or recommended items to the user. Such
system is realized by, for example, using a cooperative filtering
method described in JP-A-2003-167901 (Patent Document 1). The
cooperative filtering method is a recommendation method using
information (for example, purchase history and so on) of users
having similar taste.
SUMMARY
[0003] When using the above cooperative filtering method, it is
possible to recommend items to a new user which has no purchase
history by using information of users having similar taste to the
user. On the other hand, in a case of a contact-based filtering
method which recommends items by using purchase history of the
user, it is difficult to recommend items to the new user who has no
purchase history. Accordingly, the cooperative filtering method is
effective with respect to the user who has little purchase history
of items.
[0004] The content-based filtering method is a method of
determining items to be recommended by analyzing purchase history
of respective users, therefore, the method has an advantage that
tastes of respective users can be reflected easily. In a state
where the number of users is small and users having similar taste
are difficult to be found, the content-based filtering method is
more efficient than the cooperative filtering method.
[0005] At the on-line store, the purchase history of items, access
history of detailed information and so on are analyzed by using the
above filtering methods and so on, thereby preparing information
for recommending items to the user. However, sufficiently long time
is necessary for analyzing processing using the above filtering
methods and so on. On the other hand, recommendation results of
items have large effect on sales in the on-line store, therefore, a
system of presenting items corresponding to the user's taste
immediately is requested.
[0006] However, in present on-line stores, a system in which
purchase history and access history of detailed information are
stored for a given period of time (for example, one day) and
analysis processing by using the above filtering methods and so on
is executed at given intervals (hereinafter referred to as batch
processing) is used. Accordingly, when the user purchases an item,
the purchase history is not reflected on the recommendation result
at the time of purchasing. For example, when a user A purchases an
item X, the purchase history of the user A is not reflected on the
recommendation result with respect to a user B purchasing the item
X until the batch processing is completed.
[0007] The example in which purchase history obtained when the user
purchases items at on-line stores is used for recommendation is
shown as the above, however, evaluation results obtained when the
user evaluates a certain item, access history obtained when user
accesses to detailed information of a certain item are also used
for recommendation of items. In the following explanation, actions
(for example, purchase, evaluation, information access and so on)
of the user performed with respect to items or services
(hereinafter referred to as items) are referred to as
feedbacks.
[0008] In view of the above, it is desirable to provide a novel and
improved item recommendation system, an item recommendation method
and a program capable of realizing recommendation reflecting a
feedback of the user immediately.
[0009] According to an embodiment of the present disclosure, there
is provided an item recommendation system including a feedback
receiving unit receiving feedbacks with respect to items by users,
a feedback history storage unit storing information of feedbacks
received at present or in the past by the feedback receiving unit
in units of combinations of users and items, a relevance
calculation unit calculating relevance between items base on first
information indicating the degree in which the same user gives
feedbacks to respective items and second information indicating the
degree in which many users give feedbacks by using feedback
information stored by the feedback history storage unit when
receiving the feedback by the feedback receiving unit, and a
recommendation unit immediately recommending items having higher
relevance with respect to the item to the user who has given the
feedback to the item by the feedback receiving unit based on the
relevance calculated by the relevance calculation unit.
[0010] The recommendation system may further include a variation
detection unit detecting a combination of items i, j, the relevance
of which varies to the highest degree between a time t.sub.1 and a
time t.sub.2 by comparing a relevance J (t.sub.1) between items
calculated at the time t.sub.1 by the relevance calculation unit
and a relevance J (t.sub.2) between items calculated by the
relevance calculation unit at the time t.sub.2 which is the time
next to the time t.sub.1 and a relevance prediction unit predicting
a relevance J (t.sub.3) between items at a time t.sub.3 next to the
time t.sub.2 from a difference between relevances J (t.sub.2)-J
(t.sub.1) concerning items i, j detected by the variation detection
unit. In this case, the recommendation unit recommends items based
on the relevance calculated by the relevance calculation unit at a
given timing and recommends items based on the relevance predicted
by the relevance prediction unit at a timing different from the
given timing.
[0011] The recommendation system can further include a taste
information generation unit generating taste information indicating
user's taste based on feedback information stored by the feedback
history storage unit. In this case, the taste information
generation unit generates taste information based on feedback
information stored by the feedback history storage unit just before
receiving a feedback by the feedback receiving unit, and the
recommendation unit recommends items in which the taste is higher
for the user who has given the feedback to the item by the feedback
receiving unit from items having higher relevance based on taste
information generated just before receiving the feedback by the
feedback receiving unit.
[0012] When a priority coefficient is set to each item, the
recommendation unit can be configured to recommend items having
higher relevance with respect to the item to the user who has given
the feedback to the item by the feedback receiving unit based on
relevance calculated by multiplying the priority coefficient.
[0013] The feedback history storage unit can be configured to store
the feedback history in a storage device connected through a
network.
[0014] When the feedback information stored by the feedback history
storage unit is P(i, u) at the time of receiving the feedback by a
user u with respect to an item i by the feedback receiving unit,
and when P (i, u)=1 is set in the case of receiving the feedback
and P (i, u)=0 is set in the case of not receiving the feedback, a
relevance J (i, j) between the item i and an item j (j.noteq.1) is
expressed by the later-described expressions (1), (2).
[0015] Also according to another embodiment of the present
disclosure, there is provided an item recommendation method
including receiving feedbacks with respect to items by users,
storing information of feedbacks received at present or in the past
by the feedback receiving unit in units of combinations of users
and items, calculating relevance between items base on first
information indicating the degree in which the same user gives
feedbacks to respective items and second information indicating the
degree in which many users give feedbacks by using feedback
information stored by the step of storing feedback history when
receiving the feedback in the step of receiving the feedback and
recommending items having higher relevance with respect to the item
immediately to the user who has given the feedback to the item in
the step of receiving the feedback based on the relevance
calculated in the step of calculating relevance.
[0016] Also according to still another embodiment of the present
disclosure, there is provided a program for allowing a computer to
execute a feedback receiving function receiving feedbacks with
respect to items by users, a feedback history storage function
storing information of feedbacks received at present or in the past
by the feedback receiving function in units of combinations of
users and items, a relevance calculation function calculating
relevance between items base on first information indicating the
degree in which the same user gives feedbacks to respective items
and second information indicating the degree in which many users
give feedbacks by using feedback information stored by the feedback
history storage function when receiving the feedback by the
feedback receiving function and a recommendation function
immediately recommending items having higher relevance with respect
to the item to the user who has given the feedback to the item by
the feedback receiving function based on the relevance calculated
by the relevance calculation function.
[0017] Also according to yet another embodiment of the present
disclosure, there is provided a recording medium in which the above
program is recorded, which can be read by the computer.
[0018] As described above, according to the embodiments of the
present disclosure, it is possible to realize recommendation
immediately reflecting the feedback of the user.
BRIEF DESCRIPTION OF THE DRAWINGS
[0019] FIG. 1 is an explanatory view for explaining a configuration
of a recommendation system according to an embodiment of the
present disclosure;
[0020] FIG. 2 is an explanatory diagram for explaining a
configuration of a server device and a configuration of a storage
device included in the recommendation system according to the
embodiment;
[0021] FIG. 3 is an explanatory chart for explaining a
configuration of a history table used in the recommendation system
according to the embodiment;
[0022] FIG. 4 is an explanatory chart for explaining a
configuration of an intermediate value table used in the
recommendation system according to the embodiment;
[0023] FIG. 5 is an explanatory chart for explaining a
configuration of a relevance table used in the recommendation
system according to the embodiment;
[0024] FIG. 6 is an explanatory chart for explaining a
configuration of taste information used in the recommendation
system according to the embodiment;
[0025] FIG. 7 is an explanatory chart for explaining a
configuration of a history matrix (P-matrix) used in the
recommendation system according to the embodiment;
[0026] FIG. 8 is an explanatory chart for explaining a
configuration of an intermediate value matrix (D-matrix) used in
the recommendation system according to the embodiment;
[0027] FIG. 9 is an explanatory chart for explaining a
configuration of a relevance matrix (J-matrix) used in the
recommendation system according to the embodiment;
[0028] FIG. 10 is an explanatory chart for explaining a
recommendation method of items according to the embodiment;
[0029] FIG. 11 is an explanatory chart for specifically explaining
the recommendation method of items according to the embodiment;
[0030] FIG. 12 is an explanatory chart for specifically explaining
the recommendation method of items according to the embodiment;
[0031] FIG. 13 is an explanatory chart for explaining an update
method of the history matrix (P-matrix) according to the
embodiment;
[0032] FIG. 14 is an explanatory chart for explaining an update
method of the intermediate value matrix (D-matrix) according to the
embodiment;
[0033] FIG. 15 is an explanatory chart for explaining an update
method of the relevance matrix (J-matrix) according to the
embodiment;
[0034] FIG. 16 is an explanatory chart for explaining a recommend
method using taste information according to the embodiment;
[0035] FIG. 17 is an explanatory chart for explaining a
configuration of a table of the number of purchases used in the
recommendation system according to the embodiment;
[0036] FIG. 18 is an explanatory chart for explaining a
configuration of a table of the number of co-occurrences used in
the recommendation system according to the embodiment;
[0037] FIG. 19 is an explanatory chart for explaining an update
method of the relevance table (J-matrix) according to a first
modification example of the embodiment;
[0038] FIG. 20 is an explanatory chart for explaining the update
method of the relevance table (J-matrix) according to the first
modification example of the embodiment;
[0039] FIG. 21 is an explanatory chart for explaining the update
method of the relevance table (J-matrix) according to the first
modification example of the embodiment;
[0040] FIG. 22 is an explanatory chart for explaining an update
method of the relevance table (J-matrix) using bias parameters
according to a second modification example of the embodiment;
and
[0041] FIG. 23 is an explanatory diagram for explaining a hardware
configuration capable of realizing functions of the server device
and user terminals according to the embodiment.
DETAILED DESCRIPTION
[0042] Hereinafter, a preferred embodiment of the present
disclosure will be explained in detail with reference to the
attached drawings. In the specification and the drawings, the same
signs are given to components having substantially the same
functional configurations to omit repeated explanation.
[Flow of Explanation]
[0043] The flow of explanation concerning the embodiment of the
present disclosure described as follows will be briefly
described.
[0044] First, a configuration of a recommendation system 10
according to the embodiment will be explained with reference to
FIG. 1. Next, a configuration of a server device 11 and a
configuration of a storage device 12 included in the recommendation
system 10 of the embodiment will be explained with reference to
FIG. 2. In the explanation, configurations of a history table, an
intermediate value table, a relevance table and taste information
used in the recommendation system 10 according to the embodiment
will be also explained with reference to FIG. 3 to FIG. 6.
[0045] Next, configurations of a history matrix (P-matrix), an
intermediate value matrix (D-matrix) and a relevance matrix
(J-matrix) used in the recommendation system according to the
embodiment will be explained with reference to FIG. 7 to FIG. 9.
Additionally, a recommendation method of items according to the
embodiment using the history matrix (P-matrix), the intermediate
value matrix (D-matrix) and the relevance matrix (J-matrix) will be
explained with reference to FIG. 10.
[0046] Next, the recommendation method of items according to the
embodiment will be explained more specifically with reference to
FIG. 11 and FIG. 12. In the explanation, update methods of the
history matrix (P-matrix), the intermediate value matrix (D-matrix)
and the relevance matrix (J-matrix) and a recommendation method
using the taste information according to the embodiment will be
specifically explained with reference to FIG. 13 to FIG. 16.
Furthermore, configurations of a table of the number of purchases
and a table of the number of co-occurrences used in the
recommendation system according to the embodiment will be explained
with reference to FIG. 17, FIG. 18 and FIG. 19.
[0047] Next, an update method of the relevance matrix (J-matrix)
according to a first modification example of the embodiment will be
explained with reference to FIG. 19 to FIG. 21. Next, an update
method of the relevance matrix (J-matrix) using bias parameters
according to a second modification example of the embodiment will
be explained with reference to FIG. 22. Next, a hardware
configuration capable of realizing functions of the server device
11 and user terminals 13 according to the embodiment will be
explained with reference to FIG. 23. Lastly, technical ideas of the
embodiment are summarized, then, operation and effect obtained from
the technical ideas are briefly explained.
(Explanation Items)
1: Embodiment
[0048] 1-1: Configuration of a recommendation system 10
[0049] 1-2: Functional configurations of a server device 11 and a
storage device 12
[0050] 1-3: Recommendation method using P-matrix, D-matrix and
J-matrix
[0051] 1-3-1: Configurations of P-matrix, D-matrix and J-matrix
[0052] 1-3-2: General flow of processing
[0053] 1-3-3: Specific flow of processing
[0054] 1-3-4: Recommendation method using taste information
[0055] 1-3-5: About efficiency method
1-4: Modification examples
[0056] 1-4-1: Approximate update method of J-matrix (first
modification example)
[0057] 1-4-2: Update method of J-matrix based on bias parameters
(second modification example)
1-5: Hardware configuration
2: Brief
1: EMBODIMENT
[0058] The embodiment of the present disclosure will be explained.
The embodiment relates to the system executing recommendation
reflecting the user's taste and a trend of the whole users
immediately in consideration of effects with respect to the trend
of the whole users in addition to the user's taste when receiving a
feedback from the user.
1-1: CONFIGURATION OF A RECOMMENDATION SYSTEM 10
[0059] First, the configuration of the recommendation system 10
according to the embodiment will be explained with reference to
FIG. 1. FIG. 1 is an explanatory view for explaining the
configuration of the recommendation system 10. The system
configuration of the recommendation system 10 shown in FIG. 1 is
one example.
[0060] As shown in FIG. 1, the recommendation system 10 includes
the server device 11, the storage device 12 and plural user
terminals 13. The server device 11 is connected to the plural user
terminals 13 through a network 14. A personal computer and a
cellular phone are shown as examples of the user terminals 13 in
FIG. 1, however, the forms of the user terminals 13 are not limited
to the above and an information terminal, a game machine, a video
device, information home appliances and so on can be applied.
[0061] The server device 11 is a means for receiving a feedback of
the user and providing services for the user in accordance with the
received feedback. For example, the server device 11 has a function
of an on-line store, providing a sales service of items for the
user. The server device 11 provides item information to the user
terminals 13 through the network 14 and provides services such as
an on-line payment service in accordance with feedbacks received
from the user terminals 13. The server device 11 further stores
information of feedbacks (for example, purchase history and an
evaluation value) in the storage device 12 when receiving feedbacks
from the user terminals 13. The storage device 12 can be provided
inside a casing of the server device 11.
[0062] Information of the feedbacks stored in the storage device 12
is used for analyzing the user's taste or for analyzing relevance
between items. For example, the server device 11 analyzes the taste
of each user based on the number of purchasing items by each user
and evaluation values added to respective items by each user and
generates taste information in which high scores are assigned to
items having the large number of purchases and items having high
evaluation values (for example, refer to FIG. 6). The server device
11 also analyzes relevance between items based on existence of the
feedback added with respect to respective items by each user and
calculates relevance between items (for example, refer to FIG. 5).
The server device 11 further recommends items relating to the item
to which the feedback has been given to the user who has given the
feedback based on taste information and relevance between
items.
[0063] The above is the explanation of the configuration of the
recommendation system 10. The recommendation of items in the
recommendation system 10 is realized by functions of the server
device 11 and the storage device 12. Accordingly, the explanation
will be made with a focus on the functional configurations of the
server device 11 and the storage device 12 as follows.
1-2: FUNCTIONAL CONFIGURATIONS OF THE SERVER DEVICE 11 AND THE
STORAGE DEVICE 12
[0064] The functional configuration of the server device 11 and the
storage device 12 according to the embodiment will be explained
with reference to FIG. 2. FIG. 2 is an explanatory view for
explaining functional configurations of the server device 11 and
the storage device 12 according to the embodiment.
[0065] As shown in FIG. 2, the server device 11 includes a feedback
receiving unit 111, a feedback information update unit 112, an
intermediate-value table generation unit 113, a relevance
calculation unit 114, an item recommendation unit 115 and a taste
analyzing unit 116. The storage device 12 chiefly includes a
history table storage unit 121, an intermediate-value table storage
unit 122 and a relevance table storage unit 123.
[0066] First, a feedback given from the user by using the user
terminal 13 is inputted into the feedback receiving unit 111. For
example, when the user executes purchase operation of an item,
information indicating purchase of the item is inputted into the
feedback receiving unit 111 as the feedback. The feedback
information inputted in the feedback receiving unit 111 is inputted
into the feedback information update unit 112. When the feedback
information is inputted, the feedback information update unit 112
stores the inputted feedback information in the history table
storage unit 121.
[0067] The history table as shown in FIG. 3 is stored in the
history table storage unit 121. The history table stores
information indicating combinations of items which have received
feedbacks and the users who has given the feedbacks is stored so
that which user has given the feedback to which item can be
seen.
[0068] In the history table shown in FIG. 3 as an example, "User
ID" indicates identification information of users who have given
feedbacks. "Item ID" indicates identification information of items
to which feedbacks are given. Accordingly, it is found that a user
having "User ID"=1 has given the feedback with respect to items
having "Item ID"=1, 2 from the example of FIG. 3. Similarly, a user
having "User ID"=2 has given the feedback with respect to the item
having "Item ID"=1 and a user having "ID item"=3 has given the
feedback with respect to the item having "ID item"=2.
[0069] When the feedback information is stored in the history table
storage unit 121 by the feedback information update unit 112, the
intermediate-value table generation unit 113 generates the
intermediate value table (refer to FIG. 4) based on the history
table stored in the history table storage unit 121. The
intermediate value table indicates information concerning relevance
between items. The intermediate value table generated by the
intermediate-value table generation unit 113 is stored in the
intermediated-value table storage unit 122.
[0070] When the intermediate value table is stored in the
intermediate-value table storage unit 122, the relevance
calculation unit 114 calculates the relevance between items by
using the intermediate value table stored in the intermediate-value
table storage unit 122. Then, the relevance between items
calculated by the relevance calculation unit 114 is stored in the
relevance table storage unit 123 as the relevance table (refer to
FIG. 5).
[0071] When the relevance between items is stored in the relevance
storage unit 123, the item recommendation unit 115 selects an item
group having high relevance with respect to the item to which the
feedback has been given based on the relevance table stored in the
relevance storage unit 123. Next, the item recommendation unit 115
selects a given number of items corresponding to the user's taste
from the selected item group and recommends the items to the user
who has given the feedback. At this time, the item recommendation
unit 115 selects the given number of items corresponding to the
user's taste from the selected item group based on taste
information generated by the taste analyzing unit 116.
[0072] For example, the taste analyzing unit 116 assigns higher
scores to items to which many feedbacks were given in the past by
the users who have given the feedbacks, storing scores assigned to
respective items as taste information (referred to FIG. 6). In the
example of FIG. 6, the user having "User ID"=1 has a higher taste
for the item having "Item ID"=1. It is also preferable that the
taste analyzing unit 116 has a configuration of assigning higher
scores to items to which the user who has given the feedback gave
good feedbacks in the past.
[0073] The above is the explanation of the functional
configurations of the server device 11 and the storage device 12.
However, in the above explanation, configurations of the history
table, the intermediate value table and the relevance table as well
as a method of calculating stored values of the intermediate value
table from stored values of the history table and a method of
calculating stored values of the relevance table from stored values
of the intermediate value table are omitted. Accordingly, a
specific method of calculating relevance between items from
feedback information and a recommendation method of items based on
the relevance will be explained in more detail.
1-3: RECOMMENDATION METHOD USING P-MATRIX, D-MATRIX AND
J-MATRIX
[0074] The server device 11 according to the embodiment uses the
history table, the intermediate value table and the relevance table
for selecting items to be recommended to the user who has given the
feedback. However, the history matrix (hereinafter referred to as
P-matrix), the intermediate value matrix (hereinafter referred to
as D-matrix) and the relevance matrix (hereinafter referred to as
J-matrix) as shown in FIG. 7 to FIG. 9 are actually used instead of
using the history table, the intermediate value table and the
relevance table shown in FIG. 3 to FIG. 5.
1-3-1: Configurations of P-Matrix, D-Matrix and J-Matrix
[0075] Here, configurations of P-matrix, D-matrix and J-matrix will
be explained.
(P-Matrix)
[0076] The P-matrix is a matrix indicating the relation between
users who have given feedbacks and items receiving the feedbacks as
shown in FIG. 7. When a user u gives a feedback to an item i, 1 is
inputted to an element P (i, u) corresponding to the user u and the
item i. For example, 1 is inputted to a column (element P (1, 2) of
the P-matrix) corresponding to a user 2 (User 2) and an item 1
(Item 1), therefore, it is found that the user 2 has given the
feedback with respect to the item 1.
[0077] The P-matrix is stored in the history table storage unit
121. When the feedback receiving unit 111 receives the feedback,
the feedback information update unit 112 updates the P-matrix
stored in the history table storage unit 121.
(D-Matrix)
[0078] The D-matrix is a matrix indicating the relation between
items as shown in FIG. 8. The D-matrix is used for calculating the
later-described J-matrix, which is calculated based on the
P-matrix. An element D (i, j) of the D-matrix concerning an item i
and an item j is defined by the following expression (1). The
D-matrix is generated by the intermediate-value table generation
unit 113 and stored in the intermediate-value table storage unit
122. As the D-matrix is symmetrical with respect to i, j, it is
sufficient that only elements of i>j are stored in the
intermediate-value table storage unit 122 as shown in FIG. 8.
D ( i , j ) = v P ( i , v ) + v P ( j , v ) - v P ( i , v ) P ( j ,
v ) ( 1 ) ##EQU00001##
(J-Matrix)
[0079] The J-matrix is a matrix indicating relevance between items
as shown in FIG. 9. The J-matrix is calculated based on the
P-matrix and the D-matrix. An element J (i, j) of the J-matrix
concerning an item i and an item j is defined by the following
expression (2). The J-matrix is calculated by the relevance
calculation unit 114 and stored in the relevance table storage unit
123. As the J-matrix is symmetrical with respect to i, j, it is
sufficient that only elements of i>j are stored in the relevance
table storage unit 123 as shown in FIG. 9.
J ( i , j ) = v P ( i , v ) P ( j , v ) D ( i , j ) ( 2 )
##EQU00002##
[0080] The above is the explanation of the configurations of
P-matrix, D-matrix and J-matrix. Hereinafter, the update methods of
the above P-matrix, D-matrix and J-matrix and the recommendation
method of items using the above methods will be explained.
1-3-2: General Flow of Processing
[0081] First, the general flow of processing relating to the
recommendation method of items using P-matrix, D-matrix and J
matrix will be explained with reference to FIG. 10. FIG. 10 is an
explanatory chart for explaining the general flow of processing
relating to the recommendation method of items using P-matrix,
D-matrix and J matrix.
[0082] As shown in FIG. 10, first, the server device 11 determines
whether there is a feedback or not (S101). When receiving the
feedback by the feedback receiving unit 111, the server device 11
allows the process to proceed to Step S102. On the other hand, when
the feedback is not received by the feedback receiving unit 111,
the server device 11 returns the process to Step S101. When the
process proceeds to Step S102, the server device 11 updates the
P-matrix stored in the history table storage unit 121 based on
feedback information received by the feedback receiving unit 111
using the function of the feedback information update unit 112
(S102).
[0083] Subsequently, the server device 11 which has updated the
P-matrix stored in the history table storage unit 121 updates the
D-matrix stored in the intermediate-value table storage unit 122
based on the updated P-matrix using the function of the
intermediate-value table generation unit 113 (S103). Next, the
server device 11 which has updated the D-matrix stored in the
intermediate-table storage unit 122 updates the J-matrix stored in
the relevance table storage unit 123 based on the updated D-matrix
and the P-matrix using the function of the relevance calculation
unit 114 (Step S104). Next, the server device 11 recommends items
relating to the item receiving the feedback based on the updated
J-matrix and taste information using the function of the item
recommendation unit 115 (S105).
[0084] The above is the explanation of the general flow of the
processing relating to the recommendation method of items using
P-matrix, D-matrix and J-matrix.
1-3-3: Specific Flow of Processing
[0085] Next, specific update processing of P-matrix, D-matrix and
J-matrix will be explained with reference to FIG. 11 and FIG. 12.
FIG. 11 and FIG. 12 are explanatory charts for explaining specific
update processing of P-matrix, D-matrix and J-matrix. Here, the
specific processing flow executed when the user u purchases the
item i will be explained as an example. The feedbacks other than
the purchase are the same as this case.
[0086] First, as shown FIG. 11, the server device 11 determines
whether the element P (i, u) of the P-matrix is 0 or not by
referring to the P-matrix stored in the history table storage unit
121 (S111). When the element P (i, u) of the P-matrix is 0, the
server device 11 allows the process to proceed to Step S112. On the
other hand, when the element P (i, u) of the P-matrix is not 0, the
server device 11 allows the process to proceed to Step S113. The
case where the element P (i, u) of the P-matrix is 0 indicates a
case where the user u has not given the feedback to the item i
before.
[0087] When the process proceeds to Step S112, the server device 11
sets 1 in the element P (i, u) of the P-matrix and updates the
P-matrix stored in the history table storage unit 121 using the
function of the feed back information update unit 112 (S112). For
example, when the user 5 purchased the item 2, the element P (2, 5)
of the P-matrix is updated to 1 as shown in FIG. 13. The server
device 11 who has updated the P-matrix allows the process to Step
S113.
[0088] When the process proceeds to Step S113, the server device 11
starts a processing loop concerning a parameter k (k=1 to i-1)
(S113). First, the server device 11 determines whether an element P
(k, u) of the P-matrix is 1 or not by referring to the updated
P-matrix stored in the history table storage unit 121 (S114). When
the matrix P (k, u) of the P-matrix is 1, the server device 11
allows the process to Step S115. On the other hand, when the matrix
P (k, u) of the P-matrix is not 1, the server device 11 allows the
process to proceed to Step S116.
[0089] When the process proceeds to Step S115, the server device 11
updates the J-matrix stored in the relevance storage unit 123 by
using an element D (i, k) of the D-matrix by referring to the
D-matrix stored in the intermediate-value table storage unit 122
(S115). Specifically, the server device 11 adds 1/D (i, k) to the
element J (i, k) of the J-matrix stored in the relevance storage
unit 123 to calculate a new element J (i, k) of the J-matrix using
the function of the relevance calculation unit 114. Then, the new
element J (i, k) of the J-matrix is stored in the relevance storage
unit 123 and the process proceeds to Step S118.
[0090] When the process proceeds to Step S116 in Step S114, the
server device 11 adds 1 to the element D (i, k) of the D-matrix
stored in the intermediate-value table storage unit 122 to
calculate a new element D (i, k) of the D-matrix by the function of
the intermediate-value table generation unit 113 (S116). Then, the
new element D (i, k) of the D-matrix is stored in the
intermediate-value table storage unit 122 and the process proceeds
to Step S117.
[0091] Next, the server device 11 calculates a new element J (i, k)
of the J-matrix by using the updated D-matrix stored in the
intermediate-value table storage unit 122 and the J-matrix stored
in the relevance table storage unit 123 by the function of the
relevance calculation unit 114 (S117). At this time, the server
device 11 calculates the new element J (i, k) of the J-matrix based
on the following expression (3) by using the function of the
relevance calculation unit 114. Then the new element J (i, k) of
the J-matrix is stored in the relevance table storage unit 123 and
the process proceeds to Step S118.
J ( i , k ) = J ( i , k ) D ( i , k ) - 1 D ( i , k ) ( 3 )
##EQU00003##
[0092] When the process proceeds to Step S118, the server device 11
determines whether the process loop with respect to the parameter k
is completed or not (S118). Specifically, whether the parameter k
is i-1 or not is determined in Step S118. When k<i-1, 1 is added
to the parameter k and the process returns to Step S113. On the
other hand, when k=i-1, the process proceeds to Step S119 (FIG.
12).
[0093] Here, supplementary explanation of the update processing of
the D-matrix (refer to S114, S116) and the update processing of the
J-matrix (refer to S114, S115 and S117) will be made with specific
examples while referring to FIG. 14 and FIG. 15. In this case, a
case where the user 5 has purchased the item 2 will be considered
as an example (u=5, i=2). The update method of the P-matrix when
the user 5 has purchased the item 2 is the same as in the above
explanation made with reference to FIG. 13.
[0094] In the case where the user 5 has purchased the item 2, the
element P (2, 5) of the P-matrix is updated to 1 as shown in FIG.
13 (corresponding to Step S112).
[0095] When P (2, 5) is updated, the processing loop of the
parameter k (k=1 to 4) in Step S113 to S118 is started. First,
whether P (k, 5) is 1 or not is determined in Step S114. In the
example of FIG. 14, P (1, 5).noteq.1, P (2, 5)=1, P (3, 5)=1, P (4,
5).noteq.1, in the updated P-matrix, therefore, the process
proceeds to Step S115 when k=1, 4, and the process proceeds to Step
S116 when k=2, 3.
[0096] The update processing of the D-matrix is not executed when
the process proceeds to Step S115. That is, when k=2, 3, the update
processing of the D-matrix is not executed. On the other hand, when
the process proceeds to Step S116, the update processing of the
D-matrix is executed. Specifically, when k=1, 4, the update
processing of the D-matrix in Step S116 is executed. Therefore, in
the example of FIG. 14, only D (1,2) and D (2, 4) are update
targets in the D-matrix and 1 is added to be updated to new D (1,2)
and D (2, 4) respectively.
[0097] Next, the update processing of the J-matrix is executed.
When the process proceeds to Step S115 (when k=2, 3), 1/D (2, 3) is
added to J (2, 3) to be updated to a new J (2, 3) as shown in FIG.
15. On the other hand, when the process proceeds to Step S117 (when
k=1, 4), elements J are updated to new J (1, 2) and J (2, 4) based
on the above expression (3) as shown in FIG. 15.
[0098] As explained above, the D-matrix and the J-matrix are
updated in the processing loop (S113 to S118) concerning the
parameter k (k=1 to i-1). When the processing loop (S113 to S118)
is completed, the process proceeds to Step S119 (FIG. 12).
[0099] When the process proceeds to Step S119 (FIG. 12), the server
device 11 starts the processing loop (k=i+1 to n) concerning the
parameter k (S119). First, the server device 11 determines whether
the element P (k, u) of the P-matrix is 1 or not by referring to
the updated P-matrix stored in the history table storage unit 121
(S120). When the element P (k, u) of the P-matrix is 1, the server
device 11 allows the process to proceed to Step S121. On the other
hand, when element P (k, u) of the P-matrix is not 1, the server
device 11 allows the process to proceed to Step S122.
[0100] When the process proceeds to Step S121, the server device 11
updates the J-matrix stored in the relevance table storage unit 123
by using the element D (k, i) of the D-matrix by referring to the
D-matrix stored in the intermediate-value table storage unit 122
(S121). Specifically, the server device 11 adds 1/D (k, i) to the
element J (k, i) of the J-matrix stored in the relevance table
storage unit 123 to calculate a new element J (k, i) of the
J-matrix using the function of the relevance calculation unit 114.
Then, the new element J (k, i) of the J-matrix is stored in the
relevance table storage unit 123 and the process proceeds to Step
S124.
[0101] When the process proceeds to Step S122 in Step S120, the
server device 11 adds 1 to the element D (k, i) of the D-matrix
stored in the intermediate-value table storage unit 122 to
calculates a new element D (k, i) of the D-matrix using the
function of the intermediate-value table generation unit 113
(S122). Then, the new element D (k, i) of the D-matrix is stored in
the intermediate-value table storage unit 122 and the process
proceeds to Step S123.
[0102] Next, the server device 11 calculates a new element J (k, i)
of the J-matrix by using the updated D-matrix stored in the
intermediate-value table storage unit 122 and the J-matrix stored
in the relevance table storage unit 123 using the function of the
relevance calculation unit 114 (S123). At this time, the server
device 11 calculates a new element J (k, i) of the J-matrix based
on the following expression (4) using the function of the relevance
calculation unit 114. Then, the new element J (k, i) of the
J-matrix is stored in the relevance table storage unit 123 and the
process proceeds to Step S124.
J ( k , i ) = J ( k , i ) D ( k , i ) - 1 D ( k , i ) ( 4 )
##EQU00004##
[0103] When the process proceeds to Step S124, the server device 11
determines whether the processing loop concerning the parameter k
is completed or not (S124). Specifically, whether the parameter k
is n or not is determined in Step S124. When k<n, 1 is added to
the parameter k and the process returns to Step S119. On the other
hand, when k=n, the series of processing relating to the update of
P-matrix, D-matrix and J-matrix is completed. The specific update
methods of the D-matrix and the J-matrix in the processing loop of
Step S119 to S124 are substantially the same as the specific update
methods of the D-matrix and the J-matrix in the processing loop of
Step S113 to S118 explained with reference to FIG. 14 and FIG.
15.
[0104] The above is the explanation of the update methods of
P-matrix, D-matrix and J-matrix.
1-3-4: Recommendation Method Using Taste Information
[0105] Next, the processing flow relating to a recommendation
method of items using taste information and the J-matrix will be
explained with reference to FIG. 16. FIG. 16 is an explanatory
chart for explaining the processing flow relating to the
recommendation method of items using taste information and the
J-matrix.
[0106] As shown in FIG. 16, first, the server device 11 learns the
feedback and acquires taste information based on the learned result
using the function of the taste analyzing unit 116 (S131). The
server device 11 can be configured to use taste information
inputted by a user's hand using the function of the taste analyzing
unit 116.
[0107] For example, feedbacks such as purchase history, evaluation
values and access history of detailed information are inputted to
the taste analyzing unit 116 through the feedback receiving unit
111. When such feedbacks are inputted, the taste analyzing unit 116
assigns higher scores to items having higher purchase frequency,
items having higher evaluation values and items having higher
access frequency of detailed information to generate taste
information as shown in FIG. 6. As shown in FIG. 6, the taste
information is expressed by scores indicating tastes of respective
users with respect to respective items. Therefore, it is possible
to detect items preferred by the user by comparing scores of
respective items with respect to the user who gave feedbacks and
extracting items having higher scores in the taste information.
[0108] Here, supplementary explanation of the generation method of
taste information will be made. As described above, the taste
information is generated based on feedbacks given by the user.
However, there are various types of feedbacks such as purchase,
evaluation and access to detailed information. For example, it is
conceivable that the user's taste is higher in the purchased items
than in items detailed information of which has been just accessed.
Naturally, it is conceivable that the user's taste is higher in
items to which higher evaluation values were added than in items to
which lower evaluation values were added. Therefore, it is
necessary to consider weight corresponding to types of feedbacks
when generating taste information.
[0109] For example, when a score indicating a taste degree of the
user u with respect to the item i is expressed as UP (u, i), a
weight with respect to the k-th feedback is expressed as a.sub.k,
and a score added when the k-th feedback with respect to the item i
is given (the score not depending on types of feedbacks) is
expressed as x.sub.i, k, the score UP (u, i) is expressed by the
following expression (5). The score x.sub.i, k can be set to a
given value not depending on k as well as set to a score varying in
accordance with k. For example, it is also preferable that the
score X.sub.i, k is increased in the order of timing in which
feedbacks were given. As the weight a.sub.k, for example, a large
weight value (1.0) is set to the purchase, an intermediate weight
(0.6) is set to the evaluation and a low weight value (0.3) is set
to the access to the detailed information.
UP ( u , i ) = k a k x i , k ( 5 ) ##EQU00005##
[0110] When the taste information is generated by the taste
analyzing unit 116 as described above, the server device 11 inputs
the taste information generated by the taste analyzing unit 116 to
the item recommendation unit 115. The server device 11 also
extracts an item group relating to the item which has received the
feedback (hereinafter referred to as a related item group) by the
function of the item recommendation unit 115 (S132). At this time,
the server device 11 refers to the J-matrix stored in the relevance
table storage unit 123 to extracts the item group having higher
correlation values (values of the J-matrix) with respect to the
item receiving the feedback by the function of the item
recommendation unit 115.
[0111] The server device 11 which has extracted the related item
group selects a given number of items corresponding to the user's
taste (having higher scores) from the related item group by using
taste information inputted by the taste analyzing unit 116 by the
function of the item recommendation unit 115. Then, the server
device 11 generates a recommendation list including items selected
from the related item group (hereinafter, referred to as related
items) by the function of the item recommendation unit 115 (S133).
Subsequently, the server device 11 recommends related items
included in the recommendation list to the user who has given the
feedback by the function of the item recommendation unit 115
(S134).
[0112] The above is the explanation of the recommendation method
using the taste information.
1-3-5: About Efficiency Method
[0113] Next, devices for increasing calculation efficiency at the
time of calculating the J-matrix and use efficiency of a memory
will be explained with reference to FIG. 17 and FIG. 18. FIG. 17
and FIG. 18 are explanatory charts for explaining devices for
increasing the calculation efficiency at the time of calculating
the J-matrix and the use efficiency of the memory. As described
above, the calculation of the J-matrix corresponds to calculation
of the above expressions (1), (2). When referring to the above
expressions (1), (2), it is found that two kinds of sums S.sub.1,
S.sub.2 concerning elements of the P-matrix (the following
expressions (6), (7)) are included in the calculation.
S 1 ( i ) = v P ( i , v ) ( 6 ) S 2 ( i , j ) = v P ( i , v ) P ( j
, v ) ( 7 ) ##EQU00006##
[0114] The above S.sub.1 (i) indicates the frequency in which all
users gave feedbacks to the item i. On the other hand, the above
S.sub.2 (i, j) indicates the frequency in which the same user gave
the feedback to both of the items i, j. For example, when the type
of the feedback is the purchase, the number of purchases in each
item corresponds to the above expression (6) as shown in FIG. 17.
Additionally, the number of purchases (the number of
co-occurrences) in each combination of items corresponds to the
above expression (7) as shown in FIG. 18. Therefore, the sum
calculation represented by the above expressions (6), (7) can be
omitted by storing tables of the number of purchases and the number
of co-occurrences as shown in FIG. 17 and FIG. 18. As a result, the
calculation of the J-matrix expressed in the above expressions (1),
(2) can be made more effective.
[0115] The above is the explanation of the method of allowing the
calculation at the time of calculating the J-matrix to be more
effective.
1-4: MODIFICATION EXAMPLES
[0116] Next, modification examples of the embodiment will be
explained.
(1-4-1 Approximate Update Method of J-Matrix
First Modification Example
[0117] When applying the recommendation method of items using
P-matrix, D-matrix and J-matrix explained above, the recommendation
of related items reflecting the feedback can be executed in real
time at the time of receiving the feedback by the user. However,
there is a demand for reducing a calculation load of J-matrix to
reduce time necessary for recommending related items as short as
possible. Accordingly, the inventors of the case devised a method
of reducing the calculation load necessary for recommending related
items to reduce time necessary for recommending related items.
[0118] Hereinafter, an approximate update method of J-matrix will
be explained with reference to FIG. 19 to FIG. 21. FIG. 19 to FIG.
21 are explanatory charts for explaining the approximate update
method of J-matrix.
[0119] First, the processing flow relating to the approximate
update method of J-matrix will be explained with reference to FIG.
19. As shown in FIG. 19, first, the server device 11 calculates the
J-matrix based on feedback information at a time (t+1) (S141). The
calculation method of the J-matrix in Step S141 is the same as the
method explained above. Next, the server device 11 compares the
J-matrix at a previous time t with the J-matrix at the time (t+1)
to calculate an update value .delta. (S142).
[0120] The update value .sigma. is a parameter for approximately
calculating the J-matrix at a time (t+2). Here, supplementary
explanation of the calculation method of the update value .delta.
will be made.
[0121] As shown in FIG. 20, the server device 11 refers to the
number of feedbacks (the case of "purchase" is cited as an example)
given to respective items (the number of purchases) from the time t
to the time (t+1). In the example of FIG. 20, the number of
purchases is the largest in an item 4 (Item 4), the server device
11 selects the item 4. The server device 11 also stores the number
of purchases of the selected item 4 as n=64.
[0122] Subsequently, the server device 11 refers to the J-matrix at
the time t. At this time, the server device 11 refers to elements
of the J-matrix J (1, 4), J (2, 4), J (3, 4) relating to the item 4
the number of purchases is the largest. The server device 11
further refers to the J-matrix at the time (t+1). At this time, the
server device 11 refers to elements of the J-matrix J (1, 4), J (2,
4), J (3, 4) relating to the item 4 the number of purchases is the
largest. Then, the server device 11 compares the elements of the
J-matrix J (1, 4), J (2, 4), J (3, 4) at the time t with the
elements of the J-matrix J (1, 4), J (2, 4), J (3, 4) at the time
(t+1).
[0123] The server device 11 selects the maximum value .DELTA.J (2,
4)=0.16 from difference values .DELTA.J (1, 4)=0.12, .DELTA.J (2,
4)=0.16 and .DELTA.J (3, 4)=0.06 of the elements of the J-matrix J
(1, 4), J (2, 4), J (3, 4). The maximum value is expressed as d.
Furthermore, the server device 11 calculates the update value
.delta.=d/n=0.16/62=0.00258 by using the maximum value d=.DELTA.J
(2, 4). The explanation has been made with the specific example
(the case where the feedback type is "purchase"), however, the
update value .delta. can be calculated in the same manner also in a
case where the feedback type or the value of the J-matrix
differs.
[0124] FIG. 19 is referred to again. After calculating the update
value .delta. in Step S142, the server device 11 determines whether
there is a feedback or not at a next time (t+2) (S143). When there
is the feedback, the server device 11 allows the process to proceed
to Step S144. On the other hand, when there is no feedback, the
server device 11 returns the process to Step S143 and waits for the
feedback.
[0125] When the process proceeds to Step S144, the server device 11
updates the P-matrix based on the feedback at the time (t+2)
(S144). Subsequently, the server device 11 calculates the J-matrix
at the time (t+2) by using the update value .delta. calculated in
Step S142 (S145). However, the J-matrix calculated here is not
calculated by the above expressions (1), (2) based on the P-matrix
updated in Step S144, therefore, the J-matrix substantially
corresponds to the J-matrix at the time (t+2) predicted from the
J-matrixes at the time t and the time (t+1).
[0126] Here, supplementary explanation concerning the calculation
method of the J-matrix at the time (t+2) using the update value
.sigma. will be made with reference to FIG. 21. A case where the
user who gave the feedback (the case of "purchase" is cited as an
example in this case) to items 2, 3 in the past has given the
feedback to the item 1 at the time (t+2) will be considered.
[0127] As shown in FIG. 21, the server device 11 refers to the
J-matrix at the time (t+1) first. At this time, the server device
11 pays attention to elements (hereinafter referred to as focused
elements) concerning elements concerning combinations between the
item 1 purchased at the time (t+2) and items 2, 3 purchased before.
Next, the server device 11 adds the update value .delta. calculated
in Step S142 to values of focused elements at the time (t+1) to
calculates the J-matrix at the time (t+2).
[0128] As described above, the server device 11 executes update
processing of the J-matrix using the update value .delta.. The
server device 11 executes calculation of the J-matrix based on the
above expressions (1), (2) by the method described above at a given
timing. As described above, high-speed update processing is
realized by updating the J-matrix by using the update value
.delta.. Additionally, accumulation of error due to approximation
can be suppressed by executing calculation processing of the
J-matrix from the P-matrix at the given timing.
[0129] The approximate update method of the J-matrix will be
explained as the above.
1-4-2: Update Method of J-Matrix Based on Bias Parameters
Second Modification Example
[0130] Next, an update method of the J-matrix based on bias
parameters will be explained with reference to FIG. 22. FIG. 22 is
an explanatory chart for explaining the update method of the
J-matrix based on the bias parameters. The bias parameters in this
case indicate parameters for performing weighting to elements of
the J-matrix so as to correspond to the user's taste.
[0131] As shown in FIG. 22, the bias parameters are previously
determined in accordance with item types. In the example of FIG.
22, the bias parameter is set to 1.2.sup.n-1 when the item is game
software, the bias parameter is set to 1.1.sup.n-1 when the item is
video software and the bias parameter is set to 1.4.sup.n-1 when
the item is comics. n indicates the order of purchasing. For
example, when the user purchased the items 2, 3 (game software)
sequentially, the server device 11 calculates bias parameters
concerning items 2, 3. In this case, a bias parameter b1 of the
item 2 will be 1.2.sup.0=1.0, and a bias parameter b2 of the item 3
will be 1.2.sup.1=1.2.
[0132] After calculating the bias parameters b1, b2, the server
device 11 refers to the J-matrix and pays attention to elements of
the J-matrix J (1, 2), J (1, 3) and J (2, 3) relating to the items
2, 3. Then, the server device 11 multiplies the element of the
J-matrix J (1, 2) by the bias parameter b1 and multiplies the
elements of the J-matrix J (1, 3), J (2, 3) by the bias parameter
b2. As described above, the server device 11 updated the J-matrix
by the bias parameters generates the recommendation list based on
the updated J-matrix and recommends related items to the user based
on the generated recommendation list.
[0133] The above is the explanation of the update method of the
J-matrix based on the bias parameters.
1-5: HARDWARE CONFIGURATION
[0134] Functions of respective components included in the above
server device 11 and the user terminals 13 can be realized by using
a hardware configuration of an information processing apparatus,
for example, shown in FIG. 23. That is, functions of respective
components are realized by controlling the hardware shown in FIG.
23 by using a computer program. The form of the hardware is
optional, including, for example, a personal computer, information
terminals such as a cellular phone, a PHS and a PDA, a game machine
and various information home appliances. The PHS is an abbreviation
of "Personal Handy-phone System". The PDA is an abbreviation of
"Personal Digital Assistant".
[0135] As shown in FIG. 23, the hardware chiefly includes a CPU
902, a ROM 904, a RAM 906, a host bus 908 and a bridge 910. The
hardware further includes an external bus 912, an interface 914, an
input unit 916, an output unit 918, a storage unit 920, a drive
922, a connection port 924 and a communication unit 926. The CPU is
an abbreviation of "Central Processing Unit". The ROM is an
abbreviation of "Read Only Memory". The RAM is an abbreviation of
"Random Access Memory".
[0136] The CPU 902 functions as a calculation processing device or
a control device, controlling the whole operation of respective
components or part thereof based on various types of programs
stored in the ROM 904, the RAM 906, the storage unit 920 or a
removal recording medium 928. The ROM 904 is a means for storing
programs read by the CPU 902, data used for calculation and so on.
For example, programs to be read by the CPU 902, various parameters
varying appropriately at the time of executing the programs and so
on are stored in the RAM 906 temporarily or permanently.
[0137] These components are connected to one another, for example,
through the host bus 908 capable of transmitting data at high
speed. The host bus 908 is connected to, for example, the external
bus 912 transmitting data at relatively low speed through the
bridge 910. As the input unit 916, for example, a mouse, a
keyboard, a touch panel, buttons, switches, levers and so on are
used. Additionally, a remote controller capable of transmitting
control signals by using infrared rays or other electric waves is
occasionally used as the input unit 916.
[0138] As the output unit 918, devices capable of notifying
acquired information to users visually or auditorily can be used,
which are, for example, display devices such as a CRT, an LCD, a
PDP and an ELD, audio output devices such as a speaker, headphones
and a printer, a cellular phone, a facsimile and so on. The CRT is
an abbreviation of "Cathode Ray Tube". The LCD is an abbreviation
of "Liquid Crystal Display". The PDP is an abbreviation of "Plasma
Display Panel". The ELD is an abbreviation of "Electro-Luminescence
Display".
[0139] The storage unit 920 is a device for storing various data.
As the storage unit 920, for example, a magnetic storage device
such as a hard disk drive (HDD), a semiconductor storage device, an
optical storage device, a magneto-optical storage device and the
like are used. The HDD is an abbreviation of "Hard Disk Drive".
[0140] The drive 922 is a device reading information recorded in
the removable recording medium 928 such as a magnetic disk, an
optical disk, a magneto-optical disk, a semiconductor memory or the
like and writing information in the removable recording medium 928.
The As the removable recording medium 928, for example, DVD media,
Blue-ray media, HD DVD media, various semiconductor storage media
and so on can be used. Naturally, the removable recording medium
928 can apply, for example, an IC card on which a non-contact IC
chip is mounted, electronic devices and so on. The IC is an
abbreviation of "Integrated Circuit".
[0141] The connection port 924 is a port for connecting an external
connection device 930, which is, for example, an USB port, an
IEEE1394 port, a SCSI, an RS-232C port or an optical audio
terminal. As the external connection device 930, a printer, a
portable music player, a digital camera, a digital video camera, an
IC recorder and so on can be cited. The USB is an abbreviation of
"Universal Serial Bus". The SCSI is an abbreviation of "Small
Computer System Interface".
[0142] As the communication unit 926, communication devices for
connecting to a network 932 can be used, which are, for example,
communication cards for a wired or wireless LAN, Bluetooth
(Registered Trademark) and WUSB, a router for optical
communication, a router for ADSL, modems for various
communications. The network 932 connected to the communication unit
926 includes networks connected by wired or wireless communication,
which is, for example, a internet, a home LAN, an infrared
communication, a visible light communication, broadcasting,
satellite communication or the like. The LAN is an abbreviation of
"Local Area Network". The WUSB is an abbreviation of "Wireless
USB". Additionally, the ADSL is an abbreviation of "Asymmetric
Digital Subscriber Line".
2: BRIEF
[0143] Lastly, the technical contents according to the embodiment
of the present disclosure are briefly summarized. The technical
contents described here can be applied to various information
processing apparatuses such as a PC, a cellular phone, a portable
game machine, a portable information terminal, information home
appliances and a car navigation system.
[0144] The functional configuration of the information processing
apparatus can be expressed as follows. The information processing
apparatus includes a feedback receiving unit, a feedback history
storage unit, a relevance calculation unit and a recommendation
unit. The feedback receiving unit is a means for receiving
feedbacks with respect to items by a user. As the feedbacks,
purchase of items, access to information concerning items,
evaluation to items and so on.
[0145] The feedback history storage unit is a means for storing
information of feedbacks received at present or in the past by the
feedback receiving unit in units of combinations of users and
items. The feedback information includes, for example, purchase
history of items, access history of information concerning items,
evaluation information with respect to items and so on.
[0146] The relevance calculation unit is a means for calculating
relevance between items base on first information indicating the
degree in which the same user gives feedbacks to both items and
second information indicating the degree in which many users give
feedbacks by using feedback information stored in the feedback
history storage unit when receiving the feedback by the feedback
receiving unit. For example, when feedback information indicates
purchase history of the user, the first information will be
information indicating the number of purchases of items by the same
user. The second information will be information indicating the
number of items purchased by plural users.
[0147] The relevance between items will be the one reflecting the
user's taste by considering the above first information. On the
other hand, the relevance between items will be the one reflecting
the relevance among users through items by considering the second
information. That is, the relevance calculated by the relevance
calculation unit indicates the relevance between items reflecting
the user's taste and the relevance among users. The recommendation
unit immediately recommends items having higher relevance with
respect to the item to the user who has given the feedback to the
item by the feedback receiving unit.
[0148] As the result, it is possible to recommend the item to the
user selected in consideration of both the user's taste and the
relevance among users through the item. Additionally, after the
user gives the feedback, the feedback is reflected on the
recommendation result immediately, therefore, the item fitted to
the user's taste at the time of the feedback is recommended to the
user.
(Supplement)
[0149] The recommendation system 10 is an example of the item
recommendation system. The feedback information update unit 112 and
the history table storage unit 121 are examples of the feedback
history storage unit. The intermediate-value table generation unit
113 and the relevance calculation unit 114 are examples of the
relevance calculation unit. The item recommendation unit 115 is an
example of the recommendation unit. The relevance calculation unit
114 is an examples of a variation detection unit and a relevance
prediction unit. The taste analyzing unit 116 is an example of a
taste information generation unit.
[0150] The preferred embodiment of the present disclosure has been
explained with reference to the attached drawings, and it goes
without saying that the present disclosure is not limited to the
above example. It should be understood that those skilled in the
art can make various alternations and modifications within the
scope of the appended claims and that these naturally belong to the
technical scope of the present disclosure.
[0151] For example, two types of modification examples are shown in
addition to the basic configuration of the embodiment. It is
possible to combine the basic configuration with modification
examples and the manner of combination is not limited.
Additionally, the types of feedbacks, types of bias parameters are
specifically shown in the above explanation, however, the technical
scope of the embodiment is not limited to these specific
examples.
[0152] The present disclosure contains subject matter related to
that disclosed in Japanese Priority Patent Application JP
2010-135655 filed in the Japan Patent Office on Jun. 15, 2010, the
entire contents of which is hereby incorporated by reference.
[0153] It should be understood by those skilled in the art that
various modifications, combinations, sub-combinations and
alterations may occur depending on design requirements and other
factors insofar as they are within the scope of the appended claims
or the equivalents thereof.
* * * * *