JP2006190126A - Information processing apparatus and method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To appropriately determine a recommended item to be recommended to a user in response to various requests of the user. <P>SOLUTION: Of a shift center feature vector 21 as a base of comparison, for example, a component value 2000 of a first component corresponding to the "price" of which change is specified by the user is multiplied by an update amount specified by the user, namely a component value 0.7 of the "price" of the shift vector 23, the value 1400 obtained as a result becomes a new component value of the first component corresponding to the "price", and therefore the vector is updated to a vector 25. Then, the similarity between the center feature vector 25 after the update and each item feature vector of each of a plurality of items is calculated, and the item corresponding to the item feature vector of which calculated similarity satisfies a predetermined condition is determined as the recommended item. The present invention can be applied to a recommending system for recommending an item such as wine. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an information processing apparatus, method, and program, and more particularly, to an information processing apparatus, method, and program that can appropriately determine a recommended item to be recommended to a user in response to various requests of the user.

  In recent years, a system for recommending items to users (hereinafter simply referred to as a recommendation system) has become widespread as one of information processing systems (see, for example, Patent Document 1). An item refers to, for example, a product that can be purchased by a user, and is described as content in Patent Document 1. Details of the item will be described later.

  The demands of users are diversifying, and in recent years, items that have been recommended or presented to the user once or items prepared by the user are similar to those items, but some items with different characteristics are also recommended. There are many requests that you want.

Specifically, for example, there are many requests such as “I want you to recommend an item whose quality is similar to a predetermined item but whose price is lower”.
JP 2004-343321 A

  However, in the conventional recommendation system including Patent Document 1, it is only possible to recommend an item similar to the item once recommended to the user or an item prepared by the user. There was a problem that it was very difficult to recommend items. That is, the conventional recommendation system including Patent Document 1 has a problem that it cannot sufficiently satisfy the above-described user request.

  The present invention has been made in view of such a situation, and makes it possible to appropriately determine recommended items to be recommended to the user in response to various requests of the user.

  The information processing apparatus according to the present invention uses, as a standard vector, a vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of N components. For the comparison source vector and each of the plurality of items, the standard form vector that is the basis of the comparison is the comparison source vector, and each of the plurality of standard form vectors indicating the characteristics of the plurality of items is the comparison partner vector. An information processing apparatus that determines a recommended item to be recommended to a user from among a plurality of items using a comparison partner vector, and a change is designated by a user among N components of a comparison source vector Update means for updating the component value of the component corresponding to the first attribute using the first update amount designated by the user, and the comparison updated by the update means Calculates the degree of similarity between the vector itself or a part thereof and each of the comparison partner vectors for a plurality of items or a part thereof, and recommends an item corresponding to the comparison partner vector for which the calculated similarity satisfies a predetermined condition. And recommending means for determining as an item.

  The updating means further calculates the component value of the second component corresponding to the second attribute correlated with the first attribute among the N components of the comparison source vector, the first update amount, and the correlation. It can be made to update using the 2nd update amount determined according to this.

  When a plurality of attributes are designated as the first attribute by the user, and each of the first update amounts is individually designated for each of the plurality of first attributes, the updating unit is configured to update the N comparison source vectors. Among the components, the component values of the plurality of components corresponding to the plurality of first attributes can be respectively updated using the corresponding first update amount.

  The recommendation means further uses the respective weights set in advance for each of the N components of the standard vector to obtain the similarity between the comparison source vector itself or a part thereof and the comparison partner vector itself or a part thereof. Can be calculated.

  The recommendation means further includes a weight set for the first component corresponding to the first attribute among the weights set in advance for each of the N components of the standard vector. And the similarity between the comparison source vector itself or a part thereof and the comparison partner vector itself or a part thereof can be calculated by using the adjusted weights.

  The recommendation unit extracts each of one or more components including at least the first component from the N components of the comparison source vector, and generates a first vector including the extracted one or more components. Extract one or more components that are the same as the one or more components extracted by the comparison source vector from the N components of one vector to be compared among a plurality of comparison partner vectors. And generating a second vector composed of one or more extracted components, calculating a similarity between the first vector and the second vector, and if the calculated similarity satisfies a predetermined condition, The item corresponding to the two vectors can be determined as the recommended item.

  In the information processing method of the present invention, a vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of N components is a standard vector. For the comparison source vector and each of the plurality of items, the standard form vector that is the basis of the comparison is the comparison source vector, and each of the plurality of standard form vectors indicating the characteristics of the plurality of items is the comparison partner vector. An information processing method of an information processing apparatus for determining a recommended item to be recommended to a user from among a plurality of items using a comparison partner vector, which is changed by a user among N components of a comparison source vector An update step for updating the component value of the component corresponding to the first attribute for which the user is designated using the first update amount designated by the user, and an update step. The comparison source vector itself or a part thereof updated by the processing of the above and the comparison partner vector itself for a plurality of items or a part thereof are calculated, and the comparison partner satisfies the predetermined condition. A recommendation step of determining an item corresponding to the vector as a recommended item.

  The program of the present invention uses a standard vector as a standard vector for each of N attributes (N is an integer value of 1 or more) of an item and a predetermined one of the N components. Compare the original vector and multiple items using the original standard vector as the comparison source vector and the multiple standard vectors representing the characteristics of multiple items as the comparison target vectors. A program to be executed by a computer that controls a recommendation device that determines a recommended item to be recommended to a user from among a plurality of items using an opponent vector, and is executed by a user among N components of a comparison source vector Update the component value of the component corresponding to the first attribute for which the change is specified using the first update amount specified by the user. Calculating the degree of similarity between the step and the comparison source vector itself or a part thereof updated by the processing of the update step, and each of the comparison partner vectors for a plurality of items or a part thereof, and the calculated similarity is a predetermined value A recommendation step of determining an item corresponding to the comparison partner vector that satisfies the condition as a recommended item.

  In the information processing apparatus, method, and program of the present invention, a vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of the N components is a vector. The standard form vector is used as the comparison source vector, the comparison source vector is used as the comparison source vector, and each of the plurality of standard form vectors indicating the characteristics of the plurality of items is used as the comparison partner vector. A recommended item to be recommended to the user is determined from the plurality of items using the vector and the comparison partner vector for each of the plurality of items. Specifically, among the N components of the comparison source vector, the component value of the component corresponding to the first attribute designated to be changed by the user is updated using the first update amount designated by the user. The comparison partner vector itself or a part thereof and the comparison partner vector for a plurality of items are each calculated with a similarity, and the calculated similarity is a comparison partner vector that satisfies a predetermined condition. The item corresponding to is determined as a recommended item.

  As described above, according to the present invention, an item can be recommended to a user. In particular, it is possible to appropriately determine recommended items to be recommended to the user in response to various requests of the user.

  Embodiments of the present invention will be described below. Correspondences between constituent elements described in the claims and specific examples in the embodiments of the present invention are exemplified as follows. This description is to confirm that specific examples supporting the invention described in the claims are described in the embodiments of the invention. Therefore, even though there are specific examples that are described in the embodiment of the invention but are not described here as corresponding to the configuration requirements, the specific examples are not included in the configuration. It does not mean that it does not correspond to a requirement. On the contrary, even if a specific example is described here as corresponding to a configuration requirement, this means that the specific example does not correspond to a configuration requirement other than the configuration requirement. not.

  Further, this description does not mean that all the inventions corresponding to the specific examples described in the embodiments of the invention are described in the claims. In other words, this description is an invention corresponding to the specific example described in the embodiment of the invention, and the existence of an invention not described in the claims of this application, that is, in the future, a divisional application will be made. Nor does it deny the existence of an invention added by amendment.

  According to the present invention, an information processing apparatus is provided. This information processing apparatus (for example, the information processing apparatus of FIG. 3) has N (N is an integer value of 1 or more) attributes (for example, FIG. 5 and the like) for items (for example, wine in the present embodiment described later). A vector having a standard form (for example, FIG. 5 etc.) with each of “price”, “sweetness”, “spicy”, and “fragrance” shown in FIG. (Vectors such as price, sweetness, spiciness, and fragrance) shown in FIG. 5, the standard form vector to be compared is set as a comparison source vector (for example, vector 21 in FIG. 5), and each of a plurality of items Using each of the plurality of normal form vectors indicating features as a comparison partner vector, using the comparison source vector and the comparison partner vector for each of a plurality of items, the plurality of items from among the plurality of items Is an information processing apparatus for determining the recommended items to be recommended to over The. Specifically, among the N components of the comparison source vector, the component value of the component corresponding to the first attribute designated to be changed by the user is set as the first update amount designated by the user. (For example, in the example of FIG. 8, the first attribute is “price”, and the first update amount is the component value 0.7 of “price” of the shift vector 23. For example, FIG. In the example 9, the first attribute is “sweetness”, and the first update amount is the “sweetness” component value 2 of the shift vector 24) update means (for example, the shift parameter of FIG. 3) The calculation unit 13) calculates the degree of similarity between the comparison source vector itself or a part thereof updated by the updating means and each of the comparison partner vectors for a plurality of items or a part thereof. The similarity satisfies a predetermined condition The item corresponding to be the comparative counterpart vector, the recommendation means for determining a recommended item (e.g., recommendation list generation unit 16 of FIG. 3), characterized in that it comprises a.

  The updating unit of the information processing apparatus further includes a second attribute (for example, “sweet” in the example of FIG. 9) that is correlated with the first attribute among the N components of the comparison source vector. The component value of the second component corresponding to “spicy” correlated with “sa” ”is set to a second update amount (for example, FIG. 9) determined according to the first update amount and the correlation. In the example, it can be updated using the component value-2) of “spiciness” of the shift vector 24.

  In this information processing apparatus, when the user specifies a plurality of attributes as the first attribute, and each of the first update amounts is individually specified for each of the plurality of first attributes ( For example, when the image 1 and the image 2 in FIG. 2 are used and “price” and “taste” are specified at the same time), the updating means includes the N of the components of the comparison source vector, Each component value of the plurality of components corresponding to each of the plurality of first attributes is updated using the corresponding first update amount (for example, as indicated by a dotted arrow in FIG. 1, Update directly from vector I01 to vector I03).

  The recommendation means of this information processing apparatus is configured so that each weight set in advance for each of the N components of the standard vector (for example, each component value of the effect vector 27 in FIGS. , Each weight) can be further utilized to calculate the similarity between the comparison source vector itself or a part thereof and the comparison partner vector itself or a part thereof.

  The recommendation unit of the information processing apparatus further includes the first attribute corresponding to the first attribute among the weights set in advance for each of the N components of the standard vector. The amount of weight set for one component is adjusted (for example, “price” is the first component (shift target attribute) in the example of FIG. 10). In the example of FIG. 11, since “sweetness” is the first component (shift target attribute), the component value of “sweetness” is adjusted), and each weight after adjustment is adjusted. Can be used to calculate the similarity between the comparison source vector itself or a part thereof and the comparison partner vector itself or a part thereof.

  The recommendation unit of the information processing apparatus includes one or more vectors including at least the first component from among the N components of the comparison source vector (for example, the post-update shifted center feature vector 25 in FIG. 8). Each of the components (for example, the first component corresponding to “price”) is extracted, and a first vector (for example, the primary selection vector 25-1 in FIG. 14) composed of the one or more extracted components is extracted. One or more same as one or more of the components extracted from the comparison source vector, out of N components of one vector to be compared among the plurality of comparison partner vectors Each of the components is extracted, a second vector composed of one or more of the extracted components is generated, a similarity between the first vector and the second vector is calculated, and the calculated similarity Degree is predetermined When the condition is satisfied, an item corresponding to the second vector can be determined as the recommended item.

  According to the present invention, an information processing method is provided. This information processing method is the information processing method of the information processing apparatus of the present invention described above. Specifically, in this information processing method, the component value of the component corresponding to the attribute specified by the user among the N components of the comparison source vector is set as the update amount specified by the user. Update step (for example, steps S2 to S5 in FIG. 4) to be used and updated, the comparison source vector itself or a part thereof updated by the processing of the update step, and each of the comparison partner vectors for the plurality of items A recommendation step (for example, steps S7 to S7 in FIG. 4) that calculates the similarity with itself or a part thereof, and determines the item corresponding to the comparison partner vector that satisfies the predetermined condition as the calculated similarity. S11).

  Furthermore, according to the present invention, there are also provided a program corresponding to the information processing method of the present invention described above and a recording medium on which the program is recorded. As will be described later, this program is executed by, for example, the computer shown in FIG.

  Next, the principle of the present invention will be described before describing the embodiment of the present invention.

  In the present specification, an item refers to software or hardware (article) that can be provided to a user. For example, software such as so-called content, that is, a television broadcast program, a movie, a photograph, a song, and the like (moving image, still image, sound, or a combination thereof) is also an item in this specification. Further, for example, an article (hardware) such as wine is also an item referred to in this specification. In addition, for example, sentences and conversations are also items in this specification.

  However, in order to facilitate understanding of the present invention, in the following, when it is necessary to give specific examples of items, wine (sake) packed in a predetermined container (for example, a 750 ml bottle) is adopted. And

  First, an outline of an example of a series of processes (hereinafter simply referred to as a recommendation process) until an item is generally recommended that has been conventionally performed will be described.

  However, for simplicity of explanation, it is assumed that one information processing apparatus executes all of the recommendation processing.

  First, a conventional information processing apparatus vectorizes an item by using each of N pieces (N is an integer value of 1 or more) of information indicating the characteristics of an item as a base vector. At that time, the conventional information processing apparatus weights each component of the vector using a predetermined weighting method (gives a weight value) as necessary.

  Hereinafter, a vector obtained as a result of vectorizing items in this way is referred to as an item feature vector. Hereinafter, information indicating the feature of an item is referred to as attribute information, and the type of attribute information is referred to as an attribute. That is, when attribute information of each of N attributes is associated with a predetermined item, the weight value (N values obtained as a result of digitization) of N attribute information is set for each component. A vector having a value is an item feature vector.

  For example, when an item is wine, “price”, “variety of grapes”, “taste”, “scent”, and the like are attributes. Also, for example, for a given wine, the “price” is “10,000 yen”, the “rice variety” is “a type”, and the “taste” is evaluated as “3” out of 5 levels. It is assumed that “scent” is evaluated as “2” out of 5 levels. In this case, the attribute information for each attribute of “price”, “variety of strawberry”, “flavor”, and “fragrance” is “10,000 yen”, “a type”, “3”, It becomes each of "2". Therefore, when it is specified that the item feature vector of wine is expressed by a vector such as (price, grape variety, taste, aroma), “10,000 yen”, “a seed”, “3”, “2” Is obtained by quantifying each of α, β, γ, and θ (α, β, γ, and θ are arbitrary independent numbers) and substituting them for each of the first to fourth components. The resulting vector, ie, (α, β, γ, θ), becomes the item feature vector for a given wine.

  When attribute information is not associated with an item for a predetermined attribute of N attributes, 0 is assigned to a component corresponding to the predetermined attribute of the item feature vector of the item. And

  A vector in which the order of each of the N components is fixed in advance and the N components are arranged in accordance with the order is referred to as a standard vector. Specifically, for example, in the above-described example, a vector such as (price, grape variety, taste, fragrance) is a standard vector of wine item feature vectors.

  Conventional information processing apparatuses generate and store such standard item feature vectors for each item.

  On the other hand, a conventional information processing apparatus generates a standard vector indicating a user's preference using a user's past history and various information input by the user himself / herself. Hereinafter, a standard vector indicating the user's preference is referred to as a user preference vector or UPV (User Preference Vector).

  Then, the conventional information processing apparatus obtains a similarity such as cosine correlation between the UPV and each of the accumulated item feature vectors (matching process is performed), for example, the similarity is equal to or greater than a threshold value. An item corresponding to the item feature vector is presented to the user as a recommended item to be recommended to the user.

  The outline of the recommendation process in the conventional recommendation system has been described above.

  Further, for example, when the user requests that other items similar to the recommended item presented in this way are recommended, the conventional information processing apparatus, the item feature vector of the recommended item, The similarity with each of the item feature vectors of a plurality of other items can be obtained, and the item corresponding to the item feature vector whose similarity is equal to or greater than the threshold can be presented to the user as another item similar to the recommended item.

  However, for example, as described above, when a user requests that other items that are similar to the recommended item but have different characteristics are recommended, there is a method for responding to such a request. Since the conventional information processing apparatus does not exist, the conventional problem that the conventional information processing apparatus cannot sufficiently meet such a demand arises.

  Therefore, the present inventors have invented the following method. That is, the apparatus determines the component corresponding to the attribute designated to be changed by the user from among the components of the standard vector that is the basis of comparison (in the above example, the item feature vector of the first recommended item). The component value is updated using the update amount specified by the user. In other words, the device can be said to “shift” the component values. Therefore, hereinafter, a standard vector as a basis for comparison is referred to as a shifted center feature vector. Hereinafter, an attribute designated to be changed by the user is referred to as a shift target attribute, and an update amount designated by the user is referred to as a shift amount. After that, the apparatus calculates the similarity between the shift center feature vector in which the component value of the shift target attribute is updated according to the shift amount and the item feature vector for a plurality of other items, and calculates the similarity An item corresponding to an item feature vector whose degree satisfies a predetermined condition is determined as a recommended item. A technique corresponding to the series of processes described above is a technique newly invented by the present inventors.

  Hereinafter, with reference to FIG. 1 and FIG. 2, the method of the present invention will be further described using a specific example.

  1 and 2 show a specific example of the case where the method of the present invention is applied to wine recommendation.

  In the example of FIGS. 1 and 2, for the sake of simplicity, it is assumed that the item feature vector in the standard form of wine is a vector such as (taste, price).

  Thus, here, since the item feature vector in the standard form of wine is a two-dimensional vector (a vector composed of a first component and a second component), in FIG. "Is the vertical axis, and the price of the second component is the horizontal axis, and examples of item feature vectors for each wine are shown. That is, Δ (triangle mark) indicates the item feature vector of the comparison source wine, that is, the shift stop feature vector. In addition, ● (black circle mark) indicates the item feature vector of the wine to be compared.

  For each of the wines, the seller, the service provider of the recommendation system, etc. sampled in advance, and based on the results of the tasting, the numerical value determined to be applicable from 0 to 5 as the evaluation value of “taste” It is assumed that the determined numerical value is directly substituted into the component value of the first component corresponding to the “taste” of the item feature vector. However, with regard to “taste”, as shown in FIG. 1, the higher the value, the “sweeter” wine, and the lower the value, the “bitter” wine.

  Further, it is assumed that the value obtained as a result of normalizing the price of each wine in the range of 0 to 5 is substituted for the component value of the second component corresponding to the “price” of the item feature vector.

  In FIG. 1, for example, assume that the original feature vector of the comparison source wine, that is, the shift center feature vector before update (before shift) is the vector I01. In this case, as is apparent from FIG. 1, the item feature vectors similar to the shifted center feature vector I01 are the three vectors I1 to I3. Therefore, at the stage before update (before shifting), that is, conventionally, three wines corresponding to the item feature vectors I1 to I3 are presented to the user as recommended items.

  Specifically, for example, the wine corresponding to each of the item feature vectors I1 to I3 is a wine having the names “Chateau XX”, “Chateau XX”, and “Chateau XX”. . In this case, for example, the image 1 shown in FIG. 2 is presented to the user.

  That is, FIG. 2 shows an example of a GUI (Graphical User Interface) image to which the method of the present invention is applied.

  In the example of FIG. 2, in the display area below the message “search result” of image 1 or image 2 described later, information about each of the three wines as recommended items, that is, their names, prices, And each taste is displayed.

  Further, below the message “Find a slightly different item” below that, an image for the user to perform a GUI operation for setting the shift target attribute and the shift amount is displayed. In other words, the user can set the attribute of at least one of “price” and “taste” (or both) as a shift target attribute. The shift amount can be set (adjusted) by moving the shift target attribute “+ (plus mark)” to the left and right.

  Specifically, in the example of FIG. 2, when “price” is set as a shift target attribute, the second of the shift center feature vector I01 increases as “+ (plus sign)” is moved to the right in the figure. The amount of shift is such that the value of the component increases, that is, the amount of shift closer to “high” in “price”. When “taste” is set as the shift target attribute, the value of the first component of the shift center feature vector I01 increases as “+ (plus sign)” is moved to the right in the figure. The shift amount, that is, the shift amount that “taste” is closer to “sweet”.

  Thus, for example, the user requests that the taste remain slightly sweet (similar in terms of “taste” to the recommended item shown in image 1) but wants to recommend a cheaper wine ”. If “+” is added to “Price” of the image 1, it is simply moved to the left.

  Then, the shifted center feature vector is updated from the vector I01 to, for example, the vector I02, as shown in FIG. In other words, the component value of the second component corresponding to the “price” of the shifted center feature vector is updated so as to decrease (become a “lower” numerical value).

  As a result, as is apparent from FIG. 1, the item feature vectors similar to the shifted center feature vector I02 are now three vectors I4 to I6. Therefore, this time, three wines corresponding to each of the item feature vectors I4 to I6 are presented to the user as recommended items. Specifically, for example, it is assumed that the wine corresponding to each of the item feature vectors I4 to I6 is a wine having the names “Chateau XX”, “Chateau ΔΔ”, and “Chateau ○ Δ”. In this case, for example, the image 2 shown in FIG. 2 is presented to the user.

  In this way, by applying the method of the present invention, the user simply performs a simple operation, and “the taste may remain slightly sweet (although the taste is similar to the recommended item in FIG. 1). "I want you to recommend cheaper wines," you can easily meet the demand. That is, the conventional problem described above can be solved.

  Further, for example, the user requests that “the price range may remain as it is (similar to the recommended item shown in the image 2 and“ price ”is similar) but a little sweeter (a bit bitter) wine is recommended”. In this case, the “+ (plus sign)” of “Taste” of the image 2 is simply moved to the left.

  Then, as shown in FIG. 1, the shifted center feature vector is updated from the vector I02 to, for example, the vector I03. That is, the component value of the first component corresponding to the “taste” of the shifted center feature vector is updated so as to decrease (become a “bitter” value).

  As a result, as is apparent from FIG. 1, the item feature vectors similar to the shifted center feature vector I03 are now three vectors I7 to I9. Accordingly, although not shown, three wines corresponding to the item feature vectors I7 to I9 are presented to the user as recommended items.

  Furthermore, for example, in a state where the image 1 of FIG. 2 is displayed, the user wants to recommend wine that is cheaper and less sweet (a bit bitter) than the recommended item of the image 1. When the request is made to the device, “+ (plus sign)” of “price” of image 1 is simply moved to the left, and “+ (plus sign)” of “taste” of image 1 is simply moved to the left. It will be good.

  Then, the shifted center feature vector is directly updated from the vector I01 to the vector I03, for example, as indicated by the dotted arrow in FIG. As a result, also in this case, three wines corresponding to the item feature vectors I7 to I9 are presented to the user as recommended items.

  In order to facilitate comparison with the conventional recommendation method described above, the original (before shifting) shifted center feature vector, for example, the shifted center feature vector I01 in the example of FIG. 1, is the first recommended item. However, the present invention is not limited to this, and any vector may be used. Specifically, for example, the standard user preference vector itself may be adopted as the original (before shifting) shift center feature vector, or an appropriate initial value for each component value of the standard vector You may employ | adopt the vector obtained as a result of substituting a value. Further, when the latter vector is adopted, the initial value may be a value determined by the information processing apparatus or a value designated by the user.

  The principle (method) of the present invention has been described above.

  Next, embodiments of the present invention will be described with reference to FIG. 3 and subsequent drawings.

  FIG. 3 shows a functional configuration example of a recommendation system to which the above-described method of the present invention is applied, that is, a recommendation system to which the present invention is applied. That is, the recommendation system in FIG. 3 searches for recommended items to be recommended to the user from among a plurality of items, and presents the search results to the user.

  In the present specification, the system represents the entire apparatus including a plurality of apparatuses and processing units. Therefore, the recommendation system in FIG. 3 may be configured by one information processing apparatus or a plurality of information processing apparatuses. However, in this embodiment, it is assumed that the recommendation system in FIG. 3 is configured by one information processing apparatus. That is, FIG. 3 shows an example of the functional configuration of an information processing apparatus to which the present invention is applied.

  In the present embodiment, it is assumed that the item is wine, and the standard item feature vector is a vector such as (price, sweetness, hotness, fragrance). The “price” component value substitution method includes the above-described substitution method of the “price” component value among the vectors (taste, price) used as standard item feature vectors in FIGS. 1 and 2. The same is assumed. In addition, regarding the substitution methods for the component values of “sweetness”, “spiciness”, and “fragrance”, all of the substitution methods described above for the “taste” component value in the vector (taste, price) are used. It is assumed that the method is the same.

  In the example of FIG. 3, the information processing apparatus is configured to include the item feature vector storage unit 11 to the presentation unit 18.

  In addition, each form of the item feature vector memory | storage part 11 thru | or the presentation part 18 is not specifically limited. In particular, the user request acquisition unit 12, the shift parameter calculation unit 13, the recommendation list generation unit 16 and the like may be configured by software, hardware, or software and hardware. You may comprise as a combination.

  The item feature vector storage unit 11 stores standard item feature vectors for each of one or more wines. In the present embodiment, it is assumed that a predetermined one of the item feature vectors stored in the item feature vector storage unit 11 is used as the original (before shifting) shift center feature vector.

  When the user request acquisition unit 12 is operated by the user and the shift target attribute and the shift amount are selected, the user request acquisition unit 12 notifies the shift parameter calculation unit 13 of the selection.

  Specifically, for example, in the present embodiment, the user request acquisition unit 12 presents a GUI image such as the image 1 and the image 2 in FIG. 2 described above to accept the user's operation by the GUI. However, more precisely, in the present embodiment, the standard item feature vector is a vector such as (price, sweetness, hotness, fragrance), so that the image 1 and the image 2 themselves in FIG. 2 are presented to the user. In fact, the following image is presented to the user. That is, with respect to the image 1 and the image 2 in FIG. 2, “+ (plus) for each of“ price ”,“ sweetness ”,“ spicy ”, and“ scent ”is below“ search for slightly different items ”. In this embodiment, a GUI image on which an image for freely moving the “mark”) is displayed to the user is presented to the user.

  The shift parameter calculation unit 13 acquires a predetermined one of the item feature vectors stored in the item feature vector storage unit 11 as a shift center feature vector.

  The shift parameter calculation unit 13 is used to update the update target component among the components of the shift center feature vector by a predetermined amount based on the shift target attribute and the shift amount supplied from the user request acquisition unit 12. Generate a standard vector to be generated. Such a vector is referred to as a shifted vector. As specific examples of the shift vector, a vector 23 is shown in FIG. 6 described later, and a vector 24 is shown in FIG. 7 described later.

  In the example described above with reference to FIGS. 1 and 2, the update target of the components of the shift center feature vector is the component of the shift target attribute itself.

  However, among the attributes indicated by the components of the shifted center feature vector, there are attributes that have some correlation with other attributes. Specifically, for example, in the present embodiment, the item is wine, and the standard item feature vector is a vector such as (price, sweetness, spiciness, aroma). It is expected that there is some correlation between the attributes of “spiciness”. Specifically, here, for example, in wine, when “sweetness” is increased by a predetermined unit (when sweetened by one unit), “spiciness” is decreased by one unit ( It is assumed that there is a correlation (hereinafter, referred to as sweet and spicy correlation) that is spicy by one unit. When this assumption is satisfied, for example, when “sweetness” is set as a shift target attribute, only the “sweetness” component value of each component of the shift center feature vector is updated. If the component value of “sa” is not updated, the spicy correlation is broken in the shifted center feature vector after the update (after the shift). For this reason, wine recommendation is performed using the shifted center feature vector whose sweet and spicy correlation has collapsed, and as a result, a new problem arises that the recommended wine may not be appropriate for the user. End up.

  Therefore, in order to solve this new problem, the present inventor, in addition to the method of the present invention described above, not only the component of the shift target attribute but also the component correlated with the shift target attribute We invented a method to use as a component to be updated.

  In order to apply this technique, in the present embodiment, for each of the attributes corresponding to each of the two or more correlated attributes among the attributes corresponding to each of the components of the shifted center feature vector, It is assumed that a standard vector obtained by substituting values indicating correlation is used. Such a vector is hereinafter referred to as an inter-attribute correlation vector.

  Specifically, for example, the inter-attribute correlation vector for the sweet and spicy correlation, in other words, the inter-attribute correlation vector used when the shift target attribute is “sweet” or “spicy” is shown in FIG. It becomes a vector such as (0, 1, -1, 0) like the vector 22-B. However, in this case, it is assumed that there is a correlation between “sweetness” and “spiciness”, but there is no correlation between other “price” and “scent”.

  Further, for example, assuming that “price” has no correlation with any of the other attributes of “sweetness”, “spicy”, and “fragrance”, the shift target attribute is “price”. The attribute correlation vector used in this case is a vector such as (1, 0, 0, 0) like a vector 22-A shown in FIG.

  In the present embodiment, it is assumed that such an inter-attribute correlation vector is stored in advance in the inter-attribute correlation vector storage unit 14.

  Therefore, in the present embodiment, when the shift target attribute and the shift amount are supplied from the user request acquisition unit 12, the shift parameter calculation unit 13 stores an inter-attribute correlation vector for the shift target attribute. A shift vector is generated using the shift target attribute, the shift amount, and the inter-attribute correlation vector for the shift target attribute.

  Details of the shift vector generation method will be described later with reference to FIGS.

  The shift vector generated by the shift parameter calculation unit 13 is temporarily stored in the shift vector storage unit 15. That is, the shift vector storage unit 15 stores one or more shift vectors.

  Then, the shift parameter calculation unit 13 updates the component value to be updated of the shift center feature vector using the shift vector once stored in the shift parameter storage unit 15, and generates the updated shift center feature vector as a recommendation list. To the unit 16.

  Details of the shift center feature vector update method will be described later with reference to FIG. 4 and subsequent drawings.

  The recommendation list generation unit 16 includes a cosine correlation between the updated shift center feature vector supplied from the shift parameter calculation unit 13 and each of the plurality of item feature vectors stored in the item feature vector storage unit 11 or 17. Each similarity is obtained. That is, the item feature vector storage unit 17 also stores one or more item feature vectors.

  Next, the recommendation list generation unit 16 calculates the recommendation level of the item corresponding to each of the plurality of item feature vectors based on the similarity. The method of calculating the recommendation level is not particularly limited, but here, for simplicity of explanation, it is assumed that the similarity level is directly used as the recommendation level.

  Next, the recommendation list generation unit 16 determines, for example, one or more items corresponding to each of one or more item feature vectors having a recommendation degree equal to or higher than a threshold as recommended items.

  Then, the recommendation list generation unit 16 generates a list (hereinafter, referred to as a recommendation list) in which information that can identify each of one or more recommended items is sorted in descending order of recommendation, for example, and provides the presentation unit 18 with the list. .

  For example, when the item is wine as in the present embodiment, the information that can identify the recommended item refers to the name of the wine, the place of production, the producer (domaine or winery), and the like. It is also possible to include other information in the recommendation list, such as the price of wine and the reason for recommending that wine. Specifically, for example, the information displayed below the “search result” in the image 1 and the image 2 in FIG. 2 described above is an example of a recommendation list.

  In the example of FIG. 3, the item feature vector storage units 11 and 17 are different storage units, but may be the same storage unit.

  Next, an example of the “recommendation process” executed by the information processing apparatus of FIG. 3 will be described with reference to the flowchart of FIG. That is, FIG. 4 is a flowchart for explaining an example of “recommendation processing” to which the present invention is applied. The “recommendation process” to which the present invention is applied is hereinafter referred to as “shift recommendation process” in order to distinguish it from the conventional “recommendation process”.

  In order to facilitate understanding of the “shift recommendation process”, the “shift recommendation process” will be described below with reference to the specific examples of FIGS. 5 to 11 as appropriate. That is, each of FIGS. 5 to 11 shows a specific example of the processing result of a predetermined step in the “shift recommendation process”.

  In step S <b> 1, the user request acquisition unit 12 in FIG. 3 acquires the shift target attribute and the shift amount, and supplies them to the shift parameter calculation unit 13.

  In step S2, the shift parameter calculation unit 13 acquires a predetermined item feature vector from the item feature vector storage unit 11 as a shift center feature vector.

  Specifically, for example, suppose that the vector 21 shown in FIG. 5, that is, the vector (2000, 2, 4, 3) is acquired as the shifted center feature vector.

  Returning to FIG. 4, in step S <b> 3, the shift parameter calculation unit 13 acquires an attribute correlation vector related to the shift target attribute from the attribute correlation vector storage unit 14.

  Specifically, for example, when the shift target attribute is “price”, the vector 22-A shown in FIG. 6, that is, the vector (1,0,0,0) is used as the inter-attribute correlation vector. To be acquired.

  Further, for example, when the shift target attribute is “sweet”, the vector 22-B shown in FIG. 7, that is, the vector (0,1, -1,0) is acquired as the inter-attribute correlation vector. Is done.

  Returning to FIG. 4, in step S <b> 4, the shift parameter calculation unit 13 generates a shift vector using the shift amount and the correlation vector between attributes, and temporarily stores the shift vector in the shift vector storage unit 15.

  The generation method of the shift vector in step S4 is not particularly limited. Here, for example, each component value of the inter-attribute correlation vector is multiplied by the shift amount, and each of the multiplication values is converted into the corresponding component. Assume that a generation method is adopted in which substitution is performed and a vector obtained as a result is shifted.

  In this case, for example, when the shift target attribute is “price” and the shift amount is 0.7, as shown in FIG. 6, the correlation vector between attributes is (1,0,0, Since the vector 22-A is (0), a vector 23 (1 × 0.7, 0 × 0.7, 0 × 0.7, 0 × 0.7) = (0.7,0,0,0) is generated as a shift vector.

  Further, for example, when the shift target attribute is “sweet” and the shift amount is 2.0, the correlation vector between attributes is (0,1, -1, -1, as shown in FIG. 7). 0) vector 22-B, a vector 24 of (0 × 2,1 × 2, −1 × 2,0 × 2) = (0,2, −2,0) is generated as a shift vector. The

  Returning to FIG. 4, in step S <b> 5, the shift parameter calculation unit 13 updates the shift center feature vector using the shift vector once stored in the shift vector storage unit 15.

  The method for updating the shift center feature vector in step S5 is not particularly limited, but here, for example, the component whose shift vector component value is not “0” is the update target, and the component whose update center feature vector is to be updated is: The component value is updated by multiplying or adding the corresponding component value (component value that is not “0”) of the shift vector to the component value before the update, and the component value remains as it is for components other than the update target It is assumed that an updating method such as “No” is performed (component value is not updated). It should be noted that whether the component value corresponding to the shift vector is multiplied or added to the component value before update is determined based on the feature of the attribute corresponding to the component to be updated and the correlation between the attributes. Depends on features.

  In this case, for example, when the shift vector 23 of FIG. 6 is generated in the process of the previous step S4, the shift center feature vector 21 is (2000, 2, 4, 3) as shown in FIG. Since it is a vector and the shift vector 23 is a vector of (0.7,0,0,0), the update target is only the first component of “price”. As a result, for example, (2000 × 0.7,2,4 , 3) = (1400,2,4,3), the vector 25 becomes the shifted center feature vector after the update.

  Further, for example, when the shift vector 24 of FIG. 7 is generated in the process of the immediately preceding step S4, as shown in FIG. 9, the shift center feature vector 21 is a vector (2000, 2, 4, 3). And the shift vector 24 is a vector of (0,2, -2,0), so the update target is only the second component of “sweetness” and the third component of “spicy”, and as a result For example, the vector 26 of (2000, 2 + 2, 4 + (− 2), 3) = (2000, 4, 2, 3) becomes the shifted center feature vector after the update.

  Returning to FIG. 4, when the shift center feature vector updated in the process of step S <b> 6 is supplied from the shift parameter calculation unit 13 to the recommendation list generation unit 16, the process proceeds to step S <b> 6.

  In step S6, the recommendation list generation unit 16 updates the component value (weight value) of the shift target attribute in the effect vector so as to adjust the effect degree of the shift target attribute.

  The effect vector is the following vector. That is, as described above, when the similarity between two standard vectors is calculated, the component value of each component of the two standard vectors is used. Therefore, as a matter of course, the degree of similarity is determined according to the magnitude of the component value of each component. By the way, for the user, the attributes indicated by the components of the standard vector do not all have the same importance. In other words, there are some users who want to recommend items with particular emphasis on predetermined attributes or with the influence of predetermined attributes as small as possible. In order to meet such a demand, when the similarity between two standard vectors is calculated, the weight of each component of the two standard vectors is determined according to the importance to the user. Change it. That is, the component values of the respective components of the two standard vectors may be updated using weight values indicating importance to the user. In this case, a vector obtained as a result of substituting each component (attribute) weight value as a component value of the corresponding component is referred to as an effect vector.

  Accordingly, since the shift target attribute is an attribute that should be noticed by the user, the effect value (weight amount) of the shift target attribute is adjusted by updating the component value of the shift target attribute in the effect vector. By doing so, it becomes possible to recommend an item more suitable for the user (in this embodiment, wine). For this reason, in the “shift recommendation process” in FIG. 4, the process of step S <b> 6 is provided as a process of adjusting the effectiveness degree (weight amount) of the shift target attribute.

  The adjustment method of the effectiveness degree (weight amount) of the shift target attribute in step S6, that is, the update method of the component value of the shift target attribute in the effect vector is not particularly limited, but here, for example, the shift of the effect vector Assume that an update method is employed in which the original component value of the target attribute is multiplied by a predetermined adjustment amount and the multiplied value is used as the updated component value.

  In this case, for example, the shift target attribute is “price”, the effect vector is the vector 27 (0.7, 0.5, 0.5, 0.9) shown in FIG. 10, and the adjustment amount is “2”. Then, as shown in FIG. 10, a vector 28 of (0.7 × 2, 0.5, 0.5, 0.9) = (1.4, 0.5, 0.5, 0.9) is generated as the updated effect vector.

  Also, for example, the shift target attribute is “sweetness”, the effect vector is the vector 27 (0.7, 0.5, 0.5, 0.9) in FIG. 10 (FIG. 11), and the adjustment amount is “2”. Then, as shown in FIG. 11, a vector 29 of (0.7, 0.5 × 2, 0.5, 0.9) = (0.7, 1.0, 0.5, 0.9) is generated as the updated effect vector.

  Returning to FIG. 4, in step S <b> 7, the recommendation list generation unit 16 uses the effect vector updated in the process of step S <b> 6 and the shifted center feature vector updated in the process of step S <b> 5 and the item features of other items. The degree of similarity with the vector is calculated, and the recommendation level of other items is calculated based on the calculation result.

  The item feature vectors of other items are stored in the item feature vector storage unit 11 or 17 as described above.

  In addition, as described above, the calculation method of the recommendation level in step S7 is not particularly limited, but here, for simplicity of explanation, it is assumed that the similarity level becomes the recommendation level as it is.

  In step S8, the recommendation list generation unit 16 determines whether or not the recommendation level satisfies a predetermined criterion.

  In addition, although the predetermined reference | standard in step S8 is not specifically limited, suppose that a recommendation degree is more than a threshold value here.

  That is, if the recommendation level is less than the threshold value, it is determined in step S8 that the recommendation level does not satisfy the predetermined standard, and the process proceeds to step S10 without executing the process in step S9.

  On the other hand, when the recommendation level is equal to or higher than the threshold value, the recommendation list generation unit 16 determines that the recommendation level satisfies a predetermined criterion in step S8, and recommends other items to the user in step S9. After setting as a recommended item to be processed, the process proceeds to step S10.

  In step S10, the recommendation list generation unit 16 determines whether or not to calculate the recommendation level of another item.

  Note that the determination condition in step S10 is not particularly limited, but here, in the case where the similarity is calculated for all of the item feature vectors stored in the item feature vector storage units 11 and 17, in step S10. Assume that the determination is NO.

  In this case, if there is an item feature vector whose similarity is not calculated in the item feature vector storage unit 11 or 17, the process returns to step S7, and the subsequent processes are repeatedly executed.

  Thereafter, when the loop process of steps S7 to S10 is repeatedly executed for all of the item feature vectors stored in the item feature vector storage units 11 and 17, the process is determined to be NO in step S10. Proceed to step S11.

  In step S11, the recommendation list generation unit 16 generates a recommendation list including recommended items.

  Note that including a recommended item means that if the recommended item is wine, it does not necessarily include the wine itself, but includes information that can identify the wine and other information such as the recommendation level of the wine. It is.

  When the recommendation list is supplied from the recommendation list generation unit 16 to the presentation unit 18, the process proceeds to step S12.

  In step S12, the presentation unit 18 presents the recommended items included in the recommendation list to the user.

  In addition, when the recommended item is wine, the recommendation item is not necessarily presented as wine itself, but information included in the recommendation list, that is, information that can identify the wine and the recommendation degree of the wine It means that other information is presented.

  The “shift recommendation process” in the example of FIG. 4 has been described above.

  Incidentally, the “shift recommendation process” is not limited to the example of FIG. 4 described above, and various other processes can be employed. Specifically, for example, a process according to the flowcharts of FIGS. 12 and 13 may be employed as the “shift recommendation process”.

  That is, FIG. 12 and FIG. 13 are flowcharts for explaining a detailed example of the “shift recommendation process” different from FIG.

  The processes in steps S41 to S45 in FIG. 12 are basically the same processes as the processes in S1 to S5 in FIG. 4 described above.

  Therefore, the description of the processing from step S41 to S45 in FIG. 12 is omitted, and the processing after step S46 in FIG. 13 will be described below.

  In order to facilitate understanding of the “shift recommendation process” in the examples of FIGS. 12 and 13, hereinafter, the processes after step S46 in FIG. 13 will be referred to with reference to the specific examples in FIGS. I will explain. That is, each of FIGS. 14 to 17 shows a specific example of a processing result of a predetermined step in the “shift recommendation process” in the examples of FIGS.

  In step S46 of FIG. 13, the recommendation list generating unit 16 updates the component to be updated in step S45 (step S45) among the shift center feature vector updated in the process of step S45 of FIG. 12 and the item feature vectors of other items. The degree of similarity between the primary selection vectors) is calculated, and the recommendation level of other items is calculated based on the calculation result.

  Specifically, for example, it is assumed that the vector 25 (1400, 2, 4, 3) shown in FIG. 8 described above becomes the shifted center feature vector after the update in the process of step S45 immediately before. In this case, as shown in FIG. 8, since the shift vector used in the process of step S45 is a vector 23 of (0.7,0,0,0), the update target component is the first component of “price”. Only. Therefore, in this case, only the first component of “price” in the shifted center feature vector 25 after the update is extracted, and the resulting vector 25-1 in FIG. 14, that is, (price) = (1400). Is determined as the comparison source vector of the primary selection vector. Although not shown, for the other items as well, only the first component of “price” is extracted from the item feature vectors, and the resulting vector is compared with the primary selection vector comparison partner. Determined to be a vector. The similarity between the one-dimensional comparison source vector 25-1 (1400) and the one-dimensional comparison partner vector is calculated, and the recommendation degree of other items is calculated based on the calculation result. become.

  Further, for example, it is assumed that the vector 26 (2000, 4, 2, 3) shown in FIG. 9 described above becomes the shifted center feature vector after the update in the process of the previous step S45. In this case, since the shift vector used in the process of step S45 is the vector 24 of (0,2, -2,0), the update target component is the second component of “sweetness” and “spicy”. The third component. Accordingly, in this case, only the second component of “sweetness” and the third component of “spicyness” are extracted from the shifted center feature vector 26 after the update, and the resulting vector 26-1 in FIG. That is, a vector of (sweetness, hotness) = (4, 2) is determined as a comparison source vector of the primary selection vector. Although not shown, for other items as well, only the second component of “sweetness” and the third component of “spicyness” are extracted from the item feature vectors and obtained as a result. The vector is determined as a comparison partner vector of the primary selection vector. Then, the similarity between the two-dimensional comparison vector 26-1 (4, 2) and the two-dimensional comparison partner vector is calculated, and the recommendation degree of other items is calculated based on the calculation result. Will be.

  Returning to FIG. 13, when the recommendation list generation unit 16 calculates the recommendation level of other items in the process of step S46 as described above, in step S47, the recommendation list generation unit 16 determines whether the recommendation level satisfies a predetermined criterion.

  In addition, although the predetermined reference | standard in step S47 is not specifically limited, Suppose here that a recommendation degree is more than a threshold value.

  That is, if the recommendation level is less than the threshold value, it is determined in step S47 that the recommendation level does not satisfy the predetermined standard, and the process proceeds to step S51 without executing the processes in steps S48 to S50. That is, as a result of the primary selection, the other items have already been deselected as recommended item candidates, so the secondary selection is not executed and the process proceeds to step S51.

  On the other hand, if the recommendation level is greater than or equal to the threshold value, it is determined in step S47 that the recommendation level satisfies a predetermined criterion, and the processes in and after step S48 are executed. That is, as a result of the primary selection, other items have been selected as recommended item candidates, and therefore, the processing after step S48 is executed as processing corresponding to the secondary selection.

  In step S48, the recommendation list generation unit 16 includes components other than the update target component in step S45 among the shift center feature vector updated in the process of step S45 in FIG. 12 and the item feature vectors of other items ( The degree of similarity between the secondary selection vectors) is calculated, and the recommendation level of other items is recalculated based on the calculation result.

  Specifically, for example, it is assumed that the vector 25 (1400, 2, 4, 3) shown in FIG. 8 described above becomes the shifted center feature vector after the update in the process of step S45 immediately before. That is, it is assumed that the vector 25-1 (price) = (1400) shown in FIG. 14 described above is used as the primary selection vector in the process of step S46 immediately before. In this case, the components other than the update target component are “sweetness” of the second component, “spiciness” of the third component, and “fragrance” of the fourth component. Therefore, in this case, only the second component of “sweetness”, the third component of “spicy”, and the fourth component of “scent” of the shifted center feature vector 25 after the update are extracted, The resulting vector 25-2 in FIG. 16, that is, the vector of (sweetness, hotness, aroma) = (2,4,3) is determined as the comparison source vector of the secondary selection vector. And although not shown in figure, similarly about other items, the 2nd component of "sweetness", the 3rd component of "spicy", and the 4th of "fragrance" among the item feature vectors Only the components are extracted, and the resulting vector is determined as the comparison partner vector of the secondary selection vector. Then, the similarity between the three-dimensional comparison source vector 25-2 (2, 4, 3) and the three-dimensional comparison partner vector is calculated, and the recommendation degree of other items is calculated based on the calculation result. It will be calculated.

  Further, for example, it is assumed that the vector 26 (2000, 4, 2, 3) shown in FIG. 9 described above becomes the shifted center feature vector after the update in the process of the previous step S45. That is, it is assumed that the vector 26-1 (sweetness, hotness) = (4, 2) shown in FIG. 15 described above is used as the primary selection vector in the process of the previous step S46. In this case, the components other than the update target component are the “price” of the first component and the “scent” of the fourth component. Therefore, in this case, only the first component of “price” and the fourth component of “fragrance” are extracted from the updated shift center feature vector 26, and the resulting vector 26-2 in FIG. That is, the vector of (price, fragrance) = (2000, 3) is determined as the comparison source vector of the secondary selection vector. Although not shown, for other items as well, only the first component of “price” and the fourth component of “fragrance” are extracted from the item feature vectors and obtained as a result. The vector is determined as a comparison partner vector of the secondary selection vector. Then, the similarity between the two-dimensional comparison vector 26-2 (2000, 3) and the two-dimensional comparison partner vector is calculated, and the recommendation degree of other items is calculated based on the calculation result. Will be.

  Returning to FIG. 13, when the recommendation list generation unit 16 calculates the recommendation level of other items in the process of step S48 in this way, in step S49, the recommendation list generation unit 16 determines whether the recommendation level satisfies a predetermined criterion.

  In addition, although the predetermined reference | standard in step S49 is not specifically limited, Suppose here that a recommendation degree is more than a threshold value. Note that the threshold value used in the process of step S49 and the threshold value used in the process of step S47 described above may be the same value or different values.

  If the recommendation level is less than the threshold value, it is determined in step S49 that the recommendation level does not satisfy the predetermined standard, and the process proceeds to step S51 without executing the process in step S50. That is, as a result of the second selection, other items have been selected as recommended item candidates, so the process proceeds to step S51 without executing the process of step S50.

  On the other hand, if the recommendation level is greater than or equal to the threshold value, it is determined in step S49 that the recommendation level satisfies a predetermined criterion, and the process of step S50 is executed. In other words, as a result of the second selection, other items are formally selected as recommended items, and therefore the process of step S50 is executed to set the other items as recommended items. That is, in step S50, the recommendation list generation unit 16 sets other items as recommended items to be recommended to the user. Thereby, a process progresses to step S51.

  In step S51, the recommendation list generating unit 16 determines whether or not to calculate the recommendation level of another item.

  Note that the determination condition of step S51 is not particularly limited, but here, when the similarity is calculated for all of the item feature vectors stored in each of the item feature vector storage units 11 and 17, in step S51. Assume that the determination is NO.

  In this case, if there is an item feature vector whose similarity is not calculated in the item feature vector storage unit 11 or 17, the process returns to step S46, and the subsequent processes are repeated.

  Thereafter, when the loop process of steps S46 to S51 is repeatedly executed for all of the item feature vectors stored in the item feature vector storage units 11 and 17, the process is determined as NO in step S51. Proceed to step S52.

  In step S52, the recommendation list generating unit 16 generates a recommendation list including recommended items.

  When the recommendation list is supplied from the recommendation list generation unit 16 to the presentation unit 18, the process proceeds to step S53.

  In step S53, the presentation unit 18 presents the recommended items included in the recommendation list to the user.

  The “shift recommendation process” in the example of FIGS. 12 and 13 has been described above.

  In the “shift recommendation process” in the examples of FIGS. 12 and 13, as described above, the update target components (primary selection vectors) are used to perform primary selection of recommended items, and then other than the update target components These components (secondary selection vectors) are used to perform secondary selection of recommended items. Therefore, as compared with the “shift recommendation process” in the example of FIG. 4, the processing load of the information processing apparatus of FIG. 1 can be reduced and the processing time can be shortened. This is because both the primary selection vector and the secondary selection vector have lower dimensions than the shifted center feature vector.

  By the way, the series of processes described above can be executed by hardware, but can also be executed by software.

  In this case, the recommendation system in FIG. 3 can be configured by, for example, one or more personal computers shown in FIG. The example described above is an example in which the recommendation system in FIG. 3 is configured by a single information processing apparatus, and corresponding to this example, the recommendation system in FIG. 3 is, for example, a personal computer shown in FIG. One unit can be configured.

  In FIG. 18, a CPU (Central Processing Unit) 101 executes various processes according to a program recorded in a ROM (Read Only Memory) 102 or a program loaded from a storage unit 108 to a RAM (Random Access Memory) 103. To do. The RAM 103 also appropriately stores data necessary for the CPU 101 to execute various processes.

  The CPU 101, ROM 102, and RAM 103 are connected to each other via a bus 104. An input / output interface 105 is also connected to the bus 104.

  The input / output interface 105 includes an input unit 106 including a keyboard and a mouse, an output unit 107 including a display, a storage unit 108 including a hard disk, and a communication unit 109 including a modem and a terminal adapter. It is connected. The communication unit 109 controls communication performed with other devices (not shown) via a network including the Internet.

  A drive 110 is also connected to the input / output interface 105 as necessary, and a removable recording medium 111 made of a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is appropriately installed, and a computer program read from them is read. Are installed in the storage unit 108 as necessary.

  When a series of processing is executed by software, a program constituting the software executes various functions by installing a computer incorporated in dedicated hardware or various programs. For example, a general-purpose personal computer is installed from a network or a recording medium.

  As shown in FIG. 18, the recording medium including such a program is distributed to provide a program to the user separately from the apparatus main body, and a magnetic disk (including a floppy disk) on which the program is recorded. Removable recording media (package) consisting of optical disks (including compact disk-read only memory (CD-ROM), DVD (digital versatile disk)), magneto-optical disk (including MD (mini-disk)), or semiconductor memory (Media) 111, but also a ROM 102 on which a program is recorded and a hard disk included in the storage unit 108 provided to the user in a state of being incorporated in the apparatus main body in advance.

  In the present specification, the step of describing the program recorded on the recording medium is not limited to the processing performed in time series along the order, but is not necessarily performed in time series, either in parallel or individually. The process to be executed is also included.

  Further, in the present specification, the system represents the entire apparatus including a plurality of apparatuses and processing units.

It is a figure explaining the method of this invention. It is a figure which shows an example of the image for GUI utilized by the user when the method of this invention is embodied. It is a block diagram which shows the functional structural example of the information processing system or apparatus to which this invention is applied. 4 is a flowchart illustrating an example of “shift recommendation process” executed by the information processing apparatus of FIG. 3. It is a figure which shows the example of the shift center feature vector acquired as a result of the process of step S2 of FIG. It is a figure which shows the example of the shift vector produced | generated as a result of the process of step S4 of FIG. It is a figure which shows the example different from FIG. 6 of the shift vector produced | generated as a result of the process of step S4 of FIG. FIG. 7 is a diagram illustrating an example of a shift center feature vector that is updated as a result of using the shift vector of FIG. 6 in the process of step S5 of FIG. FIG. 8 is a diagram illustrating an example of a shift center feature vector updated as a result of using the shift vector of FIG. 7 in the process of step S5 of FIG. 4. It is a figure which shows the example of each effect vector before the process of step S6 of FIG. 4, and after a process. It is a figure which shows the example different from FIG. 10 of each effect vector before the process of step S6 of FIG. 4, and after a process. FIG. 5 is a flowchart for explaining an example of “shift recommendation process” executed by the information processing apparatus of FIG. 3 different from FIG. FIG. 5 is a flowchart for explaining an example of “shift recommendation process” executed by the information processing apparatus of FIG. 3 different from FIG. FIG. 14 is a diagram illustrating an example of a primary selection vector generated as a result of using the post-update center feature vector of FIG. 8 in step S46 of FIG. 13; FIG. 14 is a diagram showing an example of a primary selection vector generated as a result of using the post-update center feature vector of FIG. 9 in step S46 of FIG. FIG. 15 is a diagram illustrating an example of a secondary selection vector used in step S48 of FIG. 13 and corresponding to the primary selection vector of FIG. 14; FIG. 16 is a diagram illustrating an example of a secondary selection vector used in step S48 of FIG. 13 and corresponding to the primary selection vector of FIG. It is a figure which shows the hardware structural example of the information processing apparatus of FIG.

Explanation of symbols

  1, 2 GUI images, 11 Item feature vector storage unit, 12 User request acquisition unit, 13 Shift parameter calculation unit, 14 Inter-attribute correlation vector storage unit, 15 Shift vector storage unit, 16 Recommendation list generation unit, 17 Item feature vector Storage unit 18 presentation unit 21 pre-update shift center feature vector, 22-A, 22-B inter-attribute correlation vector, 23, 24 shift vector, 25, 26 post-update shift center feature vector, 25-1, 26-1 primary Selection vector, 25-2, 26-2 Secondary selection vector, 27 Effective vector before update, 28, 29 Effective vector after update, 101 CPU, 102 ROM, 103 RAM, 108 storage unit, 111 removable recording medium, I1 Thru I9 item feature vector, before I01 update Shift center feature vector, I02, I03 After update shift center feature vector

Claims (8)

A vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of N components is a standard vector,
As a comparison source vector, the standard vector that is the basis of comparison,
Each of the plurality of normal vectors indicating the characteristics of the plurality of items is used as a comparison partner vector.
In an information processing apparatus that determines a recommended item to be recommended to the user from among a plurality of items, using a comparison source vector and the comparison partner vector for each of a plurality of items,
Of the N components of the comparison source vector, the component value of the component corresponding to the first attribute designated to be changed by the user is updated using the first update amount designated by the user. Update means;
A similarity between the comparison source vector itself or a part thereof updated by the updating means and each of the comparison partner vectors for a plurality of items or a part thereof is calculated, and the similarity satisfies a predetermined condition An information processing apparatus comprising: recommendation means for determining an item corresponding to the comparison partner vector as the recommended item. The updating means further calculates a component value of a second component corresponding to a second attribute among the N components of the comparison source vector that is correlated with the first attribute. The information processing apparatus according to claim 1, wherein updating is performed using a second update amount determined according to an update amount and the correlation. When a plurality of attributes are designated as the first attribute by the user, and each of the first update amounts is individually designated for each of the plurality of first attributes,
The updating unit is configured to calculate a component value of each of a plurality of components corresponding to each of the plurality of first attributes among the N components of the comparison source vector, and a corresponding first update amount. The information processing apparatus according to claim 1, wherein each of the information processing apparatuses is updated by using the information processing apparatus. The recommendation means further uses the weights set in advance for each of the N components of the standard vector to compare the comparison source vector itself or a part thereof and the comparison partner vector itself or The information processing apparatus according to claim 1, wherein the similarity with the part is calculated. The recommendation means further applies the first component corresponding to the first attribute out of the weights set in advance for each of the N components of the standard vector. Adjusting the amount of weights set in advance, and calculating the similarity between the comparison source vector itself or a part thereof and the comparison partner vector itself or a part thereof using the adjusted weights. The information processing apparatus according to claim 4. The recommendation means is:
Extracting each of one or more components including at least the first component from the N components of the comparison source vector, and generating a first vector composed of the extracted one or more components,
Of N components of one vector to be compared among a plurality of comparison partner vectors, one or more same components as one or more of the components to be extracted by the comparison source vector Each is extracted to generate a second vector of one or more extracted components;
The similarity between the first vector and the second vector is calculated, and when the calculated similarity satisfies a predetermined condition, an item corresponding to the second vector is determined as the recommended item. The information processing apparatus according to claim 1. A vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of N components as a standard vector,
As a comparison source vector, the standard vector that is the basis of the comparison,
Each of the plurality of normal vectors indicating the characteristics of a plurality of items is used as a comparison partner vector.
In the information processing method of the information processing apparatus for determining a recommended item to be recommended to the user from among the plurality of items using the comparison source vector and the comparison partner vector for each of a plurality of items,
An update step of updating the component value of the component corresponding to the attribute designated to be changed by the user among the N components of the comparison source vector using the update amount designated by the user;
A similarity between the comparison source vector itself or a part thereof updated by the processing of the updating step and each of the comparison partner vectors for a plurality of items or a part thereof is calculated, and the similarity is a predetermined condition A recommendation step of determining an item corresponding to the comparison partner vector satisfying the recommended item as the recommended item. A vector in which each of N attributes (N is an integer value of 1 or more) of an item is a predetermined one of N components is a standard vector,
As a comparison source vector, the standard vector that is the basis of comparison,
Each of the plurality of normal vectors indicating the characteristics of the plurality of items is used as a comparison partner vector.
A program that is executed by a computer that controls a recommendation device that determines a recommended item to be recommended to the user from among a plurality of items using a comparison source vector and the comparison partner vector for each of a plurality of items. There,
An update step of updating a component value of a component corresponding to an attribute designated to be changed by a user among N components of the comparison source vector using an update amount designated by the user;
The degree of similarity between the comparison source vector itself or a part thereof updated by the processing of the updating step and each of the comparison partner vectors for a plurality of items or a part thereof is calculated, and the similarity is a predetermined condition A recommendation step of determining an item corresponding to the comparison partner vector satisfying the recommended item as the recommended item.

Images