JP2012053863A - Information recommendation processor and control program thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable accurate information recommendation by more appropriately managing the tastes of an information user.SOLUTION: The information recommendation processor is configured to, in the case of updating intensity information set in the node of taste model information, specify a low-order concept related with a selected store via a content link as a concept under consideration, and to specify a concept directly and indirectly related with the specified concept under consideration via a link as a related concept. The processor increases the intensity information set in the specified concept under consideration and related concept by weighting the increment and stores the increased intensity information as updated intensity information. It is determined whether or not the update processing of the intensity information is periodically performed, and, when at least one of the intensity information is updated, the intensity information set in each node of the taste model information is uniformly subtracted by every constant quantity and, when a node in which the intensity information after the subtraction processing is set to "0" is generated, data of the node is deleted from the taste model information.

Description

  The present invention relates to an information recommendation processing apparatus for providing recommendation information corresponding to a preference model to an information user and a program thereof.

  In recent years, when an information provider provides information to an information user on a computer network, information suitable for each information user is selected, or information is rearranged in an order suitable for each information user. An information recommendation service to be displayed has been proposed. As information to recommend, advertising information, news information, and restaurant information are promising, for example.

  As a method for realizing this kind of service, the present inventors use a conceptual structure consisting of a directed acyclic graph, that is, a graph having a characteristic that a link has a direction and does not follow the link and return to the original node. We propose a method to do this. This method uses information by setting strength information indicating the strength of preference for each information user for each concept (also referred to as a node or class) or a link indicating the relationship between concepts in the concept structure. A user preference model. Then, based on the operation history of the information user, etc., the strength information of the preference model is updated according to a predetermined strength update rule, and information suitable for the information user based on the preference model with the updated strength information ( (Referred to as Patent Document 1 or Non-Patent Document 1).

JP 2005-149340 A

IEICE Technical Report, Koji Ito et al., "Examination of Information User Understanding Model for Action Support Services", vol. 109, no. 272, LOIS2009-58, pp. 121-128, November 2009

However, the technique described in Patent Document 1 or Non-Patent Document 1 has the following problems to be improved.
The first problem is that, when the preference model is updated using the above method, the update target of the strength information is limited to the concept directly related to the selected and operated information, and the strength of the nearby concept in the concept structure is updated. It is not targeted. For this reason, the preference management of information users has to be limited, and information is not recommended in consideration of a general concept positioned at a higher level.

For example, if an information user uses a store, and the concept “spaghetti meat sauce” is associated with the store, the above method updates only the strength set in the concept of “spaghetti meat sauce”. . For this reason, when only the “spaghetti meat sauce” is selected on the preference model, it cannot be estimated that the user likes “noodles”, but prefers “spaghetti meat sauce” or “noodles” such as “soba”. Even if the tendency is remarkable, store information of “meat udon” included in the same “noodles” is not recommended.
The first problem described above can also occur when intensity information is set for a link leading to the concept instead of setting the intensity information for the concept.

  The second problem is that, when the preference model is updated using the above method, the preference model cannot be updated following the transition of the information user's preference. For example, even when a certain information user's food preference transitions from “Fried food lover” to “Fried food hate”, the preference “Fried food lover” continues to be maintained. For this reason, a preference model may become inaccurate and a recommendation result may also become inaccurate.

  The third problem is that, when the preference model is updated using the above-described method, even when the information user uses the store information by mistake, the history information is not erased but remains semipermanently. As a result, the preference model is inaccurate and the recommendation result is also inaccurate. In addition, since unnecessary data is accumulated in the database indefinitely, the system operation cost increases.

  The present invention has been made paying attention to the above circumstances, and an object of the present invention is to provide an information recommendation processing apparatus and its program capable of more accurately managing information user preferences and enabling accurate information recommendation There is to do.

  In order to achieve the above object, one aspect of the present invention uses a concept structure including a directed acyclic graph in which a plurality of hierarchized concepts are associated with links and recommendation candidate information is associated with the above concept. Preference model information reflecting the user's preference is created by setting strength information corresponding to the user's preference in each of the above concepts or links leading to the concept, and this is stored. Then, when updating the strength information of the stored preference model information according to the selection operation of the recommendation candidate information by the user, the recommendation candidate information that has been selected and operated is directly associated with the stored preference model information. The identified concept is identified as the focused concept, and the concept directly or indirectly related to the identified focused concept is identified as the related concept, and the identified focused concept and the related concept or between these concepts are identified. The strength information set in the link to be associated is increased.

  Therefore, when the user selects information, the related concept or link that is directly or indirectly related to the target concept is not limited to the strength of the target concept or link directly related to the selected information. The strength of is also updated. For this reason, it is possible to manage the user's preference in a state reflecting not only the subordinate concept but also the superordinate concept, so that more appropriate information can be recommended to the user.

The first aspect of the present invention is also characterized by comprising the following various aspects.
The first aspect further includes a correction unit that corrects the intensity information included in the stored preference model information, and the correction unit performs the following processing. That is, when the preference model information has a plurality of conceptual structures created from different viewpoints, the concept or link having the maximum strength information is extracted from each of the plurality of conceptual structures. Then, the concept or link strength information extracted from each of the plurality of conceptual structures is corrected so that the value thereof is a ratio arbitrarily set in advance between the concepts or links, and In order to maintain the ratio of intensity information between the extracted concept or link included in the concept structure and the other concept or link, the ratio other than the above is corrected according to the intensity information after correction. The concept or link strength information is corrected.

  In this way, when learning intensity information based on the selection history by the user as described above, even if the appearance probability of the concept is biased depending on the number of choices between multiple viewpoints with different concept structures It is possible to reduce the bias of the intensity information among the plurality of viewpoints and more appropriately select the recommendation candidate information for the user. Furthermore, since the correction ratio is set to an arbitrary value when the intensity between the concepts is corrected, there is an advantage that the intensity of the concept in each viewpoint can be corrected according to the importance degree between the viewpoints.

  Further, the concept or link strength information extracted from each of the plurality of concept structures may be corrected so that the value thereof is the same between the concepts or links. In this way, it is possible to easily correct the bias of concept information or link strength information between viewpoints regardless of the appearance frequency of the concept between viewpoints.

  In the second aspect, when the strength information is increased, the strength information set in the target concept or the link connected thereto is increased by the first increase width, and is set in the related concept or the link connected thereto. The intensity information is increased by a second increase width smaller than the first increase width.

  In this way, the user's preference is more strongly reflected on the concept or link that is directly related to the information selected by the user, while the related concept or link that is indirectly related to the selected information. It is possible to reflect the user's preference weakly. For this reason, it becomes possible to manage a user's preference more appropriately on a preference model.

  It should be noted that the increase width of the intensity information is not limited to the first increase width and the second increase width, and the third, fourth,. An increase width may be prepared, and the third, fourth,..., Nth increase widths may be sequentially used as the strength of the relevance to the selection information becomes weaker.

  In the third aspect, it is determined whether or not the strength information set for each concept or link of the stored preference model information has been updated within a preset period, and at least one of the strength information is within the set period. When it is determined that the information has been updated, the strength information set for each concept or link of the stored preference model information is uniformly reduced within a preset reduction range.

  If it does in this way, it will become possible to follow a user's preference transition and to update a preference model. For example, when the taste of a certain information user's food changes from “Fried food lover” to “Fried food hate”, the preference model also changes to reflect this preference transition. For this reason, the frequency with which the information related to “fried food” is recommended gradually decreases, thereby enabling information recommendation reflecting the transition of the user's preference. In addition, when the user does not use the information, forgetting is stopped, that is, it is possible to prevent a problem that forgetting exceeds learning in a user whose usage frequency is low.

In the fourth aspect, every time the intensity information is updated by the means for increasing the intensity information, the intensity information set for each concept or link of the stored preference model information is reduced in advance. It is intended to decrease uniformly.
In this way, as in the second aspect, it is possible to update the preference model following the transition of the user's preference, and the strength information is triggered every time the strength information is updated. Therefore, the preference transition can be reflected in the preference model with good responsiveness according to the user's information selection operation.

In the fifth aspect, it is determined whether or not the intensity information after the reduction process is less than or equal to a preset minimum value. When intensity information that is less than or equal to the minimum value is detected, the intensity information is The set concept is deleted from the stored preference model information.
In this way, data corresponding to the concept that the intensity information has decreased to a predetermined value or less (for example, zero) is deleted from the preference model. For this reason, even if an information user selects information by mistake and the strength of the concept related to the information is set high, the data of this concept will not remain semipermanently and will be automatically deleted later. The Therefore, it is possible to maintain the preference model in an accurate state, and it is possible to eliminate the problem that the useless data is accumulated in the database indefinitely, thereby reducing the system operation cost.

  In other words, according to the present invention, it is possible to provide an information recommendation processing apparatus and its program capable of appropriately managing information user preferences and enabling accurate information recommendation.

1 is a schematic configuration diagram of a system including an information recommendation processing device according to a first embodiment of the present invention. The block diagram which shows the function structure of the information recommendation processing apparatus concerning 1st Embodiment of this invention. The flowchart which shows the preference model update process sequence by the information recommendation processing apparatus shown in FIG. 2, and its processing content. The figure which shows an example of the conceptual structure of a preference model. The figure which shows the example which provided intensity | strength information to the link of the preference model shown in FIG. The figure which shows the example which provided intensity | strength information to the node of the preference model shown in FIG. The figure which shows the specific method of the attention concept and related concept in the preference model shown in FIG. The figure which shows an example of an intensity | strength information reduction process among the preference model update process procedures shown in FIG. The block diagram which shows the function structure of the information recommendation processing apparatus concerning 2nd Embodiment of this invention. The flowchart which shows the preference model update process sequence by the information recommendation processing apparatus shown in FIG. 9, and its processing content. The block diagram which shows the function structure of the information recommendation processing apparatus concerning the 3rd Embodiment of this invention. The flowchart which shows the procedure of the information recommendation process by the information recommendation processing apparatus shown in FIG. The figure which shows an example of the data memorize | stored in the intensity | strength database of the information recommendation processing apparatus shown in FIG. The figure which shows an example of the data memorize | stored in the information database of the information recommendation processing apparatus shown in FIG. The figure which shows an example of the preference model which combined the several viewpoint. The figure which shows the 1st example of the correction result of the intensity | strength information in the preference model shown in FIG. The figure which shows the 2nd example of the correction | amendment result of the intensity | strength information in the preference model shown in FIG.

Embodiments according to the present invention will be described below with reference to the drawings.
(First embodiment)
FIG. 1 is a schematic configuration diagram of a system including an information recommendation processing apparatus according to the first embodiment of the present invention. This system includes an information recommendation processing device SVa and an information storage device SD that stores preference model information used by the information recommendation processing device SVa for information recommendation processing. Communication between the information recommendation processing device SVa and the client terminals MS1 to MSn used by the user for receiving the information recommendation service via the communication network NW is enabled.

  Each of the client terminals MS1 to MSn includes a personal computer, a mobile phone, a smartphone, a PDA (Personal Digital Assistant), and the like, and has a communication processing function necessary for receiving an information recommendation service with the information recommendation processing device SVa. ing. As this communication processing function, for example, a Web browser is used.

  The communication network NW includes an IP (Internet Protocol) network and an access network for accessing the IP network. As the access network, a public wired telephone network, a mobile telephone network, a LAN (Local Area Network), a wireless LAN, a CATV (Cable Television) network, or the like is used.

  The information storage device SD includes a database server using, for example, a personal computer or a workstation, and includes a database 3 as shown in FIG. The database 3 stores information representing a conceptual structure created in common for each information user. Also, in the database 3, preference model information in which information indicating the strength of preference of the information user is set for each concept of the above concept structure for each information user is associated with the identification information (user ID) of the information user. Is remembered. The information storage device SD transmits / receives information to / from the information recommendation processing device SVa via the communication network NW.

  The information user's preference model information is based on the conceptual structure as described above. The conceptual structure is composed of a directed acyclic graph that allows multiple inheritance. For example, a plurality of hierarchical concepts (nodes or classes) as shown in FIG. It consists of information related as an instance. In each of the above concepts, for example, as shown in FIG. 5, information indicating the intensity of preference is set for each information user, and this is information indicating the preference model of the information user.

  By the way, the information recommendation processing device SVa is composed of, for example, a Web server using a personal computer or a workstation, and is configured as follows. FIG. 2 is a block diagram showing the functional configuration.

  That is, the information recommendation processing device SVa includes a communication interface unit 1 and a control unit 2a. The communication interface unit 1 transmits and receives information to and from the client terminals MS1 to MSn according to a communication protocol defined by the communication network NW under the control of the control unit 2a.

  The control unit 2a includes a processor and a memory, and includes a strength update control program group 21 and a strength reduction / deletion control program group 22 as control functions necessary for realizing the present invention.

The strength update control program group 21 includes an information reception unit 211, a concept concept specification unit 212, a related concept specification unit 213, and a strength information update unit 214.
When the information user performs an operation of selecting, for example, a restaurant on the client terminals MS1 to MSn, the information reception unit 211 receives information representing the selected store from the client terminals MS1 to MSn via the communication interface unit 1. Process to accept.

  The focused concept identification unit 212 performs processing for identifying a concept having a link with respect to the selection information received by the information receiving unit 211 as the focused concept.

  The related concept identification unit 213 performs processing for identifying a plurality of concepts having links directly and indirectly to the target concept identified by the target concept identification unit 212 as related concepts.

  The strength information update unit 214 sets the intensity of preference set for the target concept specified by the target concept specification unit 212 and the plurality of related concepts specified by the related concept specification unit 213 with a preset increase range. The process to increase by. The increase width is set to be smaller as the distance from the selection information becomes longer, that is, as the strength of the relevance to the selection information becomes weaker, when the increase width with respect to the strength of the concept of interest is the maximum value.

  The intensity reduction / deletion control program group 22 includes a periodic activation unit 221, a reduction process execution determination unit 222, a reduction process execution unit 223, and an unnecessary data deletion unit 224.

  The regular activation unit 221 generates an activation signal for executing the intensity information reduction / deletion processing at predetermined intervals (for example, once a day).

  The decrease process execution determination unit 222 performs the update process of the intensity information included in the preference model information by the intensity information update unit 214 during the period from the execution of the decrease process of the previous intensity information to the execution of the same process this time. Judgment is made or not.

  The reduction processing execution unit 223 is included in the corresponding preference model information read from the database 3 of the information storage device SD when any of the intensity information is updated based on the determination result by the reduction processing execution determination unit 222. The intensity information to be reduced is uniformly reduced by a certain amount, and the preference model information after the intensity reduction process is stored in the database 3. On the other hand, if no intensity information has been updated, the process ends without performing the reduction process.

  The unnecessary data deletion unit 224 determines whether or not there is a concept in which the preference intensity information is “0” in the preference model information after the uniform reduction processing is performed by the reduction processing execution unit 223. When a concept whose intensity information is “0” is detected, a process of deleting the concept data from the preference model information is executed.

Next, the update, reduction, and deletion processing of the intensity information by the information recommendation device SVa configured as described above will be described.
(1) Creation of preference model information The information recommendation processing device SVa first creates a common conceptual structure for each information user and stores the information in the database 3 in the information storage device SD. For example, when creating a conceptual structure related to food and drink, the concept of “restaurant” is set as the highest level concept as shown in FIG. 5, and “Japanese food”, “noodles”, and “meat food” are added to the nodes of this highest level concept. A plurality of intermediate concept nodes are related by links. Further, a plurality of subordinate concept nodes such as “sushi”, “soba”, “spaghetti meat sauce”, and “steak” are associated with these intermediate concept nodes by links, and a store name is associated with these subordinate concepts as an instance.

Information representing the strength of preference of the information user is set for each information user according to the selection operation history and the like for each information user, and each concept is associated with each content (instance). Strength information is also set for the content link. In this case, the class importance of the subordinate node is obtained by multiplying the strength set for the node by the attention rate, and the importance of each instance is linked to the strength set for the associated subordinate concept node. Multiplied by the strength set. The above “attention rate” is described in Non-Patent Document 1.
An instance may be associated with a node other than the subordinate concept via a content link. In this case, the importance of the instance is the strength set for the content link to the strength set for the concept node with which the instance is associated. Multiplied by.
In addition, it is not always necessary to create a conceptual structure in common for information users. For each information user, a unique conceptual structure reflecting its attribute information is created and stored in association with the ID of the information user. You may make it do. In this case, according to the instance selection operation history by the information user, strength information is set for the conceptual structure of the information user, and information indicating the conceptual structure in which the strength information is set is the preference of the information user. Information representing the model.

  Further, as another configuration example of the preference model, instead of setting the strength information in the node, it is conceivable to set the strength information in a link that associates the concepts (nodes). FIG. 6 shows an example of the setting. In this case, the class importance of the subordinate node is the strength set for the link if there is one input link, and if there are a plurality of input links, these links are used. It becomes the average value of the intensity set in. The importance of the instance is obtained by multiplying the class importance of the subordinate concept node by the strength set in the content link.

(2) Updating preference intensity information The intensity information set in the preference model is updated as follows. FIG. 3 is a flowchart showing the processing procedure and processing contents. Here, description will be made assuming that preference model information has already been created and stored for each information user according to the past selection operation history.
That is, it is assumed that a certain information user performs a restaurant selection operation on the client terminal MS1. If it does so, the information showing this selected store will be received via the communication interface part 1 under control of the information reception part 211 of the control unit 2a (step S31).

When the store selection information is received, the control unit 2a first activates the target concept specifying unit 212 in step S32, and from the database 3 in the information storage device SD under the control of the target concept specifying unit 212. Read preference model information associated with the information user of the transmission source. Then, in the read preference model information, the subordinate concept associated with the selected store via the content link is specified as the concept of interest. For example, if “store 4” is selected in the preference model shown in FIG. 5, the subordinate concept “spaghetti meat sauce” associated with this “store 4” is selected as the concept of interest.
If the preference model information corresponding to the information user of the transmission source has not been created yet, information representing the conceptual structure created in advance for each information user is stored in the database in the information storage device SD. 3 is read out. Then, a process for setting the strength information for the corresponding concept is performed on the conceptual structure information in accordance with the selection operation, and the conceptual structure information in which the strength information is set is subsequently managed as preference model information of the information user. Is done.

  Next, the control unit 2a activates the related concept specifying unit 213 in step S33, and directly and indirectly associates with the target concept specified in the preference model information under the control of the related concept specifying unit 213. Identify the concept as a related concept. Specific rules for this related concept will be described in detail below.

  When the proposition “If A, then B” holds, let B be the parent concept of A and A be the child concept of B. In addition, the child concept of A is also included in the child concept of B, and the following is repeated recursively. At this time, the parent concept of the target concept, the child concept of the parent concept of the target concept excluding the target concept, and the child concept of the target concept are set as the concept of the relation ranking “1”. Here, the relation rank represents the depth of the relation between the concept of interest and an arbitrary concept, and the lower the rank, the deeper the relation. In addition, a concept obtained by removing the concept of the relation ranking “1” and the concept of interest from the parent concept of the parent concept of the concept of interest and the child concept of the parent concept of the parent concept of the concept of interest is the related concept of the relation ranking “2” Similarly, the concept of the related ranking “N” is defined. When the related concept of the related rank α and the related concept of the related rank β are the same and α> β, the concept is regarded as a related concept of the related rank β.

  For example, in FIG. 7, when focusing on the concept M, the parent concept (F, G) of the target concept M, the child concept (K, L, N) of the parent concept of the target concept M excluding the target concept M, and the target A child concept (Q, R, S, T) of the concept M is set as a related concept of the related rank “1”. Further, from the parent concept (B, C) of the parent concept of the target concept M and the child concept of the parent concept of the parent concept of the target concept M, the related concept “1” and the target concept M are removed (E, J, H, O) are regarded as related concepts of the related rank “2”. Furthermore, from the parent concept (A) of the parent concept of the parent concept of the target concept M and the child concept of the parent concept of the parent concept of the parent concept of the target concept M, the related concept of the relation rank “1”, the relation rank “2” The related concept (D, I, P) excluding the related concept and the concept of interest M is the related concept of the related rank “3”.

  Subsequently, the control unit 2a activates the strength information update unit 214 in step S34, and under the control of the strength information update unit 214, the target concept M specified by the target concept specification unit 212 and the related concept specification The intensity of preference set in a plurality of related concepts (A to T excluding M) specified by the unit 213 is increased by a preset increase range. At this time, when the increase width with respect to the strength of the concept of interest is “W”, the increase width is gradually smaller than the above “W” for the related concepts of the related rankings “1”, “2”, and “3”. Is set as follows.

  For example, it is assumed that the information user uses the information of “Store 2” in the preference model information shown in FIG. In this case, the intensity information of the concept of interest “Soba” associated with “Store 2” via the content link is updated with the increase width “W”. In addition, the strength information of the concept group “Japanese food”, “noodles”, “sushi”, “meat udon”, and “spaghetti meat sauce” that has a relationship of “1” to the focused concept “soba” has increased. It is updated with an increase width set smaller than the width “W”. Similarly, the intensity information is also updated for the concept groups having the relations “2” and “3” in relation to the increase width that is set so that the increase width gradually decreases.

Therefore, by recommending stores using the preference model with updated strength information in this way, information users frequently find stores related to subclasses of “Japanese food” such as “sushi” and “soba”. Even if you do not use a store related to “meat udon”, it is possible to recommend “meat udon” because the general preference that you like “Japanese food” is understood. . In addition, if the information user frequently uses stores related to the lower class of “noodles” such as “soba” and “spaghetti meat sauce”, the preference that they like “noodles” is also known. “Meat udon” which is “Japanese food” and “noodles” is highly recommended.
The strength update process in steps S31 to S34 described above is repeated each time operation information is received from the accessing client terminal MS1.

(3) Uniform reduction of intensity information In the standby state, the control unit 2a monitors whether or not a predetermined time has elapsed by the periodic activation unit 221 in step S36. In this state, when a certain time, for example, 24 hours elapses, the reduction process execution determination unit 222 is activated. Then, under the control of the reduction process execution determination unit 222, the strength information update history of the past 24 hours is read from the database 3 for each information user in step S37, and the past 24 hours are read based on the strength information update history. In step S38, it is determined whether or not the information user has updated the strength information. As a result of the determination, if the update process of the intensity information has not been performed at all in the past 24 hours, it is determined that the decrease process is unnecessary and the process returns to the standby state.

  Note that the fixed time monitored by the regular activation unit 221 is not limited to 24 hours, and even if it is set to any shorter time (for example, several hours, 1 hour, or minutes), it is a long time exceeding 24 hours. (For example, the time in units of weeks or months) may be set.

  In contrast, it is assumed that at least one piece of intensity information has been updated within the past 24 hours. In this case, the control unit 2a proceeds to step S39 to activate the reduction processing execution unit 223, reads the preference model information from the database 3 under the control of the reduction processing unit 223, and each node of the preference model information Is uniformly subtracted by a certain amount, for example, “1”. For example, if strength information as shown in FIG. 8A is set in each node of the preference model information, the strength information of each node is uniformly subtracted by “1” as shown in FIG. 8B. To do.

(4) Deletion of Unnecessary Data The control unit 2a then activates the unnecessary data deletion unit 224. Under the control of the unnecessary data deletion unit 224, first, in step S40, the node after the uniform subtraction processing of the intensity information It is determined whether or not there is one whose intensity information value is “0”. As a result of this determination, when a node having a strength information value of “0” is found, the data of the node is deleted from the preference model information in step S41.

For example, if the preference model information after the uniform subtraction processing of intensity information is shown in FIG. 8B, the intensity information of “spaghetti meat sauce” is “0”. Delete from preference model information. Then, the old preference model previously stored in the database 3 is updated to the preference model after the deletion process.
The unnecessary data deletion process is not limited to the case where the intensity information value is “0”, and the intensity information value is smaller than a preset threshold value other than “0”. It may be performed in the case of.

  As described above in detail, in the first embodiment, when the strength information set in the node of the preference model is updated in accordance with the store selection operation by the information user, a content link is added to the selected store. The subordinate concepts associated with each other are identified as the concept of interest, and the concepts directly and indirectly associated with the identified concept of interest by the link are identified as related concepts. Then, the intensity information set in the identified concept of interest and related concept is increased after weighting the increase amount, and is stored as the updated intensity information after the increase.

  Therefore, by recommending stores using the preference model with updated strength information for not only the concept of interest but also related concepts around it, the preference for subordinate concepts such as “sushi” and “soba” In addition to strength and weakness, the store should be considered not only for preference for general medium concepts such as “Japanese food” and “noodles”, but also for other low-level concepts indirectly related through these intermediate concepts. It becomes possible to recommend.

  Also, for example, it is determined whether or not the strength information has been updated in the past 24 hours at a rate of once a day. If at least one of the strength information is updated, each node of the preference model information The set intensity information is uniformly subtracted by a certain amount, for example, “1”. For this reason, it becomes possible to update a preference model following a user's preference transition.

  Further, it is determined whether or not there is a node whose intensity information value is “0” among the nodes after the uniform subtraction processing of the intensity information, and a node whose intensity information value is “0” is determined. If found, the data of the node is deleted from the preference model information. For this reason, the data corresponding to the node whose intensity information has been reduced to “0” is deleted from the preference model information. For this reason, even if an information user selects information by mistake and thereby the strength of the concept related to the information is set high, the data of this concept does not remain semipermanently and is automatically deleted later. be able to. Accordingly, it is possible to always maintain the preference model information in an accurate state, and further, it is possible to eliminate the problem that the useless data is continuously stored in the database 3, thereby reducing the system operation cost.

(Second Embodiment)
While the first embodiment periodically performs uniform subtraction processing of intensity information and deletion processing of unnecessary data, the second embodiment of the present invention has a preference according to the selection operation of the information user. A uniform subtraction process of intensity information and a process of deleting unnecessary data are performed using the update of the model intensity information as a trigger.

  FIG. 9 is a block diagram showing a functional configuration of the information recommendation processing device SVb and the information storage device SD according to the second embodiment of the present invention. In the figure, the same parts as those in FIG.

  The control unit 2 b includes a strength update control program group 21 and a strength reduction / deletion control program group 23. Of these, the strength reduction / deletion control program group 23 includes a reduction processing execution unit 231 and an unnecessary data deletion unit 232 as control functions necessary for carrying out the second embodiment of the present invention.

  The reduction processing execution unit 231 is activated when a strength information update end notification is generated from the strength information update unit 214 of the strength update control program group 21, and is read from the database 3 of the information storage device SD for each information user. The intensity information of the preference model information to be reduced is uniformly reduced by a certain amount, and the preference model information after the intensity reduction process is stored in the database 3. On the other hand, if no intensity information has been updated, the process ends without performing the reduction process.

  The unnecessary data deleting unit 232 determines whether or not there is a concept in which the preference intensity information is “0” in the preference model information after the uniform reduction processing is performed by the reduction processing execution unit 231. When a concept whose intensity information is “0” is detected, a process of deleting the concept data from the preference model information is executed. Also in this embodiment, the unnecessary data deletion process may be performed when the value of the intensity information is smaller than a preset threshold value other than when the value is “0”.

Next, a preference model strength information update operation by the information recommendation processing device SVb configured as described above will be described. FIG. 10 is a flowchart showing the processing procedure and processing contents.
If the information user performs a restaurant selection operation at the client terminal MS2, information representing the selected store is received via the communication interface unit 1 under the control of the information receiving unit 211 of the control unit 2b. (Step S51).

  When the store selection information is received, the control unit 2b first activates the target concept specifying unit 212 in step S52, and from the database 3 in the information storage device SD under the control of the target concept specifying unit 212. Read preference model information associated with the information user of the transmission source. Then, in the read preference model information, the subordinate concept associated with the selected store via the content link is specified as the concept of interest. Subsequently, in step S53, the related concept specifying unit 213 is activated, and under the control of the related concept specifying unit 213, the concept directly and indirectly related to the target concept specified in the preference model information is displayed. Identify as related concept. The same rules as those described with reference to FIG. 7 in the first embodiment are also used for the related concept specifying rules.

  Next, the control unit 2b activates the strength information update unit 214 in step S54, and under the control of the strength information update unit 214, the target concept specified by the target concept specification unit 212, and the related concept specification unit The intensity of the preference set in the plurality of related concepts specified by 213 is increased by a preset increase range. At this time, as described in the first embodiment, when the increase width with respect to the strength of the concept of interest is “W”, the increase width is related to the related concepts “1”, “2”, and “3”. On the other hand, it is set to be gradually smaller than the “W”.

  Now, it is assumed that the update process of the intensity information by the intensity information update unit 214 is completed, and the occurrence of the end notification is detected in step S55. Then, the control unit 2b proceeds to step S56 and activates the reduction process execution unit 231. Under the control of the reduction process execution unit 231, the preference unit information is read from the database 3, and each node of the preference model information is read out. Is uniformly subtracted by a certain amount, for example, “1”.

  Subsequently, the control unit 2b starts up the unnecessary data deletion unit 232, and under the control of the unnecessary data deletion unit 232, first, in step S57, the value of the strength information is set in the node after the uniform subtraction processing of the strength information. It is determined whether or not there is one that becomes “0”. As a result of this determination, when a node having a strength information value of “0” is found, the data of the node is deleted from the preference model information in step S58. In this case as well, as described above, the unnecessary data deletion process is performed when the value of the intensity information is smaller than a preset threshold value other than when the value is “0”. Also good.

  As described above in detail, in the second embodiment, the uniform subtraction process of the intensity information and the unnecessary data deletion process are performed when the intensity information of the preference model information is updated. Therefore, every time the update process of the intensity information is performed, the intensity information is uniformly subtracted and the unnecessary data is deleted, so that the intensity information can always be kept up-to-date. It is possible to recommend information based on the latest preference model information.

(Third embodiment)
The third embodiment of the present invention is based on the premise that the preference model has a plurality of conceptual structures created for each viewpoint, and is recommended in response to an information provision request from an information user. When creating and returning information, a concept having the maximum strength is first extracted from a plurality of concept structures having different viewpoints. Then, the intensity is corrected so that the intensity becomes a preset ratio between the extracted concepts. At the same time, the strength of the concept after the correction is adjusted so that the ratio of the strength between the extracted concept and other concepts included in the concept structure for each conceptual structure maintains the ratio before the correction. Accordingly, the strengths of the other concepts are corrected. Then, the instance importance of each store is recalculated based on the corrected concept strength, and the recommended information is generated by selecting and rearranging the stores based on the instance importance. Is sent back to the information user.

  FIG. 11 is a block diagram showing functional configurations of the information recommendation processing device SVc and the information storage device SD according to the third embodiment of the present invention. In this figure, the same parts as those in FIG. 2 or FIG.

  The information storage device SD is provided with a conceptual structure database 31, an intensity database 32, and an information database 33. In the conceptual structure database 31, information representing a conceptual structure created in advance for each information user is stored in advance. This conceptual structure is created for each viewpoint.

FIG. 15 shows an example of this. The concept of “restaurant” is the highest level concept, and two conceptual structures are associated with the node of the highest level concept. One of these conceptual structures has “genre” as a higher-level concept node, which is associated with a plurality of middle-level concept nodes such as “Japanese food”, “noodles”, and “meat food” by links, A plurality of subordinate concept nodes “sushi”, “soba”, “spaghetti meat sauce”, and “steak” are associated with the middle concept node by links, and store names 1 to 5 are associated as instances with these subordinate concept nodes. . On the other hand, the other of the above two conceptual structures has “no smoking / smoking” as an upper concept node, and two lower concept nodes “smoking” and “no smoking” are associated with each other by a link, and these lower concept nodes are connected. The store names are selectively associated with each other.
Note that FIG. 15 shows a state in which strength is given to each concept (node) and content link, but this strength information is stored in the strength database 32 described later.

  In the strength database 32, identification information (concept ID) of a plurality of concepts (nodes) included in a common conceptual structure stored in the conceptual structure database 31 in association with identification information (user ID) of each information user. ) And information indicating the intensity of preference of the information user currently set for the concept. An example is shown in FIG.

  In the information database 33, the ID of the concept associated with the store name and the link in association with the information (information ID) for identifying the store name as the instance, and the content link strength set for the link are displayed. Representing information is stored respectively. An example is shown in FIG.

  The control unit 24 includes an information recommendation processing program group 24 in addition to the strength update control program group 21 and the strength reduction / deletion control program group 22 or 23.

  The information recommendation processing program group 24 includes an information request reception unit 241, an intensity classification unit 242, an intensity correction unit 243, an instance importance calculation unit 244, an information selection / sorting unit 245, an information transmission unit 246, A conceptual structure information reading unit 247 is provided.

  The conceptual structure information reading unit 247 executes processing for reading information representing a common conceptual structure from the conceptual structure database 31 in the information storage device SD and storing it in a storage unit in the information recommendation processing device SVc.

  The information request receiving unit 241 receives a recommendation information provision request transmitted from the user terminals MS1 to MSn via the communication interface unit 1, and relates to the requesting information user based on the received information provision request. A process of creating information (user information) and passing it to the intensity classification unit 242 is executed.

  The strength classifying unit 242 reads information representing the strength of each concept related to the information user from the strength database 32 based on the user information passed from the information request receiving unit 241. Further, based on the information on the concept structure common to the information users read and stored in advance by the concept structure information reading unit 247, each concept related to the information user and the strength of the concept are selected from a plurality of viewpoints. A process of classifying which one of them belongs is performed.

The intensity correction unit 243 has the following processing functions.
(1) A process of extracting the concept having the maximum intensity from the concept structure corresponding to the plurality of viewpoints according to the classification result by the intensity classification unit 242.
(2) Processing for correcting the intensity so that the intensity becomes a preset ratio between the extracted concepts. The ratio is input in advance by an information user or a system operator and stored in a memory in the control unit 2c.
(3) For each conceptual structure corresponding to the above-mentioned multiple viewpoints, the intensity ratio between the extracted concept of maximum intensity and the other concepts included in the conceptual structure should maintain the ratio before correction. And a process of correcting the intensity of the other concepts according to the intensity of the concept after the correction.

  The instance importance calculation unit 244 reads the content link strength information associated with the store name included in the information user's preference model from the information database 33, and the read content link strength information and the strength A process of calculating the instance importance of each store using the information on the strength of the lower concept corrected by the correcting unit 243 is executed.

  The information selection / rearrangement unit 245 selects a store to be recommended to the information user based on the instance importance of each store calculated by the instance importance calculation unit 244, and selects the selected store. Sort the items in descending order of instance importance.

  The information transmitting unit 246 creates recommended information of the stores selected and rearranged by the information selecting / sorting unit 245, and sends the created recommended information to the requesting user terminals MS1 to MSn from the communication interface unit 1. Execute the process to send to.

Next, information recommendation processing including intensity information correction processing by the information recommendation device SVc configured as described above will be described. FIG. 12 is a flowchart showing the processing procedure and processing contents.
(1) Reading conceptual structure information The control unit 2c first activates the conceptual structure information reading unit 247 in step S51, and stores the conceptual structure information in the conceptual structure database 31 of the information storage device SD under the control of the conceptual structure information reading unit 247. The conceptual structure information is read from the conceptual structure database 31 and stored in the storage unit in the information recommendation device SVc. Here, for example, as shown in FIG. 15, it is assumed that information representing a conceptual structure corresponding to two viewpoints is read from the conceptual structure database 31 and stored in a storage unit in the apparatus.

(2) Reception of information provision request The control unit 2c monitors reception of an information provision request under the control of the information request reception unit 241 in step S52. In this state, it is assumed that an information provision request is transmitted from the user terminal MS1 and this information provision request is received by the communication interface unit 1. Then, the information request receiving unit 241 generates user information for identifying the transmission source information user in step S53 based on the information regarding the transmission source information user included in the received information provision request. .

(3) Classification of Intensity Information Subsequently, the control unit 2c activates the intensity classification unit 242 in step S54, and stores the user information generated by the information request reception unit 241 under the control of the intensity classification unit 242. In addition, the intensity information of each concept related to the information user is read from the intensity database 32 in the information storage device SD. Then, based on the concept structure information read and stored in advance from the concept structure database 31, it is determined which of the plurality of viewpoints each concept related to the information user and the concept strength belongs to. Then, classification is performed in step S55.

  For example, if the information user's preference model information is as shown in FIG. 15, the subordinate concept nodes “smoking cessation” and “smoking” and the intensity information are based on the viewpoint that “smoking / no smoking” is a superordinate concept. On the other hand, nodes of medium concept such as “Japanese food”, “noodles”, “meat” and subordinate nodes such as “sushi”, “soba”, “spaghetti meat sauce”, “steak” and their strength information Is categorized into the viewpoint of “genre” as a superordinate concept.

(4) Correction of Intensity Information Next, the control unit 2 c activates the intensity correction unit 243 and executes the following processing under the control of the intensity correction unit 243.
That is, first, in step S56, in accordance with the classification result by the intensity classification unit 242, the concept having the maximum intensity value is extracted from the conceptual structure corresponding to the two viewpoints. For example, in the preference model information shown in FIG. 15, “smoking cessation” in which the intensity is set to “20” from the conceptual structure of the viewpoint of “smoking / non-smoking” as a superordinate concept, and “genre” as a superordinate concept. The “meat udon” with the strength set to “5” is extracted from the conceptual structure of the viewpoint.

  Next, in step S57, a process of correcting the intensity value is performed so that the intensity becomes an arbitrary ratio set in advance between the extracted concepts. For example, if the ratio is set to “smoking / non-smoking”: “genre” = 0.8: 1, the extracted “non-smoking” and “meat udon” are extracted in the preference model shown in FIG. The strengths “20” and “5” are corrected to “no smoking” = 0.8 and “meat udon” = 0.6, respectively, as shown in FIG.

  Subsequently, in step S58, for each conceptual structure corresponding to each of the above viewpoints, the intensity ratio between the concept of maximum intensity extracted from the conceptual structure and the other concepts included in the conceptual structuring is corrected. Based on the corrected intensity value of the extracted maximum intensity concept, the intensity of the other concepts is corrected so as to maintain the previous ratio.

  For example, as described above, if the intensity of “no smoking” and “meat udon” is corrected to “no smoking” = 0.8 and “meat udon” = 0.6, respectively, the concept of “smoking / no smoking” viewpoint As shown in FIG. 16, the intensity “5” of “smoking”, which is another concept included in the structure, maintains a ratio with the intensity “20” of “non-smoking” before correction, as shown in FIG. It is corrected to 0.16. Similarly, “sushi”, “soba”, “spaghetti meat sauce”, “steak” strengths “4”, “5”, “1”, “2” included in the conceptual structure of the “genre” viewpoint are: In order to maintain the ratio of the strength “5” of “meat udon” before correction, as shown in FIG. 16, “sushi” = 0.8, “soba” = 0.6, “spaghetti meat sauce = 0.2”, respectively. "," Steak "is corrected to 0.4.

  When the correction process for the intensity of each concept is completed, the intensity correction unit 243 updates the stored information in the intensity database 32 to the corrected intensity in step S59.

(5) Calculation of Instance Importance Next, the control unit 2c activates the instance importance calculator 244 in step S60, and under the control of the instance importance calculator 244, the information database 33 in the information storage device SD. The content link strength associated with the store name included in the information user's preference model information is read out. Then, the instance importance of each store is calculated by using the read content link strength information and the subordinate concept strength information corrected by the strength correction unit 243.

  For example, in the example of FIG. 16, the instance importance of “Store 1” is 0.8 × 0.5 + 0.8 × 0.5 = 0.8, and the instance importance of “Store 2” is 0.16 × 0. 5 + 0.6 × 0.4 = 0.32, and the instance importance of “Store 3” is 0.8 × 0.5 + 1 × 0.6 = 1.0. Similarly, the instance importance of “Store 4” is 0.8 × 0.5 + 0.2 × 0.4 = 0.48, and the instance importance of “Store 5” is 0.16 × 0.5 + 0.4 × 0. .5 = 0.28.

(6) Selection and rearrangement of instances Subsequently, the control unit 2c activates the information selection / sorting unit 245 in step S61, and under the control of the information selection / sorting unit 245, the instance importance calculation unit Based on the instance importance of each store calculated by 244, a store to be recommended to the information user is selected, and the selected stores are rearranged in descending order of instance importance.
For example, in the example described above, stores having an instance importance of 0.4 or more, that is, “Store 1”, “Store 3”, and “Store 4” are selected, and these stores are “Store 3” in descending order of instance importance. ”,“ Store 1 ”, and“ Store 4 ”.

(7) Returning recommendation information Finally, the control unit 2c activates the information transmission unit 246 in step S62, and is selected and rearranged by the information selection / rearrangement unit 245 under the control of the information transmission unit 246. The recommended information of the store is created, and the created recommended information is transmitted from the communication interface unit 1 to the requesting user terminals MS1 to MSn.

  The recommendation information transmitted from the information recommendation processing device SVc is transmitted and received to the requesting user terminal MS1 via the communication network NW. And it displays on the indicator of user terminal MS1. Therefore, the information user can know the recommended store based on the displayed recommendation information.

  Further, it is assumed that the information user selects a desired store on the user terminal MS1 in a state where the recommended information is displayed. Then, the selection information is sent from the user terminal MS1 to the information recommendation processing device SVc via the communication network NW.

  When the communication interface unit 1 receives the selection information, the information recommendation processing device SVc moves from step S63 to steps S64 and S65, and thereafter the strength update control program group 21 and the strength reduction / deletion control program group 22 or 23. Under the control, strength update control and strength reduction / deletion control are executed. The processing procedure and processing contents of the strength update control and strength reduction / deletion control are the same as those described in the first and second embodiments.

  As described above in detail, in the third embodiment, when the recommendation information is created and returned in response to the information provision request from the user terminals MS1 to MSn, a plurality of concepts corresponding to the information user of the request source The intensity information is classified for each viewpoint, and the concept having the maximum intensity information is extracted from each classified viewpoint. Then, the intensity information is corrected so that the intensity information is arbitrarily set in advance between the extracted concepts, and the extracted concept included in the idea for each idea and the others The intensity information of the other concepts is corrected in accordance with the intensity information of the concept after the correction so that the ratio of the intensity information with the concept maintains the ratio before the correction. Then, the instance importance of each store is recalculated based on the corrected concept strength information, and the recommendation information is generated by selecting and rearranging the stores based on the instance importance. Information is sent back to user terminals MS1 to MSn of information users.

  Therefore, when the preference model has a plurality of conceptual structures created under different viewpoints, even if the appearance probability of the concept is biased depending on the number of options between the viewpoints, the strength information between the viewpoints This makes it possible to more appropriately recommend stores to information users. Furthermore, since the correction ratio is set to an arbitrary value when correcting the intensity information between the concepts, there is an advantage that the intensity information of the concept in each viewpoint can be corrected according to the importance degree between the viewpoints. is there.

(Other embodiments)
In each of the above-described embodiments, the case where information representing the strength of information user's preference is set for each concept has been described as an example. However, for example, when strength information is set for a link that associates concepts as shown in FIG. In addition, the present invention can be implemented. In this case, the increase width of the strength information set for the link that directly associates the concept of interest with the related concept is set to “W”, and the strength information of the link that links the related concept is set to an increase width smaller than the above “W”. do it. In short, as the distance from the concept of interest increases, that is, as the relation rank decreases, the increase amount of the strength information set for the link is set to be small.

In the third embodiment, the intensity information extracted from each viewpoint is corrected so that the intensity information between the maximum concepts has an arbitrarily set ratio. However, these intensity values are the same value. You may correct | amend so that it may become.
For example, taking the preference model shown in FIG. 15 as an example, the concept structure of “smoking cessation” extracted from the conceptual structure of “smoking / non-smoking” as a superordinate concept and the concept of “genre” as a superordinate concept. The intensity of the “meat udon” extracted from is corrected to “1”. At the same time, the intensity “5” of “smoking”, which is a concept other than the above “non-smoking” included in the conceptual structure of the “smoking / non-smoking” viewpoint, is the ratio of the intensity “20” of “no smoking” before correction. As shown in FIG. 17, “smoking” is corrected to 0.2 so as to maintain the above. Similarly, “sushi”, “soba”, “spaghetti meat sauce”, “steak” strengths “4”, “5”, “1”, “2” included in the conceptual structure of the “genre” viewpoint are: In order to maintain the ratio of the strength “5” of “meat udon” before correction, as shown in FIG. 17, “sushi” = 0.8, “soba” = 0.6, “spaghetti meat sauce = 0.2”, respectively. "," Steak "is corrected to 0.4.
In this way, it is possible to easily correct the bias of concept strength information between viewpoints regardless of the appearance frequency of concepts between viewpoints.

  In each of the above embodiments, the information recommendation processing devices SVa, SVb, SVc and the information storage device SD are configured as separate servers, and the two devices are connected to each other via the communication network NW. The database 3 in the information storage device SD may be accommodated in the information recommendation processing devices SVa, SVb, SVc.

  In each of the above embodiments, the restaurant preference model has been described as an example. However, the present invention can be similarly applied to a taste model such as news, advertisement, and fashion. In addition, the configuration of the information recommendation processing device, the strength information update process, the processing procedure and contents of the uniform reduction process and the unnecessary data reduction process, the increase range at the time of update, the decrease range at the uniform decrease process, etc. Various modifications can be made without departing from the scope of the invention.

  In short, the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the components without departing from the scope of the invention in the implementation stage. Moreover, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in each embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  MS1 to MSn: Client terminal, SVa, SVb, SVc ... Information recommendation processing device, SD ... Information storage device, NW ... Communication network, 1 ... Communication interface unit, 2a, 2b, 2c ... Control unit, 3 ... Database, 21 ... Strength update control program group, 22, 23 ... Strength reduction / deletion control program group, 24 ... Information recommendation processing program group, 31 ... Conceptual structure database, 32 ... Strength database, 33 ... Information database, 211 ... Information receiving unit, 212 ... Target concept identification unit, 213 ... related concept identification unit, 214 ... strength information update unit, 221 ... periodic activation unit, 222 ... reduction process execution determination unit, 223, 231 ... reduction process execution unit, 224, 232 ... unnecessary data deletion unit 241 ... Information request receiving unit 242 ... Strength classification unit 243 ... Strength correction unit 244 ... Instagram Scan importance calculation section, 245 ... information selection and rearrangement unit, 246 ... information transmitting unit, 247 ... conceptual structure information reading unit.

Claims (8)

Using a conceptual structure consisting of a directed acyclic graph in which a plurality of hierarchized concepts are linked by a link and recommendation candidate information is linked to the concept, the user's preference for each concept or a link leading to the concept Storage means for creating and storing preference model information reflecting the user's preference by setting corresponding strength information;
Updating means for updating intensity information included in the stored preference model information in response to a selection operation of recommendation candidate information by the user;
Based on the preference model information in which the intensity information has been updated, it comprises a presentation means for selecting and presenting recommended candidate information suitable for the user's preference,
The updating means includes
Means for identifying a concept directly associated with the selection candidate information selected as a target concept from the stored preference model information;
Means for identifying a concept directly or indirectly associated with the identified concept of interest from the stored preference model information as a related concept;
An information recommendation processing apparatus comprising: means for increasing intensity information set in the identified concept of interest and related concept or a link for associating these concepts. A correction unit that corrects the intensity information of the stored preference model information;
The correction means includes
When the preference model information has a plurality of conceptual structures created for each of a plurality of viewpoints, means for extracting a concept or link having the maximum intensity information from each of the plurality of conceptual structures;
Inter-structure correction means for correcting the strength information of the concept or link extracted from each of the plurality of concept structures so that the value thereof is a ratio arbitrarily set in advance between the concept or link;
For each conceptual structure, the intensity after correction so that the ratio of intensity information between the extracted concept or link and other concepts or links included in the conceptual structure maintains the ratio before correction. The information recommendation processing apparatus according to claim 1, further comprising an intra-structure correction unit that corrects the strength information of the other concept or the link in accordance with information.   The inter-structure correction means corrects the strength information of a concept or link extracted from each of the plurality of conceptual structures so that the value thereof is the same between the concepts or links. 2. The information recommendation processing device according to 2.   The means for increasing the strength information increases the strength information set in the target concept or a link connected thereto by a first increase width, and the strength information set in the related concept or a link connected thereto. The information recommendation processing apparatus according to claim 1, wherein the information recommendation processing apparatus increases the second increase width smaller than the first increase width. The updating means includes
Means for determining whether the strength information set for each concept or link of the stored preference model information has been updated within a preset period;
When it is determined that at least one of the strength information has been updated within the set period, the strength information set for each concept or link of the stored preference model information is uniformly reduced by a preset reduction width. The information recommendation processing device according to claim 1, further comprising: means. The updating means includes
Each time the strength information is updated by the means for increasing the strength information, the strength information set for each concept of the stored preference model information or the link leading to the concept is reduced in a preset reduction range. The information recommendation processing device according to claim 1, further comprising means for uniformly reducing the information recommendation processing device. The updating means includes
Means for determining whether the intensity information after being subjected to the reduction process by the means for uniformly reducing becomes equal to or less than a preset minimum value;
And means for deleting, from the stored preference model information, a concept in which the intensity information is set when intensity information that is less than or equal to the minimum value is detected by the determination process. The information recommendation processing device according to claim 5 or 6.   A program for causing a computer included in the information recommendation processing device to execute processing for realizing each unit included in the information recommendation processing device according to claim 1.

Images