JP2013039914A - Method and device for supporting information provision - Google Patents

Abstract

PROBLEM TO BE SOLVED: To receive information provision corresponding to the preference of a user at a start of use without performing troublesome prearrangement operation such as the setting of preference information.SOLUTION: At a start of use, based on profile information on a driver stored in a driver profile information storing part 101 and average preference stored in a general tendency database 102, average preference corresponding to the age and sex of the driver is introduced to a Bayesian network processing part 104 to perform learning operation. Music obtained from music contents stored in a content data storing part 106 is learned as the preferable music of a user. As a result, a program list reflecting the user's preference is prepared by inference calculation by a Bayesian network, and program contents in accordance with the program list are provided. Subsequently, the operation performed by the user during listening to the programs is learned, thus enabling program provision further reflecting user's preference.

Description

  The present invention relates to an information provision support method and an information provision support apparatus that support information provision to a user in consideration of user preferences. The present invention particularly relates to an information provision support method, an information provision support device, and a user preference for learning a user preference, creating a program table reflecting the user preference, and providing a program that matches the user preference. The present invention relates to an information providing support method and an information providing support apparatus for presenting candidates according to user thinking so that learning and selection of restaurants and other drop-in facilities reflecting user preferences can be performed.

  For example, when a driver who is driving a vehicle listens to information such as music or news, the provided information must be automatically adapted to the driver's preference with as few operations as possible for safety reasons. There is. In particular, since the music preference varies greatly among individuals, it is necessary to reflect the user's preference in more detail when providing music.

  On the other hand, Patent Document 1 below discloses a technique for reducing the number of operations when a driver listens to music or the like while driving.

Japanese Patent Laid-Open No. 2001-343979 (FIGS. 1, 3, 5, and 7)

  However, according to the technique disclosed in Patent Document 1, there is a problem that the user must first perform some initial setting operations related to preferences. Further, when the user's preference changes from the user's preference input by the initial setting operation, there is a problem that the user has to perform the initial setting operation for inputting the user's preference again. In particular, the preference for music content tends to change quickly, and the technique disclosed in Patent Document 1 has a problem that it cannot cope with such a change in preference for music content.

  In order to solve the above-described problems, the present invention uses information provision (program configuration and selection of program content) that matches the user's preference without the user performing troublesome preparatory operations such as setting preference information. It is intended to be received from the beginning.

In order to achieve the above object, the information providing support method of the present invention selectively reads information stored in advance in a mobile unit mounting device mounted on a mobile unit, and
In the stochastic inference model processing unit for performing probabilistic inference processing, performing learning processing related to the inference processing based on the information read selectively;
Detecting a situation where a user is on the moving body;
Performing an inference process in the probabilistic inference model processing unit on the detected situation, inferring an attribute related to the situation from a plurality of predetermined attributes;
Adding a content score representing the user's preference for each of the information that can be provided to the user, and storing the content score for each of the information that can be provided to the user;
A target to be provided to the user while selecting the information to be provided to the user from the attributes inferred to be related to the detected situation and the information that can be provided to the user based on the content score Determining how to provide said information selected as:
Have.
By this method, the user can receive information (program configuration and selection of program content) that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information. . It is also possible to set a content score (user preference) for each piece of content information and to provide information based on the content score.

Further, the information providing support method of the present invention selectively reads information stored in advance in a mobile body mounting device mounted on a mobile body,
In the stochastic inference model processing unit for performing probabilistic inference processing, performing learning processing related to the inference processing based on the information read selectively;
Detecting a situation where a user is on the moving body;
Performing an inference process in the probabilistic inference model processing unit on the detected situation, inferring an attribute related to the situation from a plurality of predetermined attributes;
Adding a content score representing the user's preference for each of the information that can be provided to the user, and storing the content score for each of the information that can be provided to the user;
An attribute score indicating the user's preference degree defined by the probabilistic inference model processing unit with respect to an attribute of each of the information that can be provided to the user, and the contents of the information that can be provided to the user Determining an overall score for each of the information that can be provided to the user based on the score; and
A target to be provided to the user while selecting the information to be provided to the user from among the attributes inferred to be related to the detected situation and the information that can be provided to the user based on the total score Determining how to provide said information selected as:
Have.
By this method, the user can receive information (program configuration and selection of program content) that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information. . Also, a content score (user preference) is set for each piece of content information, and information is provided based on a total score obtained by combining the content score and an attribute score representing the preference for each attribute. Is possible.

In addition to the above method, the information provision support method of the present invention generates a random number and selects information to be provided to the user from information that can be provided to the user in consideration of the random number. Has steps.
By this method, it is possible to set a content score (user preference) for each piece of content information and to provide information based on a value obtained by adding randomness to the content score.

In addition to the above method, the information provision support method of the present invention generates a random number and selects information to be provided to the user from information that can be provided to the user in consideration of the random number. Has steps.
By this method, content score (user's preference) is set for each content information, and information based on a value in which randomness is added to a total score obtained by combining this content score and an attribute score representing preference for each attribute The provision can be made.

In addition to the above method, the information providing support method of the present invention provides the user with the information based on the information selected as a target to be provided to the user and the method of providing the information. Steps to do,
Recording operation performed by the user as operation history information during the provision of the information;
In the stochastic inference model processing unit, performing a learning process based on the operation history information,
Have.
By this method, an operation that is naturally performed when the user is receiving information is recorded, and the preference obtained from the operation history can be reflected in further information provision.

In addition to the above method, the information providing support method of the present invention provides the user with the information based on the information selected as a target to be provided to the user and the method of providing the information. Steps to do,
Recording operation performed by the user as operation history information during the provision of the information;
Changing the content score based on the operation history information;
Have.
By this method, an operation that is naturally performed when the user is receiving information is recorded, and the preference obtained from the operation history can be reflected in further information provision.

In the information provision support method of the present invention, in addition to the above method, the operation performed by the user recorded as the operation history information includes at least one of a skip operation, a repeat operation, and a direct selection operation.
With this method, the user's palatability can be obtained from natural operations performed when the user is receiving information.

In addition to the above method, the information provision support method of the present invention provides a method in which the predetermined information is provided when the skip operation or the repeat operation is performed while the predetermined information is being provided. A step of recording the time until,
In the stochastic inference model processing unit, performing a learning process based on the time together with the operation history information,
Have.
By this method, when a skip operation or a repeat operation is performed, it is possible to grasp the intention or operation of the operation in more detail.

In addition to the above method, the information provision support method of the present invention is an attribute inferred to be related to the situation in the selection of information to be provided to the user and the determination of the method of providing the information. Create a program with no corners.
With this method, the user can enjoy a program including a corner having an attribute that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information.

In addition to the above-described method, the information providing support method of the present invention creates a program including a plurality of attribute corners that are inferred to be related to the situation, depending on the degree of preference for each attribute. To proportionally distribute the length of each corner.
With this method, the length of each corner of the program can be determined according to the user's preference.

In addition to the above method, the information providing support method of the present invention causes a combination of attributes to be inferred by combining a plurality of attributes, and the attribute inferred to be related to the situation. Create a program with a combination of corners.
This method makes it possible to determine a corner constituting a program from a combination of a plurality of attributes.

In order to achieve the above object, the information providing support device of the present invention includes means for selectively reading information stored in advance in the mobile unit mounting device mounted on the mobile unit,
Probabilistic inference model processing means for performing probabilistic inference processing and performing learning processing related to the inference processing based on the selectively read information;
Situation detecting means for detecting a situation in which a user is on the moving body;
Content score giving means for attaching a content score representing the user's preference for each of the information that can be provided to the user;
Content score storage means for storing the content score for each of the information that can be provided to the user;
An inference process is performed on the detected situation by the probabilistic inference model processing means, an attribute related to the situation is inferred from a plurality of predetermined attributes, and it is inferred that the situation is related And selecting the information to be provided to the user from among the information that can be provided to the user based on the attribute and the content score, and providing the information selected as a target to be provided to the user Information providing support means to be determined
Have.
With this configuration, the user can receive information (program configuration and selection of program content) that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information. . It is also possible to set a content score (user preference) for each piece of content information and to provide information based on the content score.

Further, the information providing support device of the present invention is a means for selectively reading out information stored in advance in the mobile body mounting device mounted on the mobile body,
Probabilistic inference model processing means for performing probabilistic inference processing and performing learning processing related to the inference processing based on the selectively read information;
Situation detecting means for detecting a situation in which a user is on the moving body;
Content score giving means for attaching a content score representing the user's preference for each of the information that can be provided to the user;
Content score storage means for storing the content score for each of the information that can be provided to the user;
An attribute score indicating the user's preference degree defined by the probabilistic inference model processing means with respect to the attributes of each of the information that can be provided to the user, and the contents of the information that can be provided to the user A total score determining means for determining a total score of each piece of information that can be provided to the user based on the score;
An inference process is performed on the detected situation by the probabilistic inference model processing means, an attribute related to the situation is inferred from a plurality of predetermined attributes, and it is inferred that the situation is related And selecting the information to be provided to the user from among the information that can be provided to the user based on the attribute and the total score, and providing the information selected as a target to be provided to the user Information providing support means to be determined
Have.
With this configuration, the user can receive information (program configuration and selection of program content) that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information. . Also, a content score (user preference) is set for each piece of content information, and information is provided based on a total score obtained by combining the content score and an attribute score representing the preference for each attribute. Is possible.

In addition to the above configuration, the information providing support device of the present invention further includes means for generating a random number, and includes information provided to the user from information that can be provided to the user in consideration of the random number. A selection is made.
With this configuration, it is possible to set a content score (user preference) for each piece of content information, and to provide information based on a value obtained by adding randomness to the content score.

In addition to the above configuration, the information providing support device of the present invention further includes means for generating a random number, and includes information provided to the user from information that can be provided to the user in consideration of the random number. A selection is made.
With this configuration, a content score (user preference) is set for each content information, and information based on a value obtained by adding randomness to a total score obtained by combining this content score and an attribute score representing the preference for each attribute The provision can be made.

In addition to the above configuration, the information providing support device of the present invention provides the user with the information based on the information selected as a target to be provided to the user and a method of providing the information. Information providing means,
An operation history information recording unit that records an operation performed by the user as operation history information during the provision of the information;
In the stochastic inference model processing unit, model update control means for performing learning processing based on the operation history information,
Have.
With this configuration, an operation that is naturally performed when the user is receiving information is recorded, and the preference obtained from the operation history can be reflected in further information provision.

In addition to the above configuration, the information providing support device of the present invention provides the user with the information based on the information selected as a target to be provided to the user and a method of providing the information. Information providing means,
An operation history information recording unit that records an operation performed by the user as operation history information during the provision of the information;
Content score changing means for changing the content score based on the operation history information;
Have.
With this configuration, an operation that is naturally performed when the user is receiving information is recorded, and the preference obtained from the operation history can be reflected in further information provision.

In addition to the above configuration, the information providing support device of the present invention includes at least one of a skip operation, a repeat operation, and a direct selection operation in the operation performed by the user recorded as the operation history information.
With this configuration, user preference can be obtained from natural operations performed when the user is receiving information.

In addition to the above configuration, the information providing support device according to the present invention provides the predetermined information when the skip operation or the repeat operation is performed while the predetermined information is being provided. An elapsed time information recording unit that records the time until the probabilistic reasoning model processing unit performs learning processing based on the time together with the operation history information.
With this configuration, when a skip operation or a repeat operation is performed, it is possible to grasp the intention or operation of the operation in more detail.

In addition to the above configuration, the information providing support device of the present invention is configured such that the information providing support means creates a program having a corner having an attribute inferred to be related to the situation. .
With this configuration, the user can enjoy a program including a corner having an attribute that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information.

In addition to the above configuration, the information providing support apparatus according to the present invention relates to each attribute when the information providing support means creates a program including a plurality of attribute corners that are inferred to be related to the situation. The length of each corner is proportionally distributed according to the degree of preference.
With this configuration, the length of each corner of the program can be determined according to the user's preference.

In addition to the above configuration, the information providing support device of the present invention causes the information providing support means to infer a combination of attributes related to the situation by combining a plurality of attributes, and is related to the situation. It is configured to create a program having a corner with a combination of attributes inferred to be present.
With this configuration, it is possible to determine a corner that configures a program from a combination of a plurality of attributes.

  The present invention has the above configuration, and the user uses information provision (program configuration and selection of program content) that matches the user's preference without performing troublesome preparatory operations such as setting preference information. There is an effect that it can be received from the beginning. This also provides an effect that the driver can enjoy a program configuration and program content that matches the driver's preference without performing troublesome preparatory operations such as setting preference information. In addition, by reflecting the user's operation during the information provision process in the user preference that is referred to when selecting the information to be provided, the information that dynamically responds to the user's preference change and faithfully reflects the user's preference Offering will be realized.

An example of the configuration of a program guide creation device including a content preference learning device according to the first embodiment of the present invention is illustrated. It is a flowchart which shows an example of the process at the time of the new use start performed with the program schedule preparation apparatus containing the content preference learning apparatus in the 1st Embodiment of this invention. It is a flowchart which shows an example of the process after the 2nd performed by the program schedule preparation apparatus containing the content preference learning apparatus in the 1st Embodiment of this invention. It is a figure which shows an example of the Bayesian network used in the 1st Embodiment of this invention. It is a figure which shows an example of the program schedule before the operation of a driver is reflected in the 1st Embodiment of this invention. It is a figure which shows an example of the program table after the operation of a driver was reflected in the 1st Embodiment of this invention. An example of the configuration of a program guide creation device including a content preference learning device according to the second embodiment of the present invention is illustrated. It is a flowchart which shows an example of the process at the time of the new use start performed with the program schedule preparation apparatus containing the content preference learning apparatus in the 2nd Embodiment of this invention. It is a flowchart which shows an example of the process after the 2nd performed by the program schedule preparation apparatus containing the content preference learning apparatus in the 2nd Embodiment of this invention. In the second embodiment of the present invention, the song score and attribute score at the start of use (first program configuration) are changed to the song score and attribute score after the start of use (second program configuration). FIG.

  Hereinafter, first and second embodiments of the present invention will be described with reference to the drawings.

  The present invention provides user decision support (proposing information to users and providing information) in consideration of the user's preferences. In the following, in particular, the preference of a driver who is riding (driving) a vehicle. A case will be described in which the present invention is applied in providing a program (configured by audio content so that it can be listened to while driving) that matches the preference of the driver reflecting the above.

  The program is composed of one or more corners (for example, has a length of about one hour). The corner is a time frame of one segment having a specific attribute (for example, having a length of several minutes to several tens of minutes), and is composed of one or more music contents and / or information contents. The music content is one music (for example, has a length of about several minutes), and the information content is news, weather forecast, etc. (for example, has a length of about several minutes).

<First Embodiment>
First, a first embodiment of the present invention will be described. FIG. 1 shows an example of the configuration of a program guide creation device including a content preference learning device according to the first embodiment of the present invention.

  A program guide creation device 100 shown in FIG. 1 is a content preference learning device for creating a program guide (time table) of a program to be reproduced in a vehicle according to the preference of a user (driver who drives the vehicle). It has the composition containing. Here, it is assumed that the program guide creation device 100 illustrated in FIG. 1 is incorporated in an in-vehicle device of a vehicle.

  1 includes a driver profile information storage unit 101, a general trend database 102, a driver trend data output unit 103, a Bayesian network processing unit 104, a boarding time information detection unit 105, and content data storage. Unit 106, preference evaluation unit 107, program guide creation unit 108, operation history information storage unit 109, and operation input unit 110. In FIG. 1, each function is represented by a function block, but each function can be realized by hardware and / or software (a program executable by a computer).

  The driver profile information storage unit 101 is an information memory for storing profile information such as the age, sex, and type of business of the driver, for example. This driver profile needs to be registered in advance. The driver profile information storage unit 101 may always store driver profile information, and if necessary, a driver acquired from an external device (a communication device other than the in-vehicle device) through a wireless communication function or the like. The profile information may be temporarily stored.

  Moreover, the general tendency database 102 is an information memory in which information representing a general (average) tendency of preference degree without bias corresponding to each profile is stored. More specifically, the general trend database 102 stores information representing the preferences of general people belonging to each age, each gender, each business type, and the like. Similar to the driver profile information storage unit 101, the general trend database 102 is also located remotely from the in-vehicle device, and information indicating general people's preference is acquired through a wireless communication function or the like as necessary. May be.

  In addition, the driver tendency data output unit 103 is information that represents the preference of general people (for example, people of the same age, gender, and type of business as the driver) corresponding to the profile to which the driver applies based on the profile information of the driver. It has the function to acquire.

  The Bayesian network processing unit 104 has a function of performing inference calculation based on the Bayesian network model. In this specification, a Bayesian network model is used, but various models that perform arbitrary probabilistic inference processing can be employed.

  For example, the Bayesian network in the first embodiment of the present invention has a structure as shown in FIG. In the Bayesian network shown in FIG. 3, the events corresponding to the driver's profile such as the driver's age, gender, occupation (business type), the presence / absence of passengers, events that can be observed when riding such as day of the week, time, etc. Events corresponding to the current scene specified from the event (for example, commuting to work, commuting home, shopping, homecoming, etc.), attributes of each corner and each content considered when composing a program ( Genres such as genre, age, and artist in the case of music content, genres such as news, weather, and traffic information in the case of information content, and more detailed genres such as economic news and sports news) Each node and the dependency of each of these nodes are defined, and the dependency of each of these nodes is the conditional probability It is represented by (conditional probability tables).

  For example, Bayesian network learning is performed by information representing general people's preference corresponding to the driver supplied from the driver tendency data output unit 103, content stored in the content data storage unit 106 (the preference of the driver). Attribute information of music content and information content received via wireless communication), operation history information stored in the operation history information storage unit 109 (operation history information indicating an operation performed by the driver when listening to a program), and the like Based on. As will be described later, different operations are performed in the new learning of the Bayesian network in the case of newly configuring the program guide and the second and subsequent learning of the Bayesian network in the case of creating the existing program guide. However, details will be described later.

  The boarding information detection unit 105 has a function of detecting information such as the presence / absence of a passenger, day of the week, and time when a program is provided to the driver.

  In addition, the content data storage unit 106 temporarily stores an information memory storing music content data processed by a predetermined compression format or the like, or information content acquired via a wireless communication function (not shown). Information memory to be stored, such as HDD (Hard Disk Drive) or RAM (Random Access Memory) in which music content data ripped from a CD (Compact Disc) is stored It can be realized by such as. That is, the content data in the present invention is not limited to music content and information content data statically stored in advance (locally in the program guide creation device 100), but also from a server connected to the network via wireless communication. It may be music content data or information content data that has been downloaded.

  In the case of music content, information actually required by the Bayesian network processing unit 104 is music content attribute information stored in the content data storage unit 106. Here, it is assumed that the attribute information of each piece of music content can be obtained by referring to the content data storage unit 106, but if necessary, processing for acquiring the attribute information (for example, a remotely located CDDB (CD Inquiry to Database) may be performed.

  Further, the preference level evaluation unit 107 refers to and compares the posterior probability values of the nodes obtained as a result of the inference calculation in the Bayesian network processing unit 104 (comparison between the posterior probability values or the posterior probability value and a predetermined threshold value). It is possible to infer the degree of preference of each attribute. Specifically, the posterior probability value of the attribute node of the corner (or content) considering the current scene can be obtained from, for example, the driver's profile information and information related to the current situation (information that can be observed when boarding), For example, in the posterior probability value corresponding to the attribute of a genre, which genre's posterior probability value is absolutely high (is higher than a predetermined threshold) or relatively high (higher than other genres) It is possible to evaluate the driver's favorite genre. For example, regarding the corners that make up a program, for example, only genres that exceed a predetermined threshold can be adopted, and the length of the corners can be made dependent (proportional) on the relative size of these posterior probability values. It is. In the above example, the case where a corner is created only by genre is described as an example. However, for example, a corner can be created only by an artist (or only other attributes). When creating a corner using only artists, just as in the case of creating a corner using only the above-mentioned genre, only artists with high posterior probability values are used, or based on the relative size of the posterior probability values. It is possible to determine the length of the corner.

  Furthermore, a corner may be created by combining a plurality of attributes. For example, it is possible to create corners such as “1970s J-Pop (Japanese music) corner” and “2000s J-Pop corner” by combining genres and ages. The creation of a corner combining a plurality of attributes is useful when there are a large number of contents belonging to a certain attribute (for example, genre).

  When a corner is created by combining a plurality of attributes (for example, genre and age), the posterior probability value of the J-Pop of the genre node exceeds a certain threshold as described above, and the value is too large. In the Bayesian network model, the posterior probability value of the age node when J-Pop is set as the observed value in the genre node is evaluated in the Bayesian network model. For example, when the posterior probability values in the 1970s and 2000s exceed a certain threshold, “1970s J-Pop corner” and “2000s J-Pop corner” are adopted. The lengths of these corners can be determined by, for example, proportionally allocating a value obtained by multiplying the posterior probability value of J-Pop by the posterior probability value of each age.

  In addition to the corner attribute and corner length determined based on the evaluation result by the preference level evaluation unit 107, the program guide creation unit 108 further supports, for example, a driver supplied from the driver tendency data output unit 103. It has a function of determining the order of each corner based on the information indicating the preference of general people. Furthermore, it is inserted into each corner based on the posterior probability value of attributes such as the age and artist of the Bayesian network, and information representative of general people's preference corresponding to the driver supplied from the driver tendency data output unit 103. Detailed music contents and information contents are determined, and a program guide is created in which the determined contents of each corner are arranged along the time axis. The program guide creation unit 108 can also determine the order of each corner and each content based on the operation history information stored in the operation history information storage unit 109. The program guide created by the program guide creation unit 108 is supplied to a program playback unit (not shown), and appropriate content is played back at an appropriate timing based on the program guide, whereby the program is output.

  The operation history information storage unit 109 stores operation history information in which operations performed by the driver when listening to a program are recorded. The operation performed at the time of listening to the program is an operation representing the preference of the driver related to the listening to the program. Specifically, the skip operation for skipping the currently listened content, the content currently being listened to again. It is effective that the three operations of the repeat operation for listening and the direct selection operation for directly selecting the desired content by the driver performing the selection operation at the time of listening to the program are the operations reflecting the preference of the driver. In particular, the skip operation, the repeat operation, and the direct selection operation are operations that are naturally performed by the driver in order for the driver to meet his / her preference.

  The operation input unit 110 is an arbitrary information input interface such as a touch panel or a remote controller (remote controller). The operation input unit 110 is preferably an operation input interface that allows the driver to easily perform the above-described skip operation, repeat operation, and direct selection operation.

  Next, the process at the start of new use and the second and subsequent processes performed by the program guide creation device including the content preference learning device according to the first embodiment of the present invention will be described. FIG. 2A illustrates an example of processing at the start of new use performed by the program guide creation device including the content preference learning device according to the first exemplary embodiment of the present invention, and FIG. 2B illustrates the present invention. An example of the second and subsequent processes performed by the program guide creation device including the content preference learning device in the first embodiment is illustrated.

  In the first embodiment of the present invention, when a new program guide is created, first, as shown in FIG. 2A, first, it corresponds to the driver supplied from the driver tendency data output unit 103. The learning operation of the Bayesian network model as shown in FIG. 3 is performed using information representing general people's preference and the attribute information of the music content ripped in the content data storage unit 106 (steps). S201, S203).

  Then, after the above learning operation is completed, when the driver actually gets on and drives, the driver's profile information and information (passenger, day of the week, time, etc.) that can be observed by the boarding time information detection unit 105 When the observation value is input to the Bayesian network processing unit 104, the Bayesian network processing unit 104 performs an inference calculation of the Bayesian network, and the scene (commuting to work, commuting home, shopping, homecoming, etc.) Evaluation of posterior probability values of the attribute nodes (genre, age, artist name, news, weather forecast, traffic information, etc.) considered (inference of the preference level of each attribute node) is performed (step S205). That is, the preference level of the content attribute and the genre attribute is inferred based on the general tendency estimated from the profile information of the driver and the situation of the driver, and the attribute information of the favorite music content of the driver.

  And, for example, an attribute of the posterior probability value exceeding a predetermined threshold (for example, rock in the genre, 1970s in the age, etc.) is adopted as an attribute of the corner constituting the program, and further, according to the magnitude of the posterior probability value, The length of time allocated to the corner is determined, for example, by proportional distribution. At this time, the genre to be adopted as the corner can be determined based on a value obtained by multiplying the posterior probability value of the genre node by a random value. In addition, as described above, a corner can be created by combining a plurality of attributes. Similarly, for information contents such as news, only news field names whose posterior probability values exceed a predetermined threshold may be adopted as corners, and random fluctuations may be similarly introduced in the adoption. . Also, a combination of a corner (for example, “genre = rock corner” or “artist ABC corner”) determined from one value of a single attribute node and a different attribute node in the same program. It is also possible to mix corners determined by use (for example, “1970s Western music corner”, “1980s J-Pop corner”, etc.).

  Further, for example, the order of corners and contents is determined based on information representing general preference of people corresponding to the driver supplied from the driver tendency data output unit 103. Here, for example, for information content, the order of corners is determined by rules such as traffic information at the beginning of a program, news and weather being continued. As for music, it is possible to determine the order of corners and contents in a program based on the relative order preferred between genres, or to determine each content at random. Thus, the program configuration at the start of use (a newly created program configuration) is determined (step S207).

  Through the above processing, the corner attributes, the corner lengths, and the order of the corners are determined, and the new program configuration is completed.

  On the other hand, in the first embodiment of the present invention, when an existing program table is updated, as shown in FIG. 2B, the driver is listening to the program. The content skip operation, repeat operation, and direct selection operation that are naturally performed to match the user's preference are recorded as operation history information (step S211). The operation history information is recorded so that it is possible to grasp at which timing of the program (while content is being played back).

  Then, for example, at the time when the listening of the program is finished and the process is in an idle state (or any timing), the Bayesian network is learned based on the recorded operation history information (step S213). At this time, basically, the skip-operated content does not meet the driver's preference, the repeat-operated content matches the driver's preference, and the content selected as a result of the direct selection operation is the driver's preference. It is desirable that learning that reflects these tendencies be performed as what suits the preference or what is desired to be heard at that timing. For example, when the music content data stored in the content data storage unit 106 is updated, it is desirable that learning based on the update information is also performed at the same time. Furthermore, for example, a reproduction history (such as the number of reproductions and total reproduction time) of each content is recorded, and Bayesian network learning using this reproduction history may be performed. At this time, for example, it is desirable to perform learning that reflects this tendency, assuming that content with a high number of playback times and a total playback time value matches the preference of the driver.

  After this learning operation, as in step S205, when the driver actually gets on and drives, the driver's profile information and information that can be observed by the boarding time information detection unit 105 (passenger, day of the week, time, etc.) ) Is input to the Bayesian network processing unit 104, the Bayesian network processing unit 104 performs inference calculation of the Bayesian network, and scenes (commuting to work, commuting home, shopping, homecoming, etc.) ) Is evaluated (inference of the degree of preference of each attribute node) of attribute nodes (genre, age, artist name, news, weather forecast, traffic information, etc.) in consideration of (). Further, similarly to step S207, the order of corners and the order of contents are also determined, and the program structure (updated program structure) is determined (step S217).

  With the above processing, for example, for information content such as news, basically, the content that was previously skipped is deleted, and if a direct selection operation is performed, the timing of direct selection within the program The content of the attribute directly selected in is inserted.

For example, FIG. 4A partially illustrates an example of a program guide before the driver's operation is reflected. During program playback based on this program guide,
・ Repeat operation was performed on the music content in the first music corner ・ Direct selection operation of weather forecast was performed in the middle of the first music corner ・ Direct selection operation of economic news was performed in the news corner Suppose a skip operation is performed on the music content in the second music corner.

In this case, when the operation history information is learned by the Bayesian network and the program is created next, as a result,
-Increase the time of the first music corner-Place the weather forecast corner before the first music corner-Change the attribute of the news corner from latest news to economic news-Reduce the time of the second music corner As described above, the program guide reflecting the previous operation history by the driver is updated.

  As described above, according to the first embodiment of the present invention, the user (driver) matches the preference of the user (driver) without performing troublesome preparatory operations such as setting of preference information. Information provision (determination of program structure and program content) can be received from the start of use. Also, according to the first embodiment of the present invention, based on only basic operations for adjusting information provided to the user's own preferences, such as skip, repeat, direct selection, etc. Information that more accurately reflects the user's preference is provided.

<Second Embodiment>
Next, a second embodiment of the present invention will be described. FIG. 5 shows an example of the configuration of a program guide creation device including a content preference learning device according to the second embodiment of the present invention. Note that the program guide creation device 500 shown in FIG. 5 also includes a content preference learning device and is incorporated in the in-vehicle device of the vehicle, like the program guide creation device 100 shown in FIG. And

  5 includes a driver profile information storage unit 101, a general tendency database 102, a driver tendency data output unit 103, a Bayesian network processing unit 104, a boarding time information detection unit 105, and content data storage. Section 106, preference evaluation section 107, program guide creation section 508, operation history information storage section 109, operation input section 110, content score calculation section 501, and content score storage section 502. In FIG. 5, each function is represented by a function block, but each function can be realized by hardware and / or software (a program executable by a computer).

  Among the functional blocks shown in FIG. 5, a driver profile information storage unit 101, a general tendency database 102, a driver tendency data output unit 103, a Bayesian network processing unit 104, a boarding time information detection unit 105, and a content data storage unit 106 The function of the preference level evaluation unit 107 is basically the same as that installed in the program guide creation apparatus 100 shown in FIG. 1, and a description thereof is omitted here.

  On the other hand, the content score calculation unit 501 sets an independent preference level (content score) for each content stored (or storable) in the content data storage unit 106. Specifically, the content score calculation unit 501 acquires a list of content data stored (or storable) in the content data storage unit 106, and initially, for example, a driver tendency for each content. A content score indicating a general people's preference level (so-called general popularity level of each content) corresponding to the driver supplied from the data output unit 103 is set, and set to the identification information of each content and each content. Information associated with the content score (for example, a content score list in which the correspondence between the content title and the content score is known) is stored in the content score storage unit 502. Note that the initial content score may be set based on a more general popularity ranking (for example, popularity regardless of gender or age) instead of the general people's preference corresponding to the driver profile. . Further, regarding the initial content score, the same value may be simply set for all contents. The content score calculation unit 501 simply obtains a total score (for example, the product of the content score and the attribute score) obtained by combining both the content score and the attribute score instead of the content score that represents the driver's preference for each content. Can be used to perform the processing according to the present invention.

  Further, the content score calculation unit 501 is stored in the content score storage unit 502 based on operation history information in which operations (skip operation, repeat operation, direct selection operation) performed by the driver during listening to the program are recorded. Update the content score value. For example, for content for which a skip operation has been performed, the content score of the content is decreased, and for content for which a repeat operation has been performed, an update process for increasing the content score of the content is performed. Further, for example, the content score may be set according to the reproduction history (reproduction count, total reproduction time, etc.) of each content. In particular, with regard to music content, even if music having the same attribute such as the same genre, the same age, and the same artist, a situation may occur in which the driver likes a specific music and does not like another music. In this way, the music content tends to vary greatly in the degree of preference for each content, and the setting of the content score is very effective.

  In addition to the function of the program guide creation unit 108 illustrated in FIG. 1, the program guide creation unit 508 has a content score (or content) that can be acquired from the content score list stored in the content score storage unit 502. It is possible to create a program guide in consideration of the overall score obtained by combining both the score and the attribute score. For example, regarding the attributes adopted as corners and the lengths of the corners of each attribute, the program guide creation unit 508 determines the driver's preference level and generality from the state determined by the processing in the Bayesian network processing unit 104. After determining the order of corners in consideration of the degree of preference, it is possible to preferentially select content with a high content score when selecting content to be inserted into each corner and determining the order of content. Is possible. If content is selected based on the content score, similar content may be selected every time the program is structured. Therefore, fluctuations due to random numbers are added and the content score is high. Instead of selecting only the content, the content may be selected with randomness.

  Next, the process at the start of new use and the second and subsequent processes performed by the program guide creation apparatus including the content preference learning apparatus according to the second embodiment of the present invention will be described. FIG. 6A illustrates an example of processing at the start of new use performed by the program guide creation device including the content preference learning device according to the second exemplary embodiment of the present invention, and FIG. 6B illustrates the present invention. An example of the second and subsequent processes performed by the program guide creation device including the content preference learning device according to the second embodiment is illustrated.

  In FIG. 6A, in the second embodiment of the present invention, the Bayesian network model learning operation, Bayesian network inference calculation, as in steps S201 to S205 (see FIG. 2A) in the first embodiment of the present invention, Evaluation of the posterior probability value of each attribute node (inference of the preference level of each attribute node) is performed. In the second embodiment, the preference level of the attribute node determined by the Bayesian network may be referred to as the attribute score in contrast to the content score described above. The content score is a score determined for each individual content, and the attribute score is a score determined for each individual attribute (genre, age, artist).

  In parallel with steps S201 to S205, the content score calculation unit 501 sets an initial value of the content score for each content (step S601). As described above, the initial value of the content score may be set with reference to general popularity, or the same value may be simply set for all content scores. Then, as described in the processing of step S207 illustrated in FIG. 2A, the program guide creation unit 508 determines the attributes of the corners constituting the program, the length of each corner, and the order of the corners based on the attribute score. Furthermore, it is possible to determine the priority of the content to be placed in each corner (content to be inserted in each corner) based on the content score of each content (step S603). As a result, a program guide is created based on the content score, and a program guide that more reflects the degree of preference for each content can be created. In addition, by creating a program guide based on both the content score and the attribute score, it is possible to create a program guide that more reflects the degree of preference of the driver.

  In addition, you may add the attribute showing the preference degree of the driver with respect to the attribute of each content to the content score showing the preference degree of the driver with respect to each content. That is, a value obtained as a result of the combination of the content score for each content and the attribute for each content attribute (for example, the product of the content score and the attribute score) is set as the total score, and in the above-described step S603, the total score is obtained. You may determine the priority of the content arrange | positioned in each corner based on a score. In the operations described in this specification, an operation performed on the content score can be read as an operation performed on the total score obtained from the combination of the content score and the attribute score.

  In FIG. 6B, in the second embodiment of the present invention, when an existing program table is updated, steps S211 and S213 (FIG. 2B) in the first embodiment of the present invention are performed. In the same manner as in (Refer), the operation history information is recorded, and the attribute score is calculated in consideration of the operation history information.

  In parallel with steps S211 and S213, the content score calculation unit 501 refers to the recorded operation history information and basically decreases the content score of the content that is the target of the skip operation, and repeat operation. The content score is updated so as to increase the content score of the content to be targeted (step S611). In the same manner, for example, the content score of the content selected by the direct selection operation may be increased, or the content score of the content whose playback has been stopped halfway by the direct selection operation may be decreased.

  Then, as described in the process of step S215 illustrated in FIG. 2B, the program guide creation unit 508, based on the attribute score changed in consideration of the operation history information, the attributes of the corners constituting the program, It is possible to determine the length of each corner and the order of the corners, and furthermore, it is possible to determine the priority of content to be placed in each corner (content to be inserted in each corner) based on the content score of each content ( Step S613). As a result, the content score that reflects the preference of the driver is basically updated based only on the history of operations performed by the driver. In addition, both the content score and the attribute score are basically updated based on the driver operation history so as to reflect the driver's preference, so that the driver operation history can be recorded and stored easily. Only with simple processing, it is possible to create a program guide that more reflects the degree of preference of the driver.

  Further, in FIG. 7, in the second embodiment of the present invention, from the song score and the attribute score at the start of use (first program configuration), the song score after the start of use (the program configuration after the second program) and A state in which the attribute score is changed is schematically illustrated.

  As schematically shown in FIG. 7, at the start of use (first time), the content score determined independently for each content is, for example, a preference for attributes based on general popularity. The attribute score is initially set based on, for example, the driver profile and the attribute of the music content ripped on the HDD.

  On the other hand, when the operation history information of the driver is accumulated after the start of use, an update process based on the operation history information of the driver is performed on each of the content score and the attribute score. In the case of the attribute score, as a result of the learning operation in which the operation history information of the driver is reflected in the Bayesian network, the attribute score specified by the conditional probability table (and the posterior probability value) of each attribute node is obtained. Updates are made. On the other hand, regarding the content score, the content score is updated by changing the value of the content score so as to reflect the operation history information of the driver.

  As described above, according to the second embodiment of the present invention, the program table is created based on the content score, and the degree of preference for each content is more reflected. A program guide can be created. In addition, by creating a program guide based on both the content score and the attribute score, it is possible to create a program guide that more reflects the degree of preference of the driver. Further, this makes it possible to update the content score that reflects the driver's preference based on only the operation history performed by the driver. In addition, both the content score and the attribute score are basically updated based on the driver operation history so as to reflect the driver's preference, so that the driver operation history can be recorded and stored easily. Only with simple processing, it is possible to create a program guide that more reflects the degree of preference of the driver.

  Next, the definitions of the skip operation and the repeat operation described above are supplemented. In particular, the skip operation and the repeat operation are operations that are naturally performed by the driver in order for the driver to match his / her preference.

  The skip operation is basically an operation of interrupting the reproduction of the currently listened content and starting the reproduction of the next content on the program. In the first and second embodiments of the present invention described above, it is possible to uniformly determine the learning method for the skip operation (that is, simply refer to the presence or absence of the skip operation). Even if it exists, you may consider the time, the skip frequency, etc. which listened to the content. Specifically, if you listen to content almost to the end and skip it, you can not treat it as skipped, or you can adjust the learning level considering the time you listened to the content before skipping operation is there.

  For example, as a learning method for the conditional probability table of the Bayesian network model performed in the first and second embodiments of the present invention described above, a ratio of a part of a song that was not listened to (performed by performing a skip operation) The ratio of decreasing the corresponding conditional probability value in the conditional probability table may be determined in proportion to (the length of time of the missing portion / the length of the entire song). Similarly, the content score learning method performed in the above-described second embodiment of the present invention similarly determines the rate of decreasing the content score in proportion to the proportion of the portion of the song that was not listened to. May be. In addition, when the frequency of skip operations is very high, it may be determined that the skip operation is not performed because it does not match the preference, and the skip operation may not be adopted as the learning target data.

  In addition, the repeat operation is basically an operation of restarting the reproduction of the content that is being listened to at or after the end of the reproduction of the currently listened content, or after the reproduction of the content is interrupted. In the first and second embodiments of the present invention described above, it is possible to uniformly determine the learning method for the repeat operation (that is, simply refer to the presence or absence of the repeat operation). Alternatively, the time of listening to the content, the repeat frequency, etc. may be taken into consideration. Specifically, if you repeat without listening to the content, it is determined that you did not repeat it to suit your tastes, so that it was not treated as a repeat operation, or the time you listened to the content before performing the repeat operation It is possible to adjust the learning degree in consideration of the above.

  For example, as a learning method for the conditional probability table of the Bayesian network model performed in the above-described first and second embodiments of the present invention, the ratio of the listened part in one song (listening before skip operation is performed) The ratio of increasing the corresponding conditional probability value in the conditional probability table may be determined in proportion to (the length of time of the part / the length of the entire song). Similarly, the content score learning method performed in the above-described second embodiment of the present invention similarly determines the ratio of increasing the content score in proportion to the ratio of the listened part in one song. Also good.

  According to the present invention, the user can receive information (program configuration and program content selection) that matches the user's preference from the start of use without performing troublesome preparatory operations such as setting preference information. Therefore, the present invention can be applied to a technique for providing information to a user in consideration of the user's preference and a technique for learning the user's preference for that purpose. Further, the present invention is applied to a technique for creating a program reflecting user preferences and providing a program that meets the user preferences, or a technique for selecting a restaurant or other stop-off facility reflecting the user preferences. Further, the present invention can be applied to a technique for providing information to a driver who drives a vehicle.

DESCRIPTION OF SYMBOLS 100, 500 Program schedule preparation apparatus 101 Driver profile information storage part 102 General tendency database 103 Driver tendency data output part 104 Bayesian network processing part 105 Boarding time information detection part 106 Content data storage part 107 Preference degree evaluation part 108,508 Program guide Creation unit 109 Operation history information storage unit 110 Operation input unit 501 Content score calculation unit 502 Content score storage unit

Claims (22)

Selectively reading information stored in advance in the mobile unit mounting apparatus mounted on the mobile unit;
In the stochastic inference model processing unit for performing probabilistic inference processing, performing learning processing related to the inference processing based on the information read selectively;
Detecting a situation where a user is on the moving body;
Performing an inference process in the probabilistic inference model processing unit on the detected situation, inferring an attribute related to the situation from a plurality of predetermined attributes;
Adding a content score representing the user's preference for each of the information that can be provided to the user, and storing the content score for each of the information that can be provided to the user;
A target to be provided to the user while selecting the information to be provided to the user from the attributes inferred to be related to the detected situation and the information that can be provided to the user based on the content score Determining how to provide said information selected as:
Information providing support method. Selectively reading information stored in advance in the mobile unit mounting apparatus mounted on the mobile unit;
In the stochastic inference model processing unit for performing probabilistic inference processing, performing learning processing related to the inference processing based on the information read selectively;
Detecting a situation where a user is on the moving body;
Performing an inference process in the probabilistic inference model processing unit on the detected situation, inferring an attribute related to the situation from a plurality of predetermined attributes;
Adding a content score representing the user's preference for each of the information that can be provided to the user, and storing the content score for each of the information that can be provided to the user;
An attribute score indicating the user's preference degree defined by the probabilistic inference model processing unit with respect to an attribute of each of the information that can be provided to the user, and the contents of the information that can be provided to the user Determining an overall score for each of the information that can be provided to the user based on the score; and
A target to be provided to the user while selecting the information to be provided to the user from among the attributes inferred to be related to the detected situation and the information that can be provided to the user based on the total score Determining how to provide said information selected as:
Information providing support method.   The information providing support method according to claim 1, further comprising a step of generating a random number and selecting information to be provided to the user from information that can be provided to the user in consideration of the random number.   The information provision support method according to claim 2, further comprising: generating a random number and selecting information to be provided to the user from information that can be provided to the user in consideration of the random number. Providing the user with the information based on information selected as a target to be provided to the user and a method of providing the information;
Recording operation performed by the user as operation history information during the provision of the information;
In the stochastic inference model processing unit, performing a learning process based on the operation history information,
5. The information provision support method according to claim 1, further comprising: Providing the user with the information based on information selected as a target to be provided to the user and a method of providing the information;
Recording operation performed by the user as operation history information during the provision of the information;
Changing the content score based on the operation history information;
5. The information provision support method according to claim 1, further comprising:   The information providing support method according to claim 5 or 6, wherein the operation performed by the user recorded as the operation history information includes at least one of a skip operation, a repeat operation, and a direct selection operation. When the skip operation or the repeat operation is performed while predetermined information is being provided, a step of recording a time until the predetermined information is provided;
In the stochastic inference model processing unit, performing a learning process based on the time together with the operation history information,
The information providing support method according to claim 7.   9. The method according to claim 1, further comprising: creating a program having an attribute corner inferred to be related to the situation in selecting information to be provided to the user and determining a method for providing the information. Information providing support method described in 1.   10. The information provision according to claim 9, wherein, when a program including a plurality of attribute corners inferred to be related to the situation is created, the length of each corner is proportionally distributed according to the degree of preference regarding each attribute. Support method.   The information providing support according to claim 9, wherein a combination of a plurality of attributes is inferred to infer the attribute combination related to the situation, and a program having a corner of the attribute combination inferred to be related to the situation is created. Method. Means for selectively reading out information stored in advance in the mobile unit mounting apparatus mounted on the mobile unit;
Probabilistic inference model processing means for performing probabilistic inference processing and performing learning processing related to the inference processing based on the selectively read information;
Situation detecting means for detecting a situation in which a user is on the moving body;
Content score giving means for attaching a content score representing the user's preference for each of the information that can be provided to the user;
Content score storage means for storing the content score for each of the information that can be provided to the user;
An inference process is performed on the detected situation by the probabilistic inference model processing means, an attribute related to the situation is inferred from a plurality of predetermined attributes, and it is inferred that the situation is related And selecting the information to be provided to the user from among the information that can be provided to the user based on the attribute and the content score, and providing the information selected as a target to be provided to the user Information providing support means to be determined
An information providing support device. Means for selectively reading out information stored in advance in the mobile unit mounting apparatus mounted on the mobile unit;
Probabilistic inference model processing means for performing probabilistic inference processing and performing learning processing related to the inference processing based on the selectively read information;
Situation detecting means for detecting a situation in which a user is on the moving body;
Content score giving means for attaching a content score representing the user's preference for each of the information that can be provided to the user;
Content score storage means for storing the content score for each of the information that can be provided to the user;
An attribute score indicating the user's preference degree defined by the probabilistic inference model processing means with respect to the attributes of each of the information that can be provided to the user, and the contents of the information that can be provided to the user A total score determining means for determining a total score of each piece of information that can be provided to the user based on the score;
An inference process is performed on the detected situation by the probabilistic inference model processing means, an attribute related to the situation is inferred from a plurality of predetermined attributes, and it is inferred that the situation is related And selecting the information to be provided to the user from among the information that can be provided to the user based on the attribute and the total score, and providing the information selected as a target to be provided to the user Information providing support means to be determined
An information providing support device.   13. The information provision support apparatus according to claim 12, further comprising means for generating a random number, wherein information to be provided to the user is selected from information that can be provided to the user in consideration of the random number.   14. The information provision support apparatus according to claim 13, further comprising means for generating a random number, wherein information to be provided to the user is selected from information that can be provided to the user in consideration of the random number. Information providing means for providing the information to the user based on information selected as a target to be provided to the user and a method of providing the information;
An operation history information recording unit that records an operation performed by the user as operation history information during the provision of the information;
In the stochastic inference model processing unit, model update control means for performing learning processing based on the operation history information,
The information provision support device according to any one of claims 12 to 15. Information providing means for providing the information to the user based on information selected as a target to be provided to the user and a method of providing the information;
An operation history information recording unit that records an operation performed by the user as operation history information during the provision of the information;
Content score changing means for changing the content score based on the operation history information;
The information provision support device according to any one of claims 12 to 15.   The information providing support device according to claim 16 or 17, wherein the operation performed by the user recorded as the operation history information includes at least one of a skip operation, a repeat operation, and a direct selection operation.   When the skip operation or the repeat operation is performed while predetermined information is being provided, the apparatus includes an elapsed time information recording unit that records a time until the predetermined information is provided, and the probabilistic The information provision support apparatus according to claim 18, wherein an inference model processing unit performs learning processing based on the time together with the operation history information.   The information providing support device according to any one of claims 12 to 19, wherein the information providing support means is configured to create a program having a corner having an attribute inferred to be related to the situation.   When the information providing support means creates a program that includes a plurality of attribute corners that are inferred to be related to the situation, the length of each corner is proportionally distributed according to the degree of preference for each attribute. The information provision support device according to claim 20, which is configured as follows.   The information providing support means combines a plurality of attributes to infer the attribute combination related to the situation, and creates a program having a corner of the attribute combination inferred to be related to the situation. The information providing support device according to claim 20, which is configured.