JP4712339B2 - Scene determination apparatus, scene determination method, program for causing computer to execute the method, information distribution apparatus, and information distribution system - Google Patents

Description

  The present invention relates to a scene determination device, a scene determination method, a program for causing a computer to execute the method, an information distribution device, and an information distribution system.

  In recent years, services for distributing multimedia information such as video scenes to mobile terminals and personal computers (PCs) have begun. Especially for sports games whose situation changes from moment to moment, a service that extracts important scenes in real time and distributes them to users is popular.

  For example, if a user's favorite player plays an active role, information on the scene that was active immediately can be distributed, or if the favorite team can distribute these scenes in real time when there is a chance to put a first goal or a reversal opportunity Therefore, the user's interest can be further increased with respect to the subsequent game development. Further, if the user who has started viewing from the middle can be provided with a digest that extracts important scenes up to the middle of the viewing start time, the understanding of the content of the game will progress, and it will be effective for the subsequent viewing promotion. In other words, there is currently a need to select and distribute important scenes accurately.

  In order to meet this need, the distribution server of the distribution source has a database, transmits the first content to the user's client machine, the user sends back the second content, and the distribution server sends back the second content. A technique has been considered in which distribution is performed in response to the user's needs more accurately by collecting the total contents of the contents and updating the database for use in distribution (Patent Document 1).

  Also, program video and program related information are input, program related content is generated based on a predefined tree structure, and based on the amount of viewing data set by each individual user having a mobile terminal in advance as a viewing request. For each user, a leaf node in the tree structure of program-related content is selected as a presentation candidate, and the presentation path starting from the root of the tree structure and sequentially following the leaf nodes that are presentation candidates is used as a presentation scenario. A technique has been devised for generating a presentation scenario for each user and sending the generated presentation scenario to the mobile terminal of the corresponding user. With this technology, personalized information for each user can be passively viewed without complicated interaction in a limited display area such as on a mobile device, and the information that the viewer wants to see can be viewed. It can be seen in the desired order (Patent Document 2).

JP 2002-92027 A JP 2003-32654 A

  However, in the technique of Patent Document 1, there is a problem that if the user's client machine does not send back the second content to the distribution server, the distribution server cannot accurately distribute the user's needs.

  Moreover, in the technique of patent document 2, there was a problem that important scenes could not be automatically determined and extracted in, for example, the progress of sports matches and the results of games that progress in time.

  In general, important scenes such as sports include "scenes that were good as a result" such as preemptive and reversal. However, it is not only a scene that brought about such a remarkable result, but it was an opportunity even if it did not become an analogy, and as a result the opportunity did not bear fruit. "A good scene" is also an important scene. When manually delivering a service that distributes important scenes including such scenes manually with a bulletin, fully understand the content of the game, and `` a scene that was good as a result '' and `` a scene that did not accompany the result, but a good scene The worker who performs the work while judging the importance of "" every moment is essential, and a lot of costs are required.

  On the other hand, in order to provide such a distribution service automatically instead of manually, a technology is needed to automatically determine the importance of both “scenes that are good as a result” and “scenes that are not accompanied by results” It was. In particular, in an ongoing game, it was not possible to extract the important scenes in progress due to the tension caused by facing the unpredictable reality of not knowing what will happen for a moment. There was a problem that it was not an extraction technique.

  The present invention has been made in view of the above problems, and its purpose is to generate a scene from a scene of an event that occurs, a scene determination device capable of more accurately extracting a scene important for a viewer who views the content, A scene determination method, a program for causing a computer to execute the method, an information distribution apparatus, and an information distribution system are provided.

In order to solve the above-described problems and achieve the object, the invention according to claim 1 is a scene determination device, which includes a change in the content of a content including video and text and having metadata added as a state transition. , Analyzing the prediction importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the expression means expressed using the metadata and the state transition expressed by the expression means, Determination means for determining an important scene, and the expression means includes a win probability that is a probability that a home team wins at a point in the course of the sport game when the content is a sport game. The state transition is expressed by a feature amount, and the determination unit predicts all the states that occur in the state transition and predicts the feature in the state. Get the total value of the expected value, based on a difference between the feature quantity before the transition and the sum of the expected values, the prediction that the importance calculation to analysis, characterized by.

Further, the invention according to claim 2 is a scene determination method, which includes a video and a text, and represents a change in content of the content to which metadata is added as a state transition and is expressed using the metadata. Analyzing the prediction importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition expressed in the expression step, and determining the important scene , In the expression step, when the content is a sport game, the state transition is expressed by a feature amount including a victory probability that is a probability that a home team wins at the time when the sport game is in progress. The determination step predicts all states that occur in the state transition and obtains a total value of expected values of the feature values in a predicted state, and calculates the total of the expected values. On the basis of the difference between the feature quantity before the transition and the prediction that the importance calculation to analysis, characterized by.

The invention according to claim 3 is a program, which causes a computer to execute the method according to claim 2 .

According to a fourth aspect of the present invention, there is provided an information distribution apparatus including content generation means for generating content including video and text and having metadata added thereto, and change in content of the content generated by the content generation means. Predictive importance that is the importance of the scene before the transition based on the state that is predicted to occur after the transition due to the state transition represented by the metadata and the state transition represented by the metadata Analyzing means, and determining means for determining an important scene; and distribution means for distributing content determined to be an important scene by the determining means , wherein the expressing means is a game of sports If there is a feature amount including a victory probability, which is a probability that the home team will win at the time when the game of the sport is in progress, Expressing the state transition, the determination means predicts all the states that occur in the state transition, obtain a total value of the expected value of the feature amount in the predicted state, the total value of the expected value and the The prediction importance is calculated and analyzed based on a difference from the feature amount before transition .

According to a fifth aspect of the present invention, there is provided an information distribution system, each of which is connected by a network, generates a content including video and text, and having metadata added thereto, and the content generation device generates the content The content is received via the network, a change in the content of the received content is expressed as a state transition, and expressed using the metadata, and the state transition expressed by the expression means Analyzing the predictive importance, which is the importance of the scene before the transition based on the state predicted to occur, and a scene judging device having a judging means for judging the important scene, the scene judging device judges that it is an important scene A distribution device for distributing the received content via the network, and distribution by the distribution device And an information terminal receiving via the network the content to be, the expression means, when the content of the content is a sports game, the probability that the home team to win at the time during the course of the game the sport The state transition is expressed by a feature amount including a victory probability that is, and the determination unit predicts all states that occur in the state transition and calculates a total value of expected values of the feature amount in a predicted state. And obtaining and calculating the prediction importance based on a difference between the total value of the expected values and the feature amount before the transition .

The invention according to claim 6 is the information distribution system according to claim 5 , wherein the information terminal is a mobile phone.

The invention according to claim 7 is the information distribution system according to claim 5 , wherein the information terminal is a portable information terminal.

  According to the first aspect of the present invention, the expression means expresses a change in the content of the content including the video and the text, to which the metadata is added, using the metadata as a state transition. The determination means analyzes the predicted importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the expressed state transition, and determines the important scene. With this configuration, it is possible to determine the important scene by analyzing the predictive importance, which is the importance of the scene before the state transition as seen from the state predicted after the state transition, so as a result no significant change was made in the scene Even so, it is possible to provide a scene determination device that can determine an important scene that may cause a large change in the scene.

In addition, according to the invention of claim 2 , there is provided a scene determination method, in which a change in content of content including video and text, to which metadata is added, is expressed using metadata as a state transition. And a determination step of analyzing the prediction importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition expressed in the expression process, and determining the important scene. With this configuration, it is possible to determine the important scene by analyzing the predictive importance, which is the importance of the scene before the state transition as seen from the state predicted after the state transition, so as a result no significant change was made in the scene Even so, it is possible to provide a scene determination method capable of determining an important scene that may cause a large change in the scene.

The invention according to claim 3 is a program, and has the effect of causing a computer to execute the method according to claim 2 .

According to the fourth aspect of the present invention, the content generation means generates content including video and text and having metadata added thereto. The expression means expresses a change in the content generated by the content generation means as a state transition using metadata. The determination means analyzes the prediction importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition expressed by the expression means, and determines the important scene. The distribution unit distributes the content determined to be an important scene by the determination unit. With this configuration, it is possible to determine the important scene by analyzing the predictive importance, which is the importance of the scene before the state transition as seen from the state predicted after the state transition, so as a result no significant change was made in the scene However, it is possible to provide an information distribution apparatus that can determine and distribute an important scene that may cause a large change in the scene.

According to the fifth aspect of the present invention, the content generation device generates content including video and text and having metadata added thereto. The scene determination device receives the content generated by the content generation device via the network, and expresses the change of the content of the received content as a state transition using the metadata, and is expressed by the expression means It has a determination means for analyzing the predicted importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition, and determining the important scene. The distribution device distributes the content determined as the important scene by the scene determination device via the network. The information terminal receives the content distributed by the distribution device via the network. With this configuration, it is possible to determine the important scene by analyzing the predictive importance, which is the importance of the scene before the state transition as seen from the state predicted after the state transition, so as a result no significant change was made in the scene However, it is possible to provide an information distribution system that can determine and distribute an important scene that may cause a large change in the scene, and receive and view the information by an information terminal.

Moreover, according to the invention concerning Claim 6, there exists an effect that the information delivery system which used the mobile telephone as an information terminal can be provided.

Moreover, according to the invention concerning Claim 7, there exists an effect that the information delivery system using a portable information terminal as an information terminal can be provided.

  DETAILED DESCRIPTION Exemplary embodiments of a scene determination device, a scene determination method, a program for causing a computer to execute the method, an information distribution device, and an information distribution system according to the present invention are described below with reference to the accompanying drawings. 4 will be described in detail.

  In the following description, a scene determination device according to an embodiment of the present invention is applied to an information distribution system that generates professional baseball game content into multimedia content including metadata and distributes it to an information terminal that is a client. An example will be described. However, the present invention is not limited to the examples described below.

(1. Embodiment 1)
FIG. 1 is a functional block diagram of a scene determination apparatus according to Embodiment 1 of the present invention. FIG. 2 is a diagram illustrating an example in which the scene determination device according to Embodiment 1 is applied to an information distribution system.

  The information distribution system includes an information distribution device 100 and information terminals 71, 72, and 73 connected via a network 5. The information distribution device 100 includes a content generation device 50, a scene determination device 10, and a distribution device 60. The scene determination device 10 included in the information distribution device 100 constituting this information distribution system is the scene determination device according to the first embodiment.

  The information distribution apparatus 100 receives the game data from the baseball stadium 40 where the professional baseball game is played, and the game data from time to time via the network 5. The content generation device 50 generates content for distribution to which metadata is added from received audio, video, text data, and the like.

  The scene determination device 10 receives the content generated by the content generation device 50, reads the metadata, calculates importance using the read metadata, and determines an important scene from the calculated importance.

  The distribution device 60 distributes the content determined to be important by the scene determination device 10 to the information terminals 71 to 73 that are distribution destinations via the network 5. Here, the viewer does not need to have three information terminals as shown in the figure, but can be any plural number that can be connected to the network.

  The information terminals 71 to 73 are, for example, mobile phones, and the viewer is determined to receive the content that is determined to be important by the scene determination device 10 and has already been extracted and determined according to the importance. The contents can be viewed efficiently.

  Here, as the baseball game progresses, the technology for generating and distributing content with added metadata is a publicly known technology. For example, the publicly known technology including the technology described in Patent Document 2 by the present inventor is known. The detailed description is omitted.

(1.1. Predictive importance and result importance)
The scene determination apparatus 10 according to Embodiment 1 includes an expression unit 1, an attribute determination unit 2, a prediction determination unit 3, and a progress determination unit 4 (FIG. 1). The expression unit 1 uses the metadata added to the content to express the change in the content of the content in which the game content is recorded as a state transition, using the added metadata.

  The attribute determination unit 2 analyzes the importance determined by the attributes of players, teams, and the like in the state transition expressed using the metadata by the expression unit 1 and determines the importance as attribute importance. The attribute importance is, for example, a degree representing the popularity of the appearing player or team.

  Here, in general, important scenes are: (1) Predicting the end (after the transition) before the end of the event (before the transition) There are two types of scenes: scenes with higher degrees, and (2) scenes with high importance before the end (before transition) as a result after the event ends (after transition). The former is, for example, a case where two-out is full, but no points are entered, and the latter is, for example, a case where the game is actually reversed due to, for example, a reversal full hit.

  The former importance level is referred to as a prediction importance level because it is the importance level of the scene before the transition viewed from the state predicted to occur after the transition due to the state transition. The latter importance level is referred to as a result importance level because it is an importance level of a scene before the transition as viewed from a state generated as a result after the transition due to an actual state transition.

  The prediction determination unit 3 calculates the prediction importance, that is, the importance of each scene before the transition in the state predicted to occur after the transition by the state transition expressed by the expression unit 1.

  The progress determination unit 4 is a progress scene that is an ongoing importance level that determines the importance level before the transition from the state predicted after the transition during the progress of the game based on the predicted importance level calculated by the prediction determination unit 3 Determine the importance. As a result, it is possible to determine the importance of the progress scene in the scene that is in progress and before the transition of the situation is completed, so that the progress determination unit 4 is proceeding with the scene where the tension increases before the transition ends. It can be determined as an important scene.

  Here, in addition to the prediction importance determined by the prediction determination unit 3 described above, the progress determination unit 4 considers the attribute importance by the player or team determined by the attribute determination unit 2, and determines the progress scene importance. It is desirable to do. Since the progress determination unit 4 determines the importance of the progress scene by determining the importance of the attribute and also determining the importance of the progress scene in consideration of the popularity of the players and teams, the ongoing scene with further increased tension Can be accurately determined.

  Hereinafter, the attribute importance determined by the attribute determination unit 2, the prediction importance determined by the prediction determination unit 3, and the progress scene importance determined by the progress determination unit 4 will be described in more detail.

(1.2. Judgment of important scenes based on Markov processes)
FIG. 3 is a diagram for explaining state transition of a game in a baseball game. In general, the progress of a game can be expressed by a state transition from one state to the next state.

  The baseball game starts from an initial state q0 (inning 1, score difference 0, outcount 0, no runner, batter Ka, pitcher Ku). For example, when the first batter hits a hit from this initial state (inning 1, score difference 0, outcount 0, runner 1 base, batter Is, pitcher Ku), the state transitions.

  On the other hand, when it is out, it changes to the state of (Inning 1, Score difference 0, Out count 1, No runner, Batter Is, Pitcher Ku). That is, a change in the situation in sports can be expressed as a state transition based on an event that occurred in a certain state (in the case of baseball, an event such as a hit or an out).

  FIG. 4 is another diagram for explaining the state transition of a game in a baseball game. In one inning in a baseball game, the number of runners satisfying the base is 25. Because out count is no out, one out, two out, and three out, and from no out to two out, the runner on top is {no runner, 1st, 2nd, 3rd, 12th, 13th, 13th, There are 8 runners (23km, full) and no three-out runners. That is, 8 × 3 + 1 = 25.

  And, in the case of runners on the base, state transitions, baseball scores are brought about by the state transition of how runners fill the base and reach the main base from the state without runners . For example, the transition from the no-runner state to the no-runner state without increasing the outcount is a state transition when the batter hits a home run and enters one point.

  When determining the result importance that is the importance of the “good scene” by applying a Markov process to the state transition of such a game, the importance that occurred as a result before and after the state transition Judgment is made by comparing the result importance. Then, when determining the prediction importance that is the importance of “a good scene with no result”, the importance before the transition in which the next transition state is predicted to be important is determined.

(1.3. Definition of state Q)
Here, terms used in the following description are defined. The state Q in the sports game is defined as follows.

  The state Q is expressed by items of time, H (home) team identifier, A (away) team identifier, score difference, character identifier, other attributes, and victory probability (WP). That is, it is expressed as state Q (time, H (home) team identifier, A (away) team identifier, score difference, character identifier, other attributes, victory probability (WP)). Each item of the state Q is as follows.

  The time is not only absolute time but also complex time information including various structural times according to the genre of sports (inning number for baseball, first half, second half and elapsed time for soccer, etc.).

  The H team identifier and the A team identifier are identifiers of the home team and the away team. The score difference is defined as a difference obtained by subtracting the away team score from the home team score. The character identifier is an identifier of a player who determines the state. If it is baseball, it is a pitcher and a batter, and if it is soccer, it is a plurality of players around the ball in the game. The other attributes are various attribute values depending on the genre of sports. For example, in the case of baseball, it is an outcount or a deduction situation.

(1.4. Victory probability)
The win probability (WP) is defined as the probability that the home team wins based on the current score difference in the occurrence time of a certain state in the game. In general, the probability of victory tends to be as follows.
(1) WP increases as the difference in scores {(home team score) − (away team score)} increases. Moreover, WP becomes small if it becomes small.
(2) When the score of the home team is higher (lower) than the score of the away team, the WP value becomes larger (smaller) as the remaining time of the game becomes shorter. The inventor obtained an example of the victory probability from past examples.

  FIG. 5 is a graph showing an example of the probability of winning in a baseball game. Each figure of FIG. 5 represents the winning probability at the start of each inning, that is, each score difference in no out, north strike, and no ball in each inning. The X-axis is the score difference and the Y-axis is the victory probability. This probability of victory is statistically determined from 100 professional baseball games in Japan in 2003. When the number of matches is too small to determine the probability of victory, an approximation was made. Here, WP in the state qi is represented as WP (qi).

(1.5. Definition of prediction importance)
In this embodiment, the importance of a scene is abbreviated as follows. The predictive importance is abbreviated as SI (Subjunctive Impulse), the result importance is abbreviated as RI (Result Importance), and the attribute importance is abbreviated as AI (Attribute Importance). In the first embodiment, SI and AI are used, and in the second embodiment to be described later, each scene importance is determined using RI.

  The prediction importance SI is the importance of the scene before the transition in the state predicted to occur after the transition, as already described. Here, it is assumed that the probability of occurrence of a state at a certain point in the course of the game depends on the state immediately before that.

A certain state qi transitions to a state qi + 1 due to an event occurring in qi. Therefore, when an event is considered to occur in the state qi, the occurrence probability of the state qi + 1 is expressed as follows.
P (qi + 1 | qi) (Formula 1)
That is, P (qi + 1 | qi) is the probability of transition from state qi to qi + 1.

  FIG. 6 is a diagram for explaining the prediction importance and the result importance. Now, in the actual state qi, the victory probability is WP (qi). On the other hand, when a transition is made, there are three predicted states qi + 1 after the transition in the illustrated example, and the victory probability at the transition qi + 1 is WP (qi + 1). Three victory probabilities are obtained in the case of FIG. Of these three, only one transition has actually occurred, and the winning probability corresponding to the actual state qi + 1 as a result of the transition is the winning probability WP (qi + 1) after the occurrence of the event (the actual shown in FIG. 6). Win probability).

  FIGS. 7-1, 7-2, and 7-3 are charts showing examples of state transition probabilities in Japanese professional baseball created based on the database. As shown in the chart, for example, the probability of transition from no-runner before transition to no-out runner after transition is 0.0384. Specifically, this indicates that the probability of the first batter home run immediately after the start of inning is 0.0384. In addition, for example, the probability of becoming full from no-out 2, 3 km is 0.0455. In addition, for example, the probability of 3 out without adding 2 points is 0.6875.

  In the case of a baseball game, as is easy to understand with reference to FIGS. 7-1, 7-2, and 7-3, in general, there is only one state that cannot be changed from state qi to state qi + 1. Condition, and there can be more than one. For example, it is impossible to make a transition from 2out 3 塁 to 2out 1 塁. Therefore, the corresponding portions in FIGS. 7-1, 7-2, and 7-3 are blank. In addition, there are four states that transition from 1 out 3 塁, and the probability is obtained for each of them.

  In baseball, when a single hit event occurs when the output situation is 2 or 3, the state after the transition differs depending on whether the game is full or one point is added. In order to predict the next occurrence state in a certain state, it is necessary to statistically obtain the occurrence probability of the next transition state when a corresponding event occurs in a certain state and create a DB. In the case of baseball, various score book information is stored in a DB, and a list of transition probabilities as shown in FIGS. 7-1, 7-2, and 7-3 can be obtained.

Since sports are progressing based on rules, it can be assumed that there are a finite number of events that occur in an arbitrary state and the states that transition to the next by the event. Therefore, when there are one or more states qi + 1 that can be transitioned from a certain state qi, the sum of the states multiplied by the victory probability corresponding to the transitionable state is the predicted victory probability in the state of qi + 1. That is, the predicted victory probability in the state of qi + 1 is
ΣP (qi + 1 | qi) WP (qi + 1) (Formula 2)
It is. Here, the sum Σ is performed for a transitionable state. Further, the win probability in the state qi + 1 can be obtained from FIGS. 7-1, 7-2, and 7-3.

  The difference between the predicted victory probability in the state qi + 1 and the victory probability in the state qi indicates the predicted importance, which is the importance of an ongoing scene where it is unknown what happens next to the scene before the transition. Become. In other words, when expressed as an expression:

The predicted importance SI (qi) is
SI (qi) = abs {(ΣP (qi + 1 | qi) WP (qi + 1)) − WP (qi)} (Formula 3)
It is.

(1.6. Data abstraction)
In an actual game, there may be a case where data is not accumulated regarding an event that occurs. In that case, although not individual, stepwise abstraction of the occurrence probability of the event is performed by approximating the data by using more general data.

  FIG. 8 is a diagram for explaining the abstraction of the event occurrence probability. As described above, P (qi + 1 | qi) expresses the probability that a state qi + 1 capable of transitioning in the state qi will occur. For example, in the case of baseball, the probability of occurrence of an event such as a hit, HR, or strikeout in a specific situation can be obtained from the match score of the pitcher versus batter (FIG. 8). These competition results are recorded in an official DB that records baseball game data. However, new players may have little game data and may not have enough data to determine the probability of an event. In such a case, the event occurrence probability is obtained by abstracting in stages.

  For example, in FIG. 8, the probability of occurrence of an event in pitcher C versus batter 2 is not on the DB. Therefore, referring to the average probability of occurrence of events in pitcher C (targeting all batters) and the average probability of occurrence of events in batter 2 (targeting all pitchers) and taking their weighted averages, pitcher C vs. batter 2 is the occurrence probability of the event. If there is no data for pitcher C, the probability of another occurrence is abstracted, and the average occurrence probability of events in all pitchers and the average occurrence probability of events in batter 2 are taken and a weighted average is taken. In this way, P (qi + 1 | qi) can be obtained from the data abstraction shown in FIG. 8 and the data of the baseball DB.

(1.7. Definition of attribute importance)
The attribute importance is an importance based on the attribute value in each state. For example, when there are a plurality of states having the same importance level, a state in which the character is a popular player is generally more important. In this way, the importance of the attribute value of the character in the state qi such as the popularity of the character is AI (qi). AI is defined depending on the genre of sports. For example, in baseball, it is the degree of popularity of teams and players in that state.

(1.8. Progress scene importance)
Here, the progress scene importance will be described. Using the predicted importance (SI) and the attribute importance (AI), it is possible to realize a service that promptly notifies the user of highly important information during the game. The index that gives the importance in progress in the state qi is defined as the progress scene importance (PS = Progress Significance) as follows.

PS (qi) = αAI (qi) + βSI (qi) (Formula 4)
Here, α and β are weighting factors.
The value calculated according to the above definition is an index that increases in importance before the transition from the composition of the formula while it is progressing in real time.

  When a high PS state occurs, by notifying the user to that effect, the user can know the situation of an interesting game in real time and can enjoy the subsequent game development. This service is particularly effective for sports that take a certain amount of time when transitioning from one state to the next, such as baseball.

  Here, the PS may be calculated without the attribute importance AI. This is especially for calculating the importance of the progress scene excluding the attribute importance AI of popular baseball teams and popular players.

  FIG. 9 is a flowchart for explaining a scene determination procedure according to the first embodiment. The expression unit 1 of the scene determination device 10 receives the content corresponding to the scene qi of the game to be developed and expresses the state transition using metadata (step S101). The attribute determination unit 2 analyzes attribute importance from attribute information including popularity of teams and players in the scene qi (step S102). The prediction determination unit 3 determines the prediction importance degree for which the state before the transition is determined from the state after the state is predicted to transition (step S103). The progress determination unit 4 is a progression viewed from the post-progression state predicted after the progress while the game is in progress, based on the attribute importance determined by the attribute determination unit 2 and the prediction importance determined by the prediction determination unit 3. The advancement scene importance, which is the previous importance, is calculated, and the importance of the advancement scene is determined by performing threshold processing (step S104).

(1.9. Example of scene determination based on progress scene importance)
The progress scene importance (PS value) is calculated when the state before the transition occurs before the state ends. For example, in the case of baseball, it is calculated when the batter enters the bat. Table 2 shows the top five scenes sorted by PS value in the Giant vs. Chunichi game on July 5, 2003.

  FIG. 10 is a chart showing the matching rate of scenes extracted by the PS value by the scene determination device according to the first embodiment. In the system according to the embodiment, in the case of (PS value> = 0.15), the status information is promptly notified to the user by e-mail. In the case of Table 1, the information on the upper two states was delivered to the user as a bulletin. In our experience, the scene where PS> = 0.15 is about 0.6% of all scenes, and if 6 games are held by 12 teams a day, an average of 3 Mail was delivered.

  FIG. 11 is a diagram illustrating a questionnaire result regarding mail distribution of a baseball game received by a user of the information distribution system. It can be seen that nearly 50% of users want to receive information by email about 2 to 3 per game. Except for two of the bulletins of the results of the game win and loss and the distribution of various information before the game, the optimal number of distributions of scenes that are likely to be interesting is about 0 to 1 per game. In the distribution using the threshold values set above, there were an average of three distributions for six games. This is considered to have almost coincided with the number of distributions according to the user's most desired result as the above questionnaire result.

(1.10. Effect)
Thus, according to the scene determination apparatus according to the first embodiment, the expression unit 1 expresses the state transition, the attribute determination unit 2 obtains the attribute importance of the appearance team or person in the scene, and the prediction determination unit 3 determines the prediction importance. Since the progress determination unit 4 determines the importance of the scene in progress using the attribute importance by the attribute determination unit 1 and the prediction importance by the prediction determination unit 3, it is assumed that no change has occurred as a result. Since it is possible to determine an important scene that may cause a change, an important scene determination device that can accurately determine the importance of an ongoing scene.

  When an information distribution system including an information distribution apparatus incorporating the scene extraction apparatus according to the embodiment is actually operated, 70% of the users receive the breaking news mail determined and distributed based on the importance of the progress scene. The results also show that the churn rate of users who subscribe to mail delivery is low. Therefore, when applying the scene extraction apparatus according to the present embodiment, which is determined based on the ongoing importance, there is an effect of responding to the needs of viewers accurately and prospering the information distribution service. Recognize.

(2. Embodiment 2)
FIG. 12 is a functional block diagram of a scene determination apparatus according to Embodiment 2 of the present invention. The difference between the scene determination device 20 according to the second embodiment and the scene determination device 10 according to the first embodiment is that the progress determination unit 4 included in the scene determination device 10 according to the first embodiment is not provided, but instead the result. It is a point provided with the determination part 24 and the final determination part 25. FIG.

  The result determination unit 24 calculates and determines the result importance determined by before and after the state transition of the scene that appears as a result. The final determination unit 25 calculates the importance of the end scene based on the result that has been calculated by the result determination unit 24 at the time when the game is finished or until a certain time of the game. Determine the importance.

(2.1. Importance of results)
As described above, the result importance (RI) has a higher value in the state immediately before the transitioned state, for example, when the state after the transition is reverse, preemptive, or V goal.

  When there is a certain state qi and a transition is made to the next state qi + 1, the result importance (RI) of this state qi can be expressed by the following equation.

RI (qi) = abs (WP (qi + 1) −WP (qi)) (Formula 5)
Here, WP (qi) represents the victory probability of a certain state qi.

  The meaning of Equation 5 is that when there is a certain state qi and transition to the next state qi + 1, the absolute value of the difference in victory probability (WP) is obtained, and the result is the importance RI (qi) of the state qi. After the state transition, qi which is the state before the transition when the WP is greatly changed by the result actually generated becomes an important scene as a result.

  Here, it is also possible to adopt a configuration in which the important scene is determined only by the result importance RI (qi) except for the attribute importance AI. Moreover, it is good also as a structure which adds an attribute importance to a result importance and determines and extracts an important scene.

(2.2. Last scene importance)
Here, the final scene importance will be described. The final scene importance is the degree of overall importance in scenes that have been completed up to a certain point in time. In other words, the final scene referred to here is a scene up to that point, and does not merely indicate that the game has ended.

  Using the result importance (RI), predicted importance (SI (qi)), and attribute importance (AI), the final scene importance (FS (qi)) in the state qi is defined as follows.

FS (qi) = γ * AI (qi) + δ * SI (qi) + ε * RI (qi) (Formula 6)
Here, γ, δ, and ε are weighting factors.

  The large value of this FS (qi) is the importance that ended by a certain point, combining the importance by individuals and teams, “scenes that did not come out as a result”, and “scenes that were good as a result” It becomes a high scene. When providing a digest or the like, it is only necessary to sort by the final scene importance and select a scene with a higher rank.

  Here, the FS may be calculated excluding the attribute importance AI. This is especially to calculate the importance of the final scene excluding the attribute importance AI of popular teams and popular players.

  FIG. 13 is a flowchart for explaining a scene determination procedure according to the second embodiment. The image processing procedure according to the second embodiment is the same as that up to step S103 shown in FIG.

  The result determination unit 24 of the scene determination device 20 calculates the result importance, which is the importance before the transition as seen from the state after the state transition (step S204). The final determination unit 25 determines the importance of the team and the appearing player from the attribute importance determined by the attribute determination unit 2, the prediction importance determined by the prediction determination unit 3, and the result importance determined by the result determination unit 24. Determining the importance of the scene that increases tension during the game and the importance of the scene that has become more important as a result of progress by using threshold processing. (Step S205).

(2.3. First scene determination example 1)
In the example, the probability of occurrence of an event in a scene where a specific pitcher and batter appear is obtained from the database (DB) of the Japan Professional Baseball Official Record (NPB). The occurrence probability of the state after transition when a certain event occurs in a certain state was statistically calculated for the past 100 games based on the NPB DB.

  In addition, regarding attribute importance (AI), we take a questionnaire for 807 users to select one favorite player (no answer is acceptable) and use that player as a favorite for each professional baseball player The ratio (popularity) of the players was investigated, and the popularity of the players who appeared as batters in each state was defined as the AI of that state. In the case of the pitcher, since it appears in all the states that are climbing, if the popularity of the pitcher is reflected in the AI, there will be many states with high AI, so in this implementation The AI of each state does not reflect the popularity of the pitcher.

  FIG. 14 is a chart showing the average precision when the important scenes are extracted based on the FS. A TV news program was used as the correct answer set. The average precision is 0.7 or more, and important scenes can be extracted with high accuracy.

(2.4. Example 2 of final scene determination)
FIG. 15 is a chart showing the relevance ratio at which the scene determined by the scene determination apparatus according to the second embodiment matches the scene actually selected by the television program. A scene in which the scene determined by the scene determination device matches the scene selected by the sports program of the television station is marked with a mark.

Here, the attribute importance AI is not shown, but since the attribute importance is already included in the RI and SI and calculated, the final importance FS (qi) uses the following equation.
FS (qi) = α * SI (qi) + β * RI (qi) (Formula 7)

  Here, α and β are weighting factors, and determine whether to place importance on the result importance or the prediction importance, respectively. According to the case of this table, scene selection with predictive importance and outcome importance shows a much higher precision than selecting scenes from each single importance. I understand that.

(2.5. Effect)
Thus, according to the scene determination apparatus according to the second embodiment, the expression unit 1 expresses the state transition, the attribute determination unit 2 obtains the attribute importance of the appearance team or person in the scene, and the prediction determination unit 3 determines the prediction importance. The result determination unit 24 calculates the result importance, which is the importance of the important scene generated as a result of the occurrence of the change, and the final determination unit 25 determines the attribute importance and prediction determination unit 3 by the attribute determination unit 1. By using the predicted importance by the result and the result importance by the result judging unit 24, an important scene that may cause a change even if the change is not caused as a result, and an important change as a result. Since it is possible to make a comprehensive determination according to the scene that has been brought, it becomes an important scene determination device that can accurately determine the importance of the scene up to the time of determination.

(3. Information distribution apparatus according to Embodiment 3)
The information distribution apparatus according to the third embodiment is an information distribution apparatus including the scene determination apparatus according to the first embodiment. One example is the information distribution apparatus 100 shown in FIG. The information distribution device 100 includes a content generation device 50, a scene determination device 10, and a distribution device 60.

  Each function of the content generation device 50 and the distribution device 60 included in the information distribution device 100 is as described in the description of the first embodiment. The configuration and functions of the scene determination device 10 that is a main part of the information distribution device 100 according to the third embodiment are also as described in the description of the scene determination device according to the first embodiment, and thus the description thereof is omitted.

  Here, the scene determination apparatus 20 according to the second embodiment can be applied as the scene determination apparatus included in the information distribution apparatus 100. Since the scene determination device 20 is as described in the description of the second embodiment, the description thereof is omitted.

  With the above configuration, in the information distribution apparatus according to the third embodiment, the expression unit 1 expresses the state transition, the attribute determination unit 2 calculates the attribute importance of the appearance team or person in the scene, and the prediction determination unit 3 predicts the importance. Since the progress determination unit 4 determines the importance of the scene in progress using the attribute importance by the attribute determination unit 1 and the prediction importance by the prediction determination unit 3, it does not cause a change as a result. Even if an important scene that may cause a change is determined, the importance of the scene in progress can be accurately determined, and the determined content can be distributed.

  Further, the result determination unit 24 obtains the result importance that is the importance of the important scene that has occurred as a result of the change, and the final determination unit 25 determines the attribute importance and prediction determination by the attribute determination unit 1. Using the prediction importance by the unit 3 and the result importance by the result determination unit 24, important scenes that may cause a change even if they do not change as a result, and are important as a result It becomes an information distribution apparatus capable of accurately determining an important scene comprehensively according to a scene that brings about a change and distributing information on the scene.

(4. Information distribution system according to Embodiment 4)
The information distribution system according to the fourth embodiment is an information distribution system including the information distribution apparatus according to the third embodiment. One example is the information distribution system shown in FIG.

  The information distribution system includes an information distribution device 100 and information terminals 71, 72, and 73 connected via a network 5. Since the information distribution apparatus 100 is as described in the third embodiment, the description thereof is omitted.

Here, the information terminals 71 to 73 can be mobile phones.
The information terminals 71 to 73 can be personal digital assistants (PDAs).

  FIG. 16 is a chart showing an example of a simulation in which baseball scene information selected in the information distribution system including the information distribution apparatus according to the third embodiment is distributed. In this table, the information distribution apparatus includes the scene determination apparatus according to the first embodiment. When the scene determination apparatus has a predicted importance level exceeding 1.80, it is sent to the subscriber's mobile phone by e-mail. Information distribution. Here, the game IDs 3 and 5 are scenes where the SI value is much higher than RI. As a result, the score was not included, but the possibility was quite high. In the past, such a scene had to be manually monitored by monitoring the progress of the game and manually selected. However, in the information distribution system according to the fourth embodiment, since a scene determination device is provided, It is possible to automatically determine the importance of prediction and transmit an article of the scene.

  In this way, according to the information distribution system according to the fourth embodiment, even if a change does not occur as a result, an important scene that may cause a change is determined, and the importance of the ongoing scene is accurately determined. Since the content determined and determined can be distributed to the information terminal, the importance of the ongoing scene can be accurately determined and distributed.

  Moreover, even if it does not cause a change as a result, important scenes that may cause a change and a scene that resulted in an important change are judged comprehensively and accurately. An information distribution system capable of accurately distributing information that can distribute scene information to information terminals.

(5. Hardware configuration of scene determination apparatus according to embodiment)
FIG. 17 is a hardware configuration diagram of the scene determination device according to the embodiment. The scene determination apparatus according to the present embodiment includes a control device such as a CPU 1001, a storage device such as a ROM (Read Only Memory) 1002 and a RAM 1003, an external storage device such as an HDD 1004 and a CD drive device, and an operation / display unit. An input interface 1006, a keyboard 1012, a scanner 1013, a graphic processing device 1005 and a monitor 1011, a communication device 1008 and a bus 1007 are provided, and a hardware configuration using a normal computer is provided. The ROM 1002 stores a scene determination program for causing the computer of the scene determination apparatus to execute each step (step) and each unit described above.

  The scene determination program is a file in an installable or executable format and is recorded on a computer-readable recording medium such as a CD-ROM, a flexible disk (FD), a CD-R, or a DVD (Digital Versatile Disk). May be provided. In this case, the CPU 1001 reads out the scene determination program from the storage medium and loads it onto the main storage device, thereby causing the scene determination device to realize each step (step) and each unit described above.

  The scene determination program may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Alternatively, the scene determination program may be provided or distributed via a network such as the Internet.

  As described above, the scene determination apparatus, the scene determination method, the program for causing the computer to execute the method, the information distribution apparatus, and the information distribution system according to the present invention are useful for multimedia content selection technology, and in particular, scene change. This is suitable for a technique for determining the importance of content indicating “”.

It is a functional block diagram of the scene determination apparatus by Embodiment 1 of this invention. It is a figure explaining an example which applied the scene determination apparatus by Embodiment 1 to the information delivery system. It is a figure explaining the state transition of the game in a baseball game. It is another figure explaining the state transition of the game in a baseball game. It is a graph which shows an example of the victory probability in a baseball game. It is a figure explaining prediction importance and result importance. It is a graph which shows an example of the state transition probability in Nippon Professional Baseball created based on the database. It is a graph which shows an example of the state transition probability in Nippon Professional Baseball created based on the database. It is a graph which shows an example of the state transition probability in Nippon Professional Baseball created based on the database. It is a figure explaining the abstraction of event occurrence probability. 5 is a flowchart for explaining a scene determination procedure according to the first embodiment. It is a chart which shows the adaptation rate of the scene which the scene determination apparatus by Embodiment 1 extracted by PS value. It is a figure explaining the questionnaire result about the mail delivery of the baseball game which the user of an information delivery system receives. It is a functional block diagram of the scene determination apparatus by Embodiment 2 of this invention. 10 is a flowchart illustrating a scene determination procedure according to the second embodiment. It is a graph which shows the average precision at the time of extracting an important scene based on FS. It is a graph which shows the relevance rate which the scene determined by the scene determination apparatus by Embodiment 2 matched with the scene actually selected by the television program. 10 is a chart showing an example of a simulation in which baseball scene information selected in an information distribution system including an information distribution device according to Embodiment 3 is distributed. It is a hardware block diagram of the scene determination apparatus by embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10, 20 Scene determination apparatus 100 Information distribution apparatus 50 Content generation apparatus 60 Distribution apparatus 1 Representation part 2 Attribute determination part 3 Prediction determination part 4 Progression determination part 24 Result determination part 25 Final determination part

Claims (7)

An expression means that uses video and text and expresses using the metadata as a state transition of a change in the content of the content to which the metadata is added,
A state transition expressed by the expression means, analyzing a prediction importance that is an importance of a scene before the transition based on a state predicted to occur after the transition, and a determination means for determining an important scene , and
The expression means, when the content of the content is a sport game, expresses the state transition with a feature amount including a victory probability that is a probability that a home team will win at the time when the sport game is in progress.
The determination unit obtains a total value of expected values of the feature values in a predicted state by predicting all states that occur in the state transition, and the total value of the expected values and the feature value before the transition Calculating and analyzing the predicted importance based on the difference between
Scene determining apparatus according to claim. An expression step that includes video and text, and uses the metadata to represent changes in the content of the content to which metadata is added as a state transition;
Analyzing the prediction importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition expressed in the expression step, and determining the important scene ; and
In the expression step, when the content is a sports game, the state transition is expressed by a feature amount including a victory probability that is a probability that a home team will win at the time when the sports game is in progress.
The determination step obtains a total value of expected values of the feature values in a state predicted by predicting all states occurring in the state transition, and calculates the total value of the expected values and the feature value before the transition Calculating and analyzing the predicted importance based on the difference between
A scene determination method characterized by A program causing a computer to execute the method according to claim 2 . Content generating means for generating content including video and text and having metadata added thereto;
An expression means for expressing the change of the content generated by the content generation means as a state transition using the metadata;
A determination means for analyzing the predicted importance, which is the importance of the scene before the transition based on the state predicted to occur after the transition by the state transition expressed by the expression means, and determining the important scene;
Distribution means for distributing content determined to be an important scene by the determination means ,
The expression means, when the content of the content is a sport game, expresses the state transition with a feature amount including a victory probability that is a probability that a home team will win at the time when the sport game is in progress.
The determination unit obtains a total value of expected values of the feature values in a predicted state by predicting all states that occur in the state transition, and the total value of the expected values and the feature value before the transition Calculating and analyzing the predicted importance based on the difference between
An information distribution apparatus characterized by the above. A content generation device that generates content with metadata including video and text, each connected via a network;
An expression means for receiving content generated by the content generation device via the network, and expressing the received content change as a state transition using the metadata, and
A scene determination device having a determination unit that analyzes a predicted importance level of a scene before the transition based on a state predicted to occur after the transition by the state transition expressed by the expression unit and determines an important scene; ,
A distribution device for distributing content determined to be an important scene by the scene determination device via the network;
An information terminal for receiving content distributed by the distribution device via the network;
Equipped with a,
The expression means, when the content of the content is a sport game, expresses the state transition with a feature amount including a victory probability that is a probability that a home team will win at the time when the sport game is in progress.
The determination unit obtains a total value of expected values of the feature values in a predicted state by predicting all states that occur in the state transition, and the total value of the expected values and the feature value before the transition Calculating and analyzing the predicted importance based on the difference between
An information distribution system characterized by The information distribution system according to claim 5 , wherein the information terminal is a mobile phone. The information distribution system according to claim 5 , wherein the information terminal is a portable information terminal.

Images