WO2008004576A1 - Dispositif de recherche/évaluation de contenu - Google Patents
Dispositif de recherche/évaluation de contenu Download PDFInfo
- Publication number
- WO2008004576A1 WO2008004576A1 PCT/JP2007/063362 JP2007063362W WO2008004576A1 WO 2008004576 A1 WO2008004576 A1 WO 2008004576A1 JP 2007063362 W JP2007063362 W JP 2007063362W WO 2008004576 A1 WO2008004576 A1 WO 2008004576A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- content
- distribution
- score
- search
- evaluation
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—Filtering based on additional data, e.g. user or group profiles
- G06F16/437—Administration of user profiles, e.g. generation, initialisation, adaptation, distribution
Definitions
- Content evaluation device content search device, content evaluation method, content search method, and first and second computer programs
- the present invention relates to a content evaluation apparatus and method, and a computer for evaluating a degree of conformity with respect to one search word related to one content among a plurality of contents stored in a recording device such as a music server.
- a first computer program that functions as a content evaluation apparatus, a content search apparatus and method for searching for content using the content evaluation apparatus, and a computer as such a content search apparatus In the technical field of the second computer program.
- This type of content evaluation device and a music selection device for the purpose of presenting music that matches the user's sensibility as a content search device for searching content using the content evaluation device are proposed.
- music features e.g. chord change rate '' beats
- preset personal characteristics e.g. age and gender
- sensitivity words e.g. ⁇ bright '' and ⁇ sad ''
- Patent Document 1 Japanese Patent Application Laid-Open No. 2005-209276
- the sensitivity matching value is calculated based on the distance from the average value of each feature amount for the selected sensitivity word (for example, “bright”).
- This calculation method uses the feature value that is a sample of the average value. This is based on the assumption that the distribution is a normal distribution. Therefore, if the above assumptions are not valid, such as when there are two or more peaks, when there is no clear peak, or when one peak is not symmetrical, the reliability of the evaluation is reduced. Can do.
- the present invention has been made in view of the above-described problems, for example, and is a content evaluation apparatus and method capable of suitably evaluating the content of music and the like, and a computer functioning as such a content evaluation apparatus.
- the problem is to provide computer programs that can be used. It is another object of the present invention to provide a content search apparatus and method that can suitably search for content using a content evaluation apparatus, and a computer program that causes a computer to function as such a content search apparatus.
- the content evaluation device evaluates the degree of conformity with respect to one search word related to one content among a plurality of content stored in the recording device.
- the feature quantity for quantitatively characterizing the content is extracted for each of the plurality of contents, and the one search word out of the plurality of contents is positively or negatively evaluated.
- Interpolating means for interpolating the distribution of the feature amount related to the content having the history, creation means for creating a score curve in which a score is associated with the feature amount based on the interpolated distribution, and the creation Calculating means for calculating the score based on the score curve obtained.
- the content evaluation device of the present invention for example, for evaluating the degree of conformity with respect to one search word related to one content among a plurality of content stored in a recording device such as a music server.
- Score power Calculated as follows.
- Content here refers to content and data that can be stored in a recording device and can be evaluated by the user, such as music, video, or a homepage.
- a “search term” is a term that indicates what aspect the content is evaluated for, in other words, an axis of evaluation, typically “bright”, “quiet” or “good”. Demonstrate the user's subjective sensibility. Such a search term is preliminarily determined by the user at the time of evaluation or Automatically selected from multiple candidates.
- the “score for evaluating the degree of goodness of fit” is an index that quantitatively evaluates how well a single content matches a search term in at least two stages. In other words, there are two stages: “positive evaluation of“ conforming ”)” and “UNFIT (ie,“ conforming! /, Isn't, negative evaluation) ”.
- the distribution of the feature amount related to the content having a history evaluated positively or negatively is interpolated by an interpolation unit having an arithmetic unit or the like.
- “History positively or negatively evaluated” as used herein refers to a search term that has been matched in the past (ie, “FIT”), or that has not been matched positively (ie, “FIT”).
- “UNFIT”) is a history of negative evaluation. The purpose is to estimate the score of one content to be evaluated from the distribution pattern of the feature amount of the content having a history evaluated positively or negatively.
- Interpolation means that the distribution of discrete feature values is made continuous by an interpolation technique such as a mixed Gaussian model.
- a score curve in which a score is associated with a feature amount is created by a creation unit having an arithmetic unit or the like.
- the “score curve” here is simply the force that is the interpolated distribution itself, as described below, with some further corrections to the distribution! /.
- the calculation means having an arithmetic unit or the like
- the above-mentioned score is calculated. Specifically, for example, by extracting the feature value for one search word related to one content to be evaluated, and reading the score associated with the extracted feature value on the score curve, the score is obtained. Is calculated. If there are a plurality of types of feature values, a score for each feature value may be obtained individually, and the score may be calculated by summing up the scores. At this time, change the weighting for each score for each feature amount individually, It is good also.
- the interpolation means interpolates a first distribution of feature quantities related to content having a history that has been positively evaluated among the plurality of contents. And interpolating a second distribution of feature quantities related to content having a negatively evaluated history among the plurality of contents, and the creating means interpolates from the interpolated first distribution.
- the at least one score curve is created by subtracting the second distribution.
- the first distribution of the feature amount related to the content having the history that has been positively evaluated among the plurality of contents is interpolated by the interpolation unit, and the negative evaluation is performed among the plurality of contents.
- the second distribution of the extracted feature quantity related to the content having the recorded history is interpolated.
- at least one score curve is created by subtracting the interpolated second distribution from the interpolated first distribution by the creating means, and based on this score curve, as described above. A score is calculated. In this way, not only based on “positively rated history” but also on “negatively rated history”! Since the score curve is created, the score curve reflects the user's sensibility more strongly and the evaluation accuracy is improved.
- the interpolation means further interpolates a third distribution of the feature amounts related to the plurality of contents, and based on the interpolated third distribution.
- Correction means for correcting the score curve so as to reduce the contribution of the third distribution to the shape of the score curve.
- the third distribution of the feature amounts related to the plurality of contents is further interpolated by the interpolation means.
- the “plural content” here is typically all content.
- the correction means corrects the score curve based on the interpolated third distribution so as to reduce the contribution of the third distribution to the shape of the score curve.
- the first distribution described above does not represent a distribution specific to content with a purely positive history, but is inherently more or less affected by the distribution of all content (ie, the third distribution). It is also the power that is being received.
- the second distribution As a specific example of the correction, the first and second distributions may be divided by the third distribution before the score curve is created, or the created score curve may be divided by the third distribution. The power is not limited to this. In any case, since the correction is made in this way and the contribution of the third distribution to the shape of the score curve is reduced, the evaluation accuracy is further improved.
- the interpolation means further interpolates a third distribution of the feature quantities related to the plurality of contents, and the third means for the shape of the score curve.
- Correction means for correcting the interpolated first distribution and correcting the interpolated second distribution based on the interpolated third distribution so as to reduce the contribution due to the distribution of Prepare.
- the score curve is corrected.
- the contribution of the third distribution to the shape is reduced, thus further improving the evaluation accuracy.
- a content search device is based on the content evaluation device according to any one of claims 1 to 4 and the calculated score.
- a score is calculated for each of a plurality of contents by the content evaluation device according to any one of claims 1 to 4.
- a search unit having an arithmetic unit or the like searches for a content that matches a central search word of a plurality of contents.
- a score for one search word is given, and several contents having a relatively high score are searched as content that matches one search word.
- the content searched in this way is output with a display or the like. It is output to the user by the force means.
- the content search device of the present invention it is possible to search for content suitably based on a score calculated with high accuracy, and the search accuracy is improved, which is very advantageous in practice.
- the search means searches for a content exceeding the calculated score power predetermined score threshold among the plurality of contents.
- the search means searches for a content exceeding the calculated score force predetermined score threshold among the plurality of contents.
- the “predetermined score threshold” is a value calculated in advance by experiment or simulation as a lower limit value of the score at which the content can be considered to be suitable for one search term. This predetermined score threshold may be fixed or may be changed later by the user. If such a restriction is set, even if the content is huge, only the necessary amount is output, which is effective in practice.
- the content search apparatus further includes an update unit that updates the history evaluated for the searched content according to the evaluation by the user.
- the history evaluated for the searched content is updated by the updating unit according to the evaluation by the user. For example, if the content power searched for the search term “bright” is perceived as “bright” by the user, the user positively evaluates it, that is, evaluates it as “FIT”. The history is updated according to. On the contrary, if the user does not feel “bright”, the user evaluates negatively, that is, evaluates “U NFIT”, and the history is updated according to the strong evaluation. Typically, the oldest historical power is also updated. “User evaluation” is typically done in two stages, “FIT” or “UNFIT”, or more.
- Such evaluation may be performed, for example, by the user manually selecting options on the display, or may be performed automatically by analyzing the brain waves of the user who is viewing the content. In this way, when the history is updated, a score curve is created based on the subjective judgment of the new user as much as possible. Since the content is searched, the search accuracy is improved.
- the content evaluation method of the present invention is a content evaluation method for evaluating the degree of conformity with respect to one search word related to one content among a plurality of contents stored in a recording device, An extraction step for extracting the feature quantity quantitatively characterizing the content for each of the plurality of contents, and a history of positive or negative evaluation for the one search word among the plurality of contents.
- An interpolation process for interpolating a distribution of feature quantities related to the content to be created, a creation process for creating a score curve in which a score is associated with the feature quantities based on the interpolated distribution, and the created A calculation step of calculating the score based on the score curve.
- the content evaluation method of the present invention can also adopt various aspects similar to the various aspects of the content evaluation apparatus of the present invention described above.
- the content search method of the present invention uses the content evaluation method according to claim 8 to calculate the score for each of the plurality of contents, and based on the calculated score. And a search step for searching for content that matches the central search term of the plurality of contents, and an output step for outputting the searched content to the user.
- the content search method of the present invention can also adopt various aspects similar to the various aspects of the content search apparatus of the present invention described above.
- a first computer program according to the present invention provides a composition.
- the data is made to function as the content evaluation device according to any one of claims 1 to 4.
- the first computer program product in the computer-readable medium can be executed by a computer provided in the above-described content evaluation apparatus (including various forms thereof) of the present invention.
- a program command is clearly embodied, and the computer functions as at least a part of the content evaluation device (specifically, for example, at least one of extraction means, interpolation means, creation means, and calculation means)
- the first computer program product of the present invention is stored in a recording medium such as a ROM, CD-ROM, DVD-ROM, or hard disk that stores the first computer program product. If the product is read into a computer or the first computer program product, which is a transmission wave, for example, is downloaded to the computer via communication means, the above-described content evaluation apparatus of the present invention can be implemented relatively easily. It becomes possible. More specifically, the first computer program product may also be configured with a computer readable code (or computer readable instruction) force that functions as the content evaluation apparatus of the present invention described above.
- a second computer program of the present invention causes a computer to function as the content search device according to any one of claims 5 to 7.
- a second computer program product in a computer-readable medium solves the above problem.
- a program instruction executable by a computer provided in the above-described content search device of the present invention (including various forms thereof) is clearly embodied, and the computer is used as the content search device. At least a part (specifically, for example, at least one of search means and output means).
- the second computer program product of the present invention is stored in a recording medium such as a ROM, CD-ROM, DVD-ROM, or hard disk that stores the second computer program product. If the product is read into a computer or the second computer program product, which is a transmission wave, for example, is downloaded to the computer via communication means, the above-described content search device of the present invention can be implemented relatively easily. It becomes possible. More specifically, the second computer program product may also be configured with a computer readable code (or computer readable instruction) force that functions as the content search device of the present invention described above.
- the content evaluation apparatus includes the extraction unit, the interpolation unit, the creation unit, and the calculation unit.
- the extraction step and the interpolation step Since the creation step and the calculation step are provided, it is possible to suitably evaluate the degree of conformity with respect to one search term related to one content.
- the computer program of the present invention since the computer functions as an extraction unit, an interpolation unit, a creation unit, and a calculation unit, the above-described content evaluation apparatus of the present invention can be constructed relatively easily.
- a search means and output means are provided in addition to the content evaluation device.
- a search is performed in addition to the content evaluation method. Since it includes a process and an output process, it is possible to search for content appropriately based on the evaluation results.
- FIG. 1 is a block diagram conceptually showing the basic structure of a content search apparatus equipped with a content evaluation apparatus according to an embodiment of the present invention.
- FIG. 2 is a flowchart showing an operation process of the content search device according to the embodiment.
- FIG. 3 is a list of music feature quantities extracted for each music piece according to an embodiment.
- FIG. 4 is a list of evaluation histories subjectively evaluated for a search word according to an example.
- FIG. 5 is a characteristic diagram showing an evaluation history distribution interpolation graph when the feature amount of the music is “rhythm speed” according to the embodiment.
- Fig. 6 is a characteristic diagram showing an all-music distribution interpolation graph in the case where the feature amount of the music is "Rhythm speed" according to the embodiment.
- FIG. 7 is a characteristic diagram showing a process of creating an FIT evaluation history distribution interpolation correction graph when the feature amount of the music is “rhythm speed” according to the embodiment.
- FIG. 8 is a characteristic diagram showing a process of creating a UNFIT evaluation history distribution interpolation correction graph when the feature amount of the music is “rhythm speed” according to the embodiment.
- FIG. 9 is a characteristic diagram showing a process of creating a score curve when the feature amount of the music is “rhythm speed” according to the embodiment.
- FIG. 1 is a block diagram conceptually showing the basic structure of the content search apparatus equipped with the content evaluation apparatus according to the embodiment of the present invention.
- the content search device 1 is a device that searches for music as an example of the content to be searched, and includes a music input unit 10, a feature amount extraction unit 11, Collection storage unit 12, evaluation input unit 20, evaluation history storage unit 21, distribution interpolation unit 22, evaluation history correction unit 23, score curve creation unit 24, score calculation unit 25, search word selection unit 30, search result output unit With 40.
- the music input unit 10 to the score calculation unit 25 also function as a “content evaluation device” according to the present invention. Details of each part will be described below.
- the music input unit 10 is a disc player that plays a disc such as a CD or a streaming interface that accepts streaming distribution of music data, and waveform data of the music that is an example of the "content" according to the present invention. And the music identifier.
- the waveform data of the music is input as, for example, PCM data or MP3 data via media such as CD or via the Internet.
- the feature quantity extraction unit 11 is an example of the "extraction means" according to the present invention, and includes an arithmetic circuit such as a CPU. Rhythm speed, average sound level, chord change speed, frequency spectrum center of gravity, etc. are quantitatively analyzed and extracted (see Fig. 3).
- the feature quantity storage unit 12 is configured to include a storage device such as a hard disk, and stores the feature quantities of the extracted music pieces in association with each music piece in a manner that allows identification. . Further, when updating the evaluation history, when interpolating each feature value distribution, or when calculating a score, the feature value of the stored music is read out.
- the evaluation input unit 20 is an example of the "update means" according to the present invention, and includes an input device such as a touch panel, for example. Etc.) (ie, “subjective information”) and the identifiers of the music piece M and the search word S can be input to update the evaluation history. For example, if the user feels positively that the music M they watched is “bright”, the evaluation “FIT” Conversely, if you feel negative, “not bright”, you can enter an evaluation of “UNFIT” and update the evaluation history.
- This evaluation stage may be able to input more detailed evaluations of three or more levels, not limited to two levels such as “FIT” or “UNFIT”. The default value should be adopted until such evaluation is input by the user.
- the evaluation does not necessarily have to be manually input.
- the evaluation input unit 20 capable of quantifying the sensitivity by analyzing the spectral frequency of the user's brain waves and outputting it as a sensitivity value is the music M It may be automatically evaluated on the basis of the sensitivity when viewing.
- the human brain responds to a given stimulus, and the electroencephalogram changes accordingly.
- the brain has a functional localization force S, and it uses the fact that a characteristic potential distribution is formed by the part that is involved and the part that is not involved, depending on the state of sensitivity such as anger, joy, and sadness. Is.
- the evaluation history storage unit 21 is configured to include a storage device such as a hard disk, and reads the feature amount of the input music M and adds it to the evaluation history of the search word S.
- the search word S is associated with the feature amount of the music M.
- the evaluation history is classified into FIT history and UNFIT history, and is assigned to either history based on the input evaluation results.
- the distribution interpolation unit 22 includes a CPU, a memory, and the like, and interpolates various discrete graphs based on an interpolation technique such as a mixed Gaussian model. Specifically, it reads the evaluation history, creates a distribution graph of arbitrary search terms, subjectivity, and feature quantities (also called “evaluation history distribution graph”, see Fig. 5), and interpolates this graph (“ Also called “evaluation history distribution interpolation graph” (see Figure 5). Also, a distribution graph of arbitrary feature values of all stored music (also called “all music distribution graph”, see FIG. 6) is created, and this is also interpolated (also called “all music distribution interpolation graph”). Create Figure 6).
- an interpolation technique such as a mixed Gaussian model. Specifically, it reads the evaluation history, creates a distribution graph of arbitrary search terms, subjectivity, and feature quantities (also called “evaluation history distribution graph”, see Fig. 5), and interpolates this graph (“ Also called “evaluation history distribution interpolation graph”
- the evaluation history correction unit 23 is an example of the “correction unit” according to the present invention, and includes a CPU, a memory, etc., and the evaluation history distribution interpolation graph is corrected with an all-music distribution interpolation graph.
- a distribution interpolation correction graph is created for each evaluation stage. If there are two stages of evaluation, such as “FIT” or “UNFIT”, a graph that is corrected with the all-music distribution distribution graph is created for each FIT and UNFIT evaluation history distribution complement graph ( These graphs are called “FIT evaluation history distribution interpolation correction graph” and “UNFIT evaluation history distribution interpolation correction graph”. Also say. (See Figure 7 and Figure 8.)
- the score curve creating unit 24 is an example of the “creating unit” according to the present invention, and includes a CPU, a memory, and the like, and is based on the evaluation history distribution interpolation graph or the evaluation history distribution interpolation correction graph. Thus, a score curve necessary for calculating the score is generated (see FIG. 9).
- the score curve may be the FIT evaluation history distribution interpolation graph or the FIT evaluation history distribution interpolation correction graph itself, or the UNFIT evaluation history distribution interpolation graph or the UNFIT evaluation history distribution interpolation correction graph.
- the FIT evaluation history distribution interpolation graph force is also a graph obtained by subtracting the UNFIT evaluation history distribution interpolation graph, this graph is further corrected by the all-music distribution interpolation graph, or FIT evaluation history distribution interpolation correction graph force UNFIT as described later.
- a score curve is created based on at least the interpolated feature distribution graph.
- the score calculation unit 25 is an example of the “calculation unit” according to the present invention, and includes a CPU, a memory, and the like.
- the score calculation unit 25 reads the feature amount of each content and associates it with the score curve. The obtained value is calculated as a score. That is, content is evaluated as a content evaluation device.
- a score is associated with an arbitrary feature amount on the score curve. Therefore, any value of the feature amount of the content to be evaluated can be associated with the feature amount.
- the score can be read on the score curve. Therefore, even if the feature distribution graph (first distribution) is not a normal distribution, the score accuracy is higher than when the score is calculated as the distance from the average value without interpolation. It is very advantageous in practice.
- the search word selection unit 30 is an example of the "search means" according to the present invention, and includes an input device such as a touch panel.
- the search word selection unit 30 specifies a user from a plurality of search word candidates displayed on the display.
- the search word S can be selected.
- the search result output unit 40 is an example of "output means" according to the present invention, and includes a display or the like, and uses a song corresponding to the selected search word S (that is, FIT) as a search result. It can be displayed to the user. When the user determines a favorite song from the search results, the song is played. As described above with reference to FIG. 1, the content search device 1 equipped with the content evaluation device according to the present embodiment is configured. Therefore, the score calculated with high accuracy by the content evaluation device can be obtained. Based on the content stored in the recording device, the content can be suitably searched.
- FIG. 2 is a flowchart illustrating the operation process of the content search apparatus according to the embodiment.
- the music to be searched is input as follows, and the feature value is extracted and stored.
- the waveform data of the music to be searched and the identifier of the music to be searched in the search device are input via the music input unit 10 in advance at the time of factory shipment or afterwards by the user.
- Step S10 This waveform data is quantitatively analyzed by the feature quantity extraction unit 11, and the feature quantity of the music (for example, the speed of the rhythm, the average value of the sound level, the speed of the chord change, the center of frequency spectrum, etc.) is extracted.
- the feature quantity of the music for example, the speed of the rhythm, the average value of the sound level, the speed of the chord change, the center of frequency spectrum, etc.
- the extracted feature values are standardized by, for example, a reference average and standard deviation, and stored in the feature value storage unit 12 in association with the music identifier as shown in FIG. 3 (step S).
- FIG. 3 is a list of feature amounts of the music extracted for each music according to the embodiment.
- the evaluation of the music is performed as follows at the time of factory shipment or afterwards by the user who has watched the music searched based on the search word. Specifically, first, as evaluation (FIT or UNFIT) for the music M, subjective information indicating whether the image of the music M based on the user's subjectivity is suitable for the search word S is input to the evaluation input unit 20. (Step S20). In addition, the feature amount of the song M to be evaluated is read from the feature amount storage unit 12 based on the identifier of the song M (step S121). This feature amount is stored in the evaluation history storage unit 21 as an evaluation history as shown in FIG. 4 (a) together with the subjective information of the music piece M related to the search word S (step S21).
- FIT or UNFIT subjective information indicating whether the image of the music M based on the user's subjectivity is suitable for the search word S is input to the evaluation input unit 20.
- the feature amount of the song M to be evaluated is read from the feature amount storage unit 12 based on the identifier of the song M (step
- FIG. 4 is a list of evaluation histories subjectively evaluated for the search terms according to the embodiment.
- the evaluation history stores the feature quantities of the songs that have been evaluated (FIT or U NFIT) for each search term, with a management number.
- the feature number for each song that has been evaluated as “FIT” can be assigned a management number such as management numbers 0, 1, 2, 10, 11,. Will be saved.
- the evaluation history storage unit 21 may store an identifier of a music piece having the feature amount as an evaluation history instead of the feature amount. In other words, the evaluation history shown in FIG.
- the music feature quantity shown in FIG. 3 may be associated as a so-called relational database using the music identifier as a key.
- the feature quantity can be indirectly read out from the music feature quantity storage section 12 using the music identifier without being directly read out from the evaluation history storage section 21. Therefore, it is possible to save the capacity of the evaluation history storage unit 21 while maintaining that the feature amount of the music necessary for creating the evaluation history distribution graph can be read.
- the music can be searched as follows. Specifically, first, the search word “bright” is selected by the user via the search word selection unit 30 (step S30).
- the distribution interpolation unit 22 interpolates the distribution of the evaluated music (evaluation history distribution graph) with a mixed Gaussian model etc. with respect to the feature quantities of the music recorded in the evaluation history of the selected search term “bright”. Then, an evaluation history distribution interpolation graph shown in FIG. 5 is obtained (step S22).
- FIG. 5 is a characteristic diagram showing an evaluation history distribution interpolation graph when the feature amount of the music is “rhythm speed” according to the embodiment.
- the horizontal axis shows the “rhythm speed” as an example of the feature quantity
- the vertical axis shows the “sample number” of the music before the interpolation
- the “probability” that the sample exists after the interpolation Respectively.
- Such an evaluation history distribution interpolation graph is created for each of “FIT” and “UNFIT” for “Rhythm speed” (FIT evaluation history distribution interpolation graph in FIG. 7 and UNFIT in FIG. 8). (See Evaluation history distribution interpolation graph).
- a similar evaluation history distribution complement graph is created for other features. As a result of the interpolation, it is possible to calculate the score suitably even for the music having the “rhythm speed” which is not stored in the evaluation history.
- the distribution interpolation as described above is also performed for the feature amounts of all the music pieces stored in the feature amount storage unit 12.
- the whole music distribution graph showing the distribution of all music stored in the feature quantity storage unit 12 is approximated by a mixed Gaussian model or the like, and the whole music distribution interpolation is performed. Create a graph.
- FIG. 6 is a characteristic diagram showing an all-music distribution interpolation graph in the case where the feature amount of the music is “the speed of rhythm” according to the embodiment.
- FIG. 6 is a characteristic diagram showing an all-music distribution interpolation graph in the case where the feature amount of the music is “the speed of rhythm” according to the embodiment.
- the horizontal axis indicates “rhythm speed”, which is an example of the feature quantity
- the vertical axis indicates the “number of samples” of the music before the interpolation, and the “probability” that the sample exists after the interpolation.
- a similar all-music distribution interpolation graph is created for other feature quantities.
- the evaluation history correction unit 23 corrects the evaluation history distribution interpolation graph of FIG. 5 with the all-music distribution interpolation graph of FIG. 6 to create an evaluation history distribution interpolation correction graph (step S23).
- FIG. 7 is a characteristic diagram showing a process of creating a FIT evaluation history distribution interpolation correction graph in the case where the feature amount of the music is “rhythm speed” according to the embodiment.
- FIG. 8 is a characteristic diagram showing a process of creating an UNFIT evaluation history distribution interpolation correction graph when the feature amount of the music is “rhythm speed” according to the embodiment.
- the score curve creating unit 24 creates a score curve for each feature amount based on the FIT / UNFIT evaluation history distribution interpolation correction graph (step S24).
- “Score curve FIT evaluation history distribution interpolation correction graph
- UNFIT evaluation history distribution interpolation correction Create a score curve as shown in Fig. 9.
- FIG. 9 is a characteristic diagram showing a process of creating a score curve when the musical feature quantity is “rhythm speed” according to the embodiment. A score curve is similarly created for other feature amounts. As a result, not only the positive evaluation “FIT” but also the negative evaluation “UNFIT” is taken into account, so the score curve reflects the user's preference more strongly. Is obtained.
- Score curve FIT evaluation history distribution interpolation graph
- Score curve UNFIT evaluation history distribution interpolation graph
- Score curve FIT evaluation history distribution interpolation graph — UNFIT evaluation history distribution interpolation graph
- Score curve FIT evaluation history distribution interpolation graph Z whole song distribution interpolation graph
- Score curve -UNFIT evaluation history distribution interpolation graph Z whole song distribution interpolation graph
- Score curve (FIT evaluation history distribution interpolation graph UNFIT evaluation history distribution interpolation Graph)
- a score curve may be created according to any of the formulas of “Z total music distribution interpolation graph”.
- the score curve created according to the sixth equation is similar to that shown in Fig. 9.
- the score calculation unit 25 calculates the score of each piece of music in the search word "bright” (step S25). To calculate the score of a song M in the search term “bright”, the score for each feature amount is read on the score curve for each feature amount, and the total score read for each feature amount is calculated. And the final score for the search term “bright” for song M. For example, in the search term “bright” for song M, the score for “Rhythm speed” is 0.14, the score for “Average value of voice level” is 0.1, and the score for “Speed of chord change” is 0. 2. If the score for “frequency spectrum centroid” is 0.3, the final score for the search term “bright” for song M is 0.14 + 0. 1 + 0.
- the sorted music is output together with the identifier by the search result output unit 40 (step S40).
- the output music may be all of the sorted music, or a part of the music (for example, a score that is equal to or higher than a certain positive threshold). From the list of songs output as “bright” songs in this way, the user can finally select and play the song that he / she wants to watch. After viewing, the user can input the above-described evaluation of the music via the evaluation input unit 20 (step S20). As a result, the evaluation history is updated, and based on the updated evaluation history, a search that reflects the sensitivity of the user as much as possible becomes possible.
- the operation processing shown in the present embodiment may be realized by operating the content search device based on a content search method including a search step and an output step. Or you may implement
- the content evaluation device is not necessarily used for searching. It can of course be used alone or for other purposes.
- the present invention is not limited to the above-described embodiments, and can be appropriately changed within the scope of the invention and the gist of the invention which can also read the entire specification of the claims or the spirit of the invention.
- the content evaluation device and content search device, the content evaluation method and the content search method, and the first and second computer programs are also included in the technical scope of the present invention.
- the content evaluation device and content search device, the content evaluation method and content search method, and the first and second computer programs according to the present invention are, for example, a plurality of content stored in a recording device such as a music server. Among them, it can be used for a content evaluation device for evaluating the degree of conformity of one search word related to one content.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2008523708A JP4817338B2 (ja) | 2006-07-06 | 2007-07-04 | コンテンツ評価装置及びコンテンツ検索装置、コンテンツ評価方法及びコンテンツ検索方法、並びに第1及び第2のコンピュータプログラム |
US12/305,884 US20090313242A1 (en) | 2006-07-06 | 2007-07-04 | Content assesing apparatus, content searching apparatus, content assesing method, content searching method, and first and second computer programs |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2006186321 | 2006-07-06 | ||
JP2006-186321 | 2006-07-06 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2008004576A1 true WO2008004576A1 (fr) | 2008-01-10 |
Family
ID=38894543
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2007/063362 WO2008004576A1 (fr) | 2006-07-06 | 2007-07-04 | Dispositif de recherche/évaluation de contenu |
Country Status (3)
Country | Link |
---|---|
US (1) | US20090313242A1 (ja) |
JP (1) | JP4817338B2 (ja) |
WO (1) | WO2008004576A1 (ja) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012203523A (ja) * | 2011-03-24 | 2012-10-22 | Yamaha Corp | ライブ配信システム、データ中継装置及びプログラム |
JP2014007556A (ja) * | 2012-06-25 | 2014-01-16 | Nippon Hoso Kyokai <Nhk> | 聴覚印象量推定装置及びそのプログラム |
JP2016062519A (ja) * | 2014-09-22 | 2016-04-25 | カシオ計算機株式会社 | 情報表示機器、情報表示プログラムおよび情報表示方法 |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8661029B1 (en) | 2006-11-02 | 2014-02-25 | Google Inc. | Modifying search result ranking based on implicit user feedback |
US9092510B1 (en) | 2007-04-30 | 2015-07-28 | Google Inc. | Modifying search result ranking based on a temporal element of user feedback |
JP5317716B2 (ja) * | 2009-01-14 | 2013-10-16 | キヤノン株式会社 | 情報処理装置および情報処理方法 |
US9009146B1 (en) | 2009-04-08 | 2015-04-14 | Google Inc. | Ranking search results based on similar queries |
US8447760B1 (en) | 2009-07-20 | 2013-05-21 | Google Inc. | Generating a related set of documents for an initial set of documents |
US8498974B1 (en) | 2009-08-31 | 2013-07-30 | Google Inc. | Refining search results |
US8972391B1 (en) | 2009-10-02 | 2015-03-03 | Google Inc. | Recent interest based relevance scoring |
US9623119B1 (en) * | 2010-06-29 | 2017-04-18 | Google Inc. | Accentuating search results |
US9002867B1 (en) | 2010-12-30 | 2015-04-07 | Google Inc. | Modifying ranking data based on document changes |
CN103839559B (zh) * | 2012-11-20 | 2017-07-14 | 华为技术有限公司 | 音频文件制作方法及终端设备 |
US9508329B2 (en) | 2012-11-20 | 2016-11-29 | Huawei Technologies Co., Ltd. | Method for producing audio file and terminal device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003132085A (ja) * | 2001-10-19 | 2003-05-09 | Pioneer Electronic Corp | 情報選択装置及び方法、情報選択再生装置並びに情報選択のためのコンピュータプログラム |
JP2005352754A (ja) * | 2004-06-10 | 2005-12-22 | Sharp Corp | 情報ナビゲーション装置、方法、プログラム、及び記録媒体 |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1378912A3 (en) * | 2002-07-02 | 2005-10-05 | Matsushita Electric Industrial Co., Ltd. | Music search system |
JP4302967B2 (ja) * | 2002-11-18 | 2009-07-29 | パイオニア株式会社 | 楽曲検索方法、楽曲検索装置及び楽曲検索プログラム |
JP3892410B2 (ja) * | 2003-04-21 | 2007-03-14 | パイオニア株式会社 | 音楽データ選曲装置、音楽データ選曲方法、並びに、音楽データの選曲プログラムおよびそれを記録した情報記録媒体 |
JP2005173938A (ja) * | 2003-12-10 | 2005-06-30 | Pioneer Electronic Corp | 曲検索装置、曲検索方法及び曲検索用プログラム並びに情報記録媒体 |
US7664760B2 (en) * | 2005-12-22 | 2010-02-16 | Microsoft Corporation | Inferred relationships from user tagged content |
-
2007
- 2007-07-04 WO PCT/JP2007/063362 patent/WO2008004576A1/ja active Search and Examination
- 2007-07-04 US US12/305,884 patent/US20090313242A1/en not_active Abandoned
- 2007-07-04 JP JP2008523708A patent/JP4817338B2/ja not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003132085A (ja) * | 2001-10-19 | 2003-05-09 | Pioneer Electronic Corp | 情報選択装置及び方法、情報選択再生装置並びに情報選択のためのコンピュータプログラム |
JP2005352754A (ja) * | 2004-06-10 | 2005-12-22 | Sharp Corp | 情報ナビゲーション装置、方法、プログラム、及び記録媒体 |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012203523A (ja) * | 2011-03-24 | 2012-10-22 | Yamaha Corp | ライブ配信システム、データ中継装置及びプログラム |
JP2014007556A (ja) * | 2012-06-25 | 2014-01-16 | Nippon Hoso Kyokai <Nhk> | 聴覚印象量推定装置及びそのプログラム |
JP2016062519A (ja) * | 2014-09-22 | 2016-04-25 | カシオ計算機株式会社 | 情報表示機器、情報表示プログラムおよび情報表示方法 |
Also Published As
Publication number | Publication date |
---|---|
JPWO2008004576A1 (ja) | 2009-12-03 |
JP4817338B2 (ja) | 2011-11-16 |
US20090313242A1 (en) | 2009-12-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2008004576A1 (fr) | Dispositif de recherche/évaluation de contenu | |
US7696427B2 (en) | Method and system for recommending music | |
JP5432264B2 (ja) | コレクションプロファイルの生成及びコレクションプロファイルに基づく通信のための装置及び方法 | |
US20140260912A1 (en) | Sound signal analysis apparatus, sound signal analysis method and sound signal analysis program | |
US7521620B2 (en) | Method of and system for browsing of music | |
CN104715760B (zh) | 一种k歌匹配分析方法及系统 | |
JP3892410B2 (ja) | 音楽データ選曲装置、音楽データ選曲方法、並びに、音楽データの選曲プログラムおよびそれを記録した情報記録媒体 | |
JP3797547B2 (ja) | プレイリスト生成装置、オーディオ情報提供装置、オーディオ情報提供システムおよびその方法、プログラム、記録媒体 | |
EP1798643A2 (en) | Taste profile production apparatus, taste profile production method and profile production program | |
JP4312663B2 (ja) | 楽曲選曲装置、楽曲選曲方法、プログラムおよび記録媒体 | |
WO2007077993A1 (ja) | 情報処理装置および方法、並びに記録媒体 | |
KR20180090261A (ko) | 플레이리스트 리스트 결정 방법 및 디바이스, 전자 장비 및 저장 매체 | |
CN102456342A (zh) | 音频处理装置和方法以及程序 | |
JP4560544B2 (ja) | 楽曲検索装置、楽曲検索方法および楽曲検索プログラム | |
JP2005173938A (ja) | 曲検索装置、曲検索方法及び曲検索用プログラム並びに情報記録媒体 | |
JP2012058513A (ja) | 楽曲再生システム | |
JP4921343B2 (ja) | 演奏評価装置,プログラムおよび電子楽器 | |
JP4322691B2 (ja) | 選曲装置 | |
CN109410972B (zh) | 生成音效参数的方法、装置及存储介质 | |
JP4594701B2 (ja) | 情報検索装置、情報検索方法及び情報検索用プログラム並びに情報記録媒体 | |
KR101675957B1 (ko) | 신호 성분 분석을 이용한 음악 인기도 예측 시스템 및 방법 | |
JP4475597B2 (ja) | 提示データ選択装置及び提示データ選択方法等 | |
JP2023036600A (ja) | レコメンド装置、レコメンド方法、及びプログラム | |
JP5541531B2 (ja) | コンテンツ再生装置、楽曲推薦方法およびコンピュータプログラム | |
WO2006030712A1 (ja) | 楽曲推薦装置及び方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 07790439 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2008523708 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 12305884 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
NENP | Non-entry into the national phase |
Ref country code: RU |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 07790439 Country of ref document: EP Kind code of ref document: A1 |
|
DPE1 | Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101) |