JP2008522309A - Adapting temporal similarity threshold in related content extraction - Google Patents

Abstract

Extraction of similar content items or documents is provided based on related time similarities such as generation or usage time. The time of one or more identified anchor documents, such as in a database, is determined by extracting time metadata. Thereafter, the first threshold is set based on a base time or a reference time difference determined between the current time and the anchor item time. Candidate document item times may be identified and temporal metadata for candidate content items may be extracted. The time difference between the base time and the candidate time may be determined as the candidate time difference. Candidate content items may be selected as similar for extraction based on the first candidate time difference and the first threshold. A concept of a reference time difference of the determined granularity for setting the threshold is provided.

Description

  The present invention relates to the field of content extraction, management and presentation, and to content item similarity threshold determination based on time usage and metadata.

  In recent years, the storage capacity of storage devices and databases, including hard drives and other types of storage media on personal computers, has been rapidly increasing. One expectation is that storage capacity has almost doubled each year, and network bandwidth has also increased very rapidly. As a result, the storage device stores a larger amount of content that requires user access to it. Users may be overwhelmed by content stored on a storage device or database, as well as on their own hard drive, and managed or organized in any way so that the content provides convenient access for the user. If not, you may not be able to extract content that is available on networks such as the Internet. Content that has not been indexed or organized in a way that is transparent to the user may be “missed” as far as the user needs are concerned and may not be extracted.

  Many data extraction schemes are known. Farnham et al., US Patent Application Publication No. 2003/0158855, is an automatic method in which associations are dynamically generated between objects or metadata so that the similarity expressed as a numerical value between computer files is determined. The context association is disclosed. US Patent Application Publication No. 2002/0188602 to Stubler et al., Based on the similarity between a stored image and an acquired image by extracting metadata of the acquired image, Disclosure of label generation. Platt's US Patent Application Publication No. 2003/0221541 discloses an automatic playlist generator in which multiple seed songs, including “undesirable seed” songs, are used to generate songs on a playlist. is doing. Cluts US Pat. No. 5,616,876 discloses selecting additional music similar to the first music set based on the “style label” of each music previously written by the editor. Gargi's US Patent Application Publication No. 2004/0098362 discloses automatic propagation of document metadata including generation time. However, none of the above references disclose setting a threshold for temporal similarity in target item selection or rejection. Prince's US Patent Application Publication No. 2002/0099696 discloses a fuzzy database extraction in which similarity is scored and a threshold is used to select items to be extracted. However, neither Prince nor any other reference discloses or suggests setting a threshold based on the base time, or setting a threshold based on the time difference between the base time and the identified item time.

  Of course, it is possible for the user to extract content items, but trying to identify similar items is time consuming and tedious, especially if the content type of the desired content is not specified or known by the user. Can be a painful task. Furthermore, as content items continue to accumulate in a storage device or database controlled by the user, the task of extracting content items becomes more difficult.

  Methods, systems, devices, apparatuses and computer readable media are provided that implement or perform the functions of the extraction system.

  A selected candidate content item is provided. A base time is determined. Such a base time may be the current time, for example. The first time is specified by extracting time data of the first specified content item. Thereafter, the first threshold value may be set based on a reference time difference determined between the base time and the first time. The candidate time is specified, and the time data of the candidate content item is extracted. The difference between the base time and the first candidate time may be determined as a candidate difference. The candidate content item may be selected as similar for organizing the database or for extraction based on the first candidate time difference and the first threshold, and a selection signal for the selected candidate content Is output accordingly.

  The concept of time-determined time difference for setting the threshold is provided, so that the difference granularity is greater for times closer to the base time than for times further away from the base time. Set to

  Further, a second threshold value based on the reference time difference may be set, and the second threshold value has a certain range together with the first threshold value, and becomes a candidate when the time difference to be the first candidate is within the range. A content item is selected.

  The first time may include a content item acquisition time, a most recent use time of the content item, or a time when the content item is most often used. This time may be a content item base time, a content item latest change time, or a content item generation time.

  Further identified content items may be identified, and the time and difference are determined such that the first threshold can also be set based on these determined reference time differences.

  The following description and the above drawings describe embodiments of Applicant's invention as best understood by the inventor, but many modifications of the invention are possible, and the invention is intended to It will be understood that other forms may be implemented and may be implemented in other ways without departing from. Moreover, some features of the described embodiments may be omitted, selectively or wholly combined with other embodiments, or other embodiments or one of them without departing from the spirit of the invention. It may be used to replace part features. Accordingly, the drawings and detailed description are to be taken as illustrative illustrations of certain aspects of the invention and should not be construed as limiting the scope of the invention.

  As shown in FIG. 1, the extraction system 1-1 has several modules to be described later. The module of the extraction system 1-1, a part thereof, and / or the entire extraction system can be configured by hardware, software, firmware, or a combination of the above, but some modules are configured by hardware, etc. The module may be composed of software, firmware, or a combination thereof.

  It will be understood that not all modules of the extraction system need be located or integrated in the same device. A decentralized architecture is also conceivable for an extraction system that “piggy-backs” from the appropriate modules provided by existing equipment.

  The following description refers to an extraction system 1-1 that is physically integrated or connected to the database 1-2 via a wired or wireless connection. A clock (not shown) may also be integrated or connected to the extraction system 1-1. The database 1-2 may be realized by a storage device such as a personal computer, a personal video recorder, an entertainment system, an electronic organizer, a personal portable device, a hard drive of a Jaz drive, or realized as a commercial storage device such as a disk drive. May be. It will be appreciated that the database 1-2 may include a number of storage devices that are connected to allow for the organization or grouping of content items on two or more of such devices. . Further, the database is understood to include one or more storage media such as CDs, DVDs, zip disks, floppy disks, disks including data cartridges, etc. that can be loaded into and extracted from the database 1-2. It may be. However, it will be appreciated that the extraction system 1-1 can also extract content via a network 1-9 such as a LAN, WAN, or the Internet.

  As shown in FIG. 1, the extraction system 1-1 includes a time data extraction device 1-11, which is a module that collects specific types of data from content items. The content item may be a video, video clip, movie, photo, text file, music data, audio file, or other type of multimedia data such as a JPEG file or XML data. For example, the video may be a home video shot on a digital video recorder, and the movie is commercially available film data such as a film encoded as MPEG (including MPEG-2, MPEG-3, etc.). Alternatively, the photo may be digital photo data, a photo, or a photo album group. The text file may be a file generated by a word processor, a spreadsheet, or a computer code file. The music data may be an MP3 file or the like.

  The description data extracted by the time data extraction device 1-11 includes information such as usage data or metadata regarding the content item. Such information may also include item creation time, item acquisition time, content item last / first / second etc. playback and / or editing time, most commonly used time, eg It may include time data for content items such as around 8 o'clock, mainly used for a given day of the week, January or year, or often used at night. As used herein, “mainly” may be based on average usage time, median usage time, usage time mode, and the like. Such usage history data may sometimes be known as metadata, while the type of metadata is sometimes referred to as usage history data.

  The temporal information described herein may be one or many such similarity features, or it may be the only or most important feature. The degree to which such factors are weighted (if any) will depend on the user needs and the application.

  Such descriptive data about the item includes from the item, from an index or database management file, from an external source such as the World Wide Web connected to the extraction system 1-1 via a wired or wireless connection to the Internet 1-9. It will be appreciated that it may be arranged and extracted by various methods.

  The identified content item may be identified by one of several methods. The user may specify items from which other items, sometimes called “candidate content items”, are extracted based on. Alternatively, a newly added or generated content item may be automatically designated as a specified content item from which another item is extracted based on the content item.

  The base time is determined by the base time determination device 1-13. Such base time is the current time entered or set by the user, previously programmed, determined with reference to a clock (not shown), or determined by the Internet or other network or combination of the above. There may be. It will be appreciated that the base time, the time associated with the identified content item, and the candidate content item may each include a date and / or time. For some applications, a date with no time may be sufficient or more appropriate. For many applications, both time and date are used. It will be appreciated that such date information and time information can be converted to a format that allows for time difference calculations and comparisons with other dates and times.

  The time data extraction device 1-11 determines the time related to the specified content item, and determines the time difference (that is, the elapsed time) between the time related to the specified content item and the base time. This difference is sometimes called the first reference time difference. The time associated with the identified content item refers to the metadata associated with the content item, by referring to the network 1-9, including the World Wide Web, by database index, by requesting user input, or May be determined by a combination of the above. This time difference may be determined by referring to a table, by calculation, by requesting user input, or by a combination of the above.

  The threshold setting device 1-14 sets a threshold or a range that needs to be satisfied because a candidate content item is selected. This threshold value or range is set by the threshold setting device 1-14 based on the first reference time difference.

  Candidate content item identification devices 1-12 are similar in terms of their metadata or other information in the database or from other sources over a network connection and / or first identified to base time Identify candidate content items based on their differences from the base time relative to the time difference of the content items.

  The controller 1-15 adjusts the overall function of the extraction system 1-1, interacts with the user interface 1-1, the database 1-2, the server 1-9 and generally the outside, and processes system settings.

  The selection device 1-16 selects eligible content items that are eligible, and the result output 1-17 provides a result signal for the selected and / or rejected candidate content items. The result output 1-17 interfaces with other devices and external communications, including interaction with a user (not shown). In particular, the extraction result output 1-17 notifies the user interface of the content item extracted by the extraction system 1-1. The user interface 1-3 may be an independent device or may be integrated with other devices or systems, such as a personal computer, a personal video recorder, the storage described above, and one or more of other devices.

  There may be additional identified content items available. Accordingly, the time metadata processing and / or usage extraction and time difference determination may be repeated for any number of available identified content items 1-N (N is a positive integer greater than 1). . Thereafter, selection of candidate content items is performed based on the average of all such reference time differences.

  The processing of the embodiment of the present invention will be described with reference to FIGS. As described above, automatically identified by the system via the user interface 1-3 shown in FIG. 1 or by detecting a newly added or isolated content item in the database 1-2. Is done.

  In S1, as described above, the base time determination device 1-13 determines the base time. The time data extraction device 1-11 of the extraction system 1-1 extracts the first time data of the first identified content item as shown in S2 of FIG. In S2, further identified content items may be processed in a similar manner, for example, when the user or system specifies several “anchor” documents from which the target document is extracted.

  In S3, a reference time difference is determined by the time metadata extraction device 1-11 by determining a time difference between the base time and the time of the identified content item for each identified content item. In S4, such a reference time difference may be averaged to arrive at an average reference time difference. The average used here may be determined based on arithmetic average, mode or median calculations. Furthermore, the simple sum of the values may be utilized with a statistical function appropriately selected to provide a synthetic view of the selected value.

  In S5, a threshold value or range is set based on the reference time difference or the average reference time difference. For example, a value of 1 or 0 indicates that the time difference between the base time and the first identified content item is small, and a value of 9 or 10 indicates that the time difference is large. May be assigned. Alternatively, instead of using a scale of 1-10, the threshold may be, for example, “same time”, “very close time”, “close time”, “away time”, “very far time”, or the like It may represent a specific designation. It will be understood that many other schemes of such values can be used without departing from the spirit of the invention.

  Further, the second threshold value may be selected in the same manner. The first threshold value represents the maximum time difference, the second threshold value represents the minimum time difference, and together these constitute a range with these threshold values. Candidate items are selected only if their differences fall within the range (difference between base time and candidate content item time).

  In S6, the candidate content item specifying device 1-12 shown in FIG. 1 specifies a candidate content item in the database via the network or in any case, and the time data extracting device 1-11 (FIG. 2). Extracts the time data of each candidate content item. Thereafter, in S7, the determination of the time difference between the candidate content items is executed. Further identified content items are also available, and the process of extracting time data and determining the time difference is continued for candidate content items 1-M.

  In S8, the time difference between candidate content items is compared with a threshold value or a range by the selection device 1-16. If the difference is less than or within the threshold, it is selected.

  For example, if the base time is determined as 8:00 pm on November 22, 2004 and the time of the first identified content item is determined to be 6:00 pm on November 22, 2004, the base time and A time difference of 2 hours representing a difference from the first identified content item becomes a first reference time difference. Based on the time difference of 2 hours or based on the average of the reference time differences determined from the identified content items 1-N, the time difference threshold values are “4 hours”, “same day”, “same days”, “close time difference”. Alternatively, it is set as “4” or the like. (4 is an integer assigned from 0 to 9, 0 means substantially the same time, and 9 means a very far time.) Thereafter, the difference between the candidate content items is the threshold value. In S8, the time difference of the candidate content item from the base time is within 4 hours, “same day”, “same number of days”, “near time difference”, or “4” on the time difference scale. If it is due to a threshold, a candidate item is selected.

  According to a feature of the invention, the threshold is set such that the difference granularity is greater for times closer to the base time than for times further away from the base time. Thus, for example, if the distance from the base time is ranked by a scale of 1-10, as the difference from the base time increases, a larger difference may be captured by a smaller scale gradation. Thus, if the reference time difference (difference between the base time and the identified document or content item) is 1 hour, the content item that is the first candidate if the candidate difference from the base is 5 hours May be considered dissimilar. However, if the reference time difference is 1000 days, the second candidate document may be determined to be similar even if the second candidate difference from the base is 6000 days. Such thresholds are often based on the idea that people intuitively do not consider a time difference at a previous time point as important as a more recent time difference of the same length. That is, as the person moves farther away from the associated base time in the time difference, the time difference between time points becomes less important in determining similarity. Such threshold processing is called a reference time difference of the determined granularity threshold processing.

  As described above, a range may be generated in S5 by the threshold setting device 1-14 using the maximum and minimum thresholds based on the spread in the difference set of identified content items. Thus, using the above example with only one identified content item, the maximum threshold is as described, and the minimum threshold is, for example, “different time”, “at least one hour”, very The near time is “or differential ranking 2”. At this time, the range of the selected content item is “different time of the same day”, “different time of the same period”, “1 to 4 hours”, “time proximity”, depending on the threshold processing / range system used. May be a "very close difference" or "2-4" scale.

  According to a feature of the present invention, multiple “base” times are available, and for each such base time, a reference time difference of the determined granularity is applied individually. For example, the second base time may be the time of a past important event, such as a user's wedding date, child's birthday, or anniversary, and the actual current time. For this reason, the granularity level decreases due to the difference from the base time 1 (the more elapsed time is considered to be similar as the candidate document time becomes the past), and similarly decreases due to the time difference from the base time 2. . The idea is that, for example, for a person who has a wedding on a certain date, the time difference closer to the second base time becomes more important and therefore requires a higher granularity.

  Such second, third and Lth (L is an integer greater than 3) base times can be set by the user or can be determined by the system according to the method for determining the base time described above. is there. Thus, for example, a system may have a large or large percentage of documents related to a user (such as a document on a user's computer, database, mobile phone, etc.) that has a content item time (such as the date of creation or most recent use / time) at a certain time. If so, such additional base time can be set. For example, for illustrative purposes, based on statistical functions that show a much higher concentration than usual on or near a user's past wedding date for content items such as wedding photos, wedding videos, music, email, etc. Or, using a threshold, many content items can be determined as such additional base times. Thereafter, a reference time difference can be determined, and a threshold is set according to the determined reference time difference of the granularity based on such further base time.

  According to a feature of the present invention, the extracted content item may be a content type different from the content type of the content item selected by the user. For example, if the content item selected by the user is a music file, ie, MP3 type, the extracted content item may be the content type of photo data. In this way, for example, a photo of a certain genre may be extracted so as to match the music selected by the user based on the temporal similarity.

  This (or these) selected candidate content items are provided to the user or user interface in S9. A signal may be provided directly to the database 1-2 to cause extraction of selected candidate items for the database or user interface 1-3. Alternatively (or in addition), a signal may be provided if a candidate content item is rejected.

  A notification for notifying a user (not shown) of extractable content items may be provided to the user interface 1-3. This notification may consist of identifying the content item to be extracted, a description of the content item, the URL or link of the content item, extracting the entire content item or part thereof, or a combination of the above. The system can also be used to group extracted items selected by anchor items to organize the database. In S10, the process ends.

  The embodiments of the present invention provided by the above description are intended as exemplary embodiments only. However, it will be understood that the scope of the invention is provided by the claims.

FIG. 1 is a schematic diagram of an extraction system according to an embodiment of the present invention. FIG. 2 is a flowchart of the processing of the system according to the embodiment of the present invention.

Claims (13)

Determining a base time;
Extracting the time metadata of the first identified content item as a first time;
Setting a first threshold based on a reference time difference determined between the base time and the first time;
Extracting time metadata of a content item as a first candidate as a first candidate time, and determining a difference between the base time and the first candidate time as a first candidate time difference;
Selecting the first candidate content item based on the first candidate time difference and the first threshold;
Outputting a selection signal of the first candidate content item when the first candidate content item is selected;
A content item extraction method comprising:   The method of claim 1, wherein the first threshold is set based on a determined reference time difference in granularity.   Setting a second threshold having a range together with the first threshold based on the reference time difference, and selecting the first candidate content item when the first candidate time difference is within the range; The method of claim 1 further comprising:   At least one of the first time and the first candidate time includes a content item acquisition time, a time when the content item was most recently used, a time when the content item was used first, and a time when the content item is used most often. The method of claim 1.   The method of claim 1, wherein at least one of the first time and the first candidate time comprises a content item last modification time and a content item generation time. Extracting time metadata of the second identified content item as a second time;
Setting the first threshold based also on a second reference time difference determined as a difference between the base time and the second time;
The method of claim 1, further comprising: A base time extractor configured to determine a base time;
The time metadata of the first identified content item is extracted as a first time, the time difference between the base time and the first time is determined as a reference time difference, and the metadata of the first candidate content item is A time metadata extraction device configured to extract as one candidate time and determine a time difference between the base time and the first candidate time as a first candidate time difference;
A threshold setting device configured to set a first threshold based on a reference time difference determined as a difference between the base time and the first time;
A selection device configured to select the first candidate content item based on the first candidate time difference and the first threshold;
A result output configured to output a selection signal for the first candidate content item when the first candidate time is selected;
A content item extraction system.   The system according to claim 7, wherein the first threshold is set based on a reference time difference of the determined granularity. The threshold setting device further sets a second threshold having a range together with the first threshold based on the reference time difference,
The system according to claim 7, wherein the selection device selects the first candidate content item when the first candidate time difference is within the range.   At least one of the first time and the first candidate time has a content item acquisition time, a time when the content item was most recently used, a time when the content item was used first, and a time when the content item was most used, The system of claim 7.   The system according to claim 7, wherein at least one of the first time and the first candidate time includes a content item latest change time and a content item generation time.   8. The at least one of the first time and the first candidate time has one of a frequently used time, a time most recently used by a user, and a time most often used by a user. System. The time data extraction device is configured to extract time metadata of a second identified content item as a second time;
8. The system of claim 7, wherein the threshold setting device is configured to set the first threshold based also on a second reference time difference determined as a difference between the base time and the second time. .

Images