HK1197945B - Media object metadata association and ranking - Google Patents

Description

Media object metadata association and ranking

Divisional Application Instructions

This application is a divisional application of Chinese invention patent application No. 200680022266.X (PCT international application PCT/US2006/015263), filed on April 20, 2006, and entitled “Media Object Metadata Association and Ranking”.

Cross-reference to related applications

This application claims priority to U.S. Application No. 60/674,109, filed April 21, 2005, entitled "GENERATION AND USE OF METADATA FORMEDIA OBJECTS," which is incorporated herein by reference in its entirety.

This application is related to concurrently filed U.S. Application No. _____, entitled “INTERESTINGNESS RANKING OF MEDIA OBJECTS,” which is incorporated herein by reference in its entirety.

Technical Field

The present invention relates generally to the organization and display of media objects, and more particularly to the association of metadata with media objects (eg, images) and their ranking.

Background Art

Existing websites allow users to annotate and review media such as books or movies. However, traditional websites are limited in the type of information that users can provide directly or indirectly through user actions about media objects, and in the use of that information to generate search results.

Search results rely on rankings of items to determine the most relevant items to present to the searcher. These rankings can be based on criteria such as the number of times a particular item has been "clicked" or viewed by users. It would be desirable to make available a variety of user-derived information about media objects and to develop more relevant rankings for media objects based on this information.

Summary of the Invention

Embodiments of the present invention enable source data to be associated with media objects by providing one or more media objects for display to one or more users and accepting input from the users regarding the one or more media objects. The media objects may include images (e.g., still or moving images) or audio media objects (e.g., audio tracks). The input may include at least two different types of metadata, including tags, ratings, annotations, descriptions, and adding to favorites ("Set to Favorites") or playlists ("Set to Playlist"). Tag metadata may include location metadata. The logic for implementing this embodiment may be located at a server, and each user may be associated with a corresponding client computer.

Embodiments of the present invention may also perform a search for media objects using at least one metadata set and rank one or more images returned from the search. The ranking may be based at least in part on user actions associated with the media object, including the amount of metadata associated with the media object input by the user, the number of users who have assigned metadata to the media object, and/or access patterns associated with the media object. The access patterns may be based at least in part on the number of click-throughs or views of the media object. The ranking may also be based at least in part on the relationship between the publisher of the media object and the user who initiated the search and/or the passage of time associated with the media object. Embodiments of the present invention may base the ranking at least in part on the relevance of one or more tags to the media object, where the relevance itself is based on relevance input from one or more users.

Embodiments of the present invention may include statistical logic for determining a metric based, at least in part, on how often a first set of metadata and a second set of metadata are commonly assigned to the same media object. The statistical logic may determine clusters of related metadata based on the metric for a plurality of metadata associated with a plurality of media objects, and rank the associated plurality of media objects within each cluster. The statistical logic may also provide the first metadata and the second metadata for display as related metadata if the determined metric exceeds a threshold relevance value. Embodiments of the present invention may determine the appropriateness of the first metadata and the second metadata for an image based, at least in part, on an appropriateness input from a user, in which case the statistical logic may change the relevance metric based, at least in part, on the determined appropriateness.

Embodiments of the present invention may determine the frequency with which at least one specific metadata set is assigned to a plurality of media objects in a predetermined time period and/or determine the number of media objects to which the at least one specific metadata set is assigned.

Embodiments of the present invention may provide information used by an ad server to associate advertisements with images or other media objects. In response to a search or other user action that results in access to the media object, the media object may be provided for display to a user. Metadata logic may accept metadata about the media object from multiple users. Statistics logic may determine a relevance metric based, at least in part, on the frequency with which at least one metadata item is assigned to the media object. The metadata set and metric may be made available to the ad server.

The ad server may associate an advertisement with the media object based at least in part on the metadata and, optionally, on a relevance metric. In an embodiment of the present invention, the statistical logic may determine the metric based at least in part on how often the first set of metadata and the second set of metadata are jointly assigned to the media object. The first and second metadata may be made available to the ad server.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 shows a client-server system according to an embodiment of the present invention.

FIG. 2 is a screenshot illustrating entry of tag metadata into a media object according to an embodiment of the present invention.

FIG. 3 illustrates adding annotation metadata according to an embodiment of the present invention.

FIG. 4 illustrates setting permissions according to an embodiment of the present invention.

DETAILED DESCRIPTION

The following description is given to enable those skilled in the art to implement and use the present invention. The description of specific devices, techniques and applications is provided as an example only. It will be clear to those skilled in the art that various modifications to the examples described herein, and the general principles defined herein can be applied to other examples and applications without departing from the spirit and scope of the present invention. Therefore, the present invention is not intended to be limited to the examples described and shown herein, but rather to conform to the scope consistent with the claims.

FIG1 illustrates a client-server system according to an embodiment of the present invention. A media server according to an embodiment of the present invention may include HTTP web server logic, a scripting engine (e.g., a PHP scripting engine), a database, and a convergence engine. The media server may communicate with multiple clients over a network (e.g., the Internet).

The scripting engine may include authentication logic, upload logic, metadata processing logic, and license processing logic. The authentication logic authenticates users logging into the media server website. The upload logic may be used to upload data from the client to the server in any media format, such as still photos (e.g., JPEG, TIFF), video (e.g., MPEG, AVI), or audio (e.g., MP3, OGG). The aggregation engine may include a statistics engine and a search engine. Clients for use with the server of the present invention may include typical web browser applications. Many of the features of the present invention can be observed at www.flickr.com, which is incorporated herein by reference in its entirety.

According to an embodiment of the present invention, a user at a client uses a browser to access a media server and request an upload of a media object. In response, the upload logic stores the media object from the client in a database. For convenience, we will frequently use images as an example of media objects manipulated by the system, but those skilled in the art will recognize that the present invention is also applicable to other media objects, with appropriate modifications and using other functions, where appropriate (for example, viewing a media object can be applied to viewing still images or moving images, or listening to an audio media object, such as a soundtrack).

The metadata processing logic allows users to enter metadata to describe each image. (As described herein, "metadata" can refer to a metadata set or multiple metadata.) Referring to FIG2 , the metadata can take the form of one or more tags for each image, for example, four different tags entered as a spatially bounded list "cloudsseagullbirdssky" for an image of seagulls in flight. Other types of metadata include titles (e.g., "Last gull (for now)"), descriptions of the image, annotations, and comments. Annotations are descriptive notes displayed directly on a portion of the annotated image. Annotations can be hidden from view until the user moves the cursor over the annotated portion. Referring to FIG3 , for example, a user can add a note near the seagull's wings, such as "Notice the sunlight shining through the wings." Comments can be entered into a text input box similar to that used for entering comments on a message board. Multiple comments from any authorized user can be made and displayed for a media object.

Referring to FIG4 , the permission logic enables the user to set permissions regarding who is allowed to view each image. For example, the user can set permissions to allow access only to the user, a limited group of people (e.g., family members and/or friends), or the public (e.g., the entire user base). Additionally, the permission logic allows others to provide metadata about each image. For example, the user can allow friends and/or family members, any other users, or certain people from the user's contact list to add tags, comments (e.g., "I love the way the seagulls are circling"), or annotations, or not allow anyone to add tags, comments, or annotations.

The database makes it possible to organize media objects in a variety of ways. For example, a user's media objects can be organized by time, which allows the user to search for media objects by date. This organization allows media objects (e.g., photo images, still images representing video files, or icons representing audio files) to be presented on the user's display along a timeline or according to a calendar format (e.g., images selected from each day are displayed at the corresponding date entry). Media objects can be displayed according to the date the media object was created, or the date it was uploaded, according to the user's display format selection. In addition, a script engine allows media objects to be displayed in a slideshow format.

For photos, the creation date can be based on metadata provided by the device, such as metadata from the camera that took the photo, including information about aperture, shutter speed, and other settings. This metadata can be extracted from, for example, the EXIF (Exchangeable Image File) headers stored in files uploaded by many digital cameras, or the IPTC (International Press Telecommunications Council) headers stored in files uploaded by many photo management and organization software packages. The temporal organization of photo images can be referred to as a "photostream."

The database also allows users to organize uploaded media objects into collections, identified and described by user-provided collection identifiers and descriptions. Each image collection is similar to a photo album, for example. Each media object can belong to multiple collections. The collection identifiers and descriptions themselves are also metadata.

In contrast to collections that include media objects from only one user, media objects from multiple users can be aggregated into "groups" using a database. Each group is identified by a group identifier provided by the user who created the group. Grouping all groups together represents all media objects of a particular type (e.g., images) accessible on the media server that hosts the media objects. The group creator can set various permission levels for accessing and adding media objects to the group. Permission levels can include, for example, public to the entire user base or private to friends/family members or a user-defined social network. Users who are granted access to a group are also permitted to add tags, comments, and/or annotations. Similar to displaying and organizing an individual user's media objects, the database enables group media objects to be organized and displayed in a timeline or calendar format arranged by date, or in a slideshow format.

A media server may include an RSS feed generator that allows a user to subscribe to a "feed" of media objects, such as media objects belonging to a particular group or identified by a particular tag, sorted, for example, by the date they were published. (A "group" may refer to any collection, such as all groups of media objects, a single group of media objects for multiple users, all media objects for an individual user, or a collection (i.e., a subset) of an individual's media objects.) An RSS reader on a user's client computer may be configured so that only new media objects (e.g., updates) published after the most recently accessed media object are presented to the reader. Similarly, the reader may be configured so that only the most popular comment threads related to a particular media object are displayed using this functionality.

The statistics engine generates statistics and other metrics based on the aggregated metadata. In one embodiment, the statistics engine determines the popularity of metadata (e.g., tags) within a group of media objects over a predetermined time period. For example, the statistics engine can determine the number of unique users who have assigned a particular tag to one or more media objects in all groups, a single group, or a collection of media objects in the system over the last 24 hours. The aggregation engine can determine (and display) a histogram of the tags and can determine the most frequently assigned tags (at any point in time or over a predetermined time period) by determining those tags whose frequency exceeds a minimum threshold frequency or falls within a predetermined number of the most popular tags.

In one embodiment of the present invention, a predetermined number of metadata (e.g., tags) or terms within metadata (e.g., terms within comments) may have their frequency indicated by the size of the font used to display them. For example, the 100 most popular tags within all groups may be arranged in alphabetical order on the user's display, with increasing popularity indicated by increasing font size.

In another embodiment, the statistics engine can determine the "relatedness" of metadata, i.e., the frequency with which a particular set of metadata (e.g., a tag) (or items within a metadata set (e.g., a comment)) is assigned to a media object and a co-occurrence measure with at least one other particular set of metadata (or items within a metadata set). In one embodiment, the co-occurrence measure can determine the frequency with which metadata of the same type co-occurs. For example, of all 100 images tagged with "Italy," 50 of these images can also be tagged with "Rome," 25 with "Venice," 10 with "Florence," and 2 with "Sienna." The co-occurrence indices are: 50 "Italy-Rome," 25 "Italy-Venice," 10 "Italy-Florence," and 2 "Italy-Sienna," respectively. In summary, a tag MD can include locations, including locations as a subset of tags.

In another embodiment, the relevance metric can be user-specific, such that it is a frequency measure of the number of users who assigned at least one media object to a particular metadata set (e.g., a tag) (or an item within a metadata set (e.g., a comment)) and a co-occurrence measure with at least one other particular metadata set (or item within a metadata set). For example, of all 100 users who tagged at least one image with the word "Italy," 50 users might have also tagged images tagged "Italy" with "Rome," 25 with "Venice," 10 with "Florence," and 2 with "Sienna." The co-occurrence indices are: 50 "Italy-Rome," 25 "Italy-Venice," 10 "Italy-Florence," and 2 "Italy-Sienna," respectively.

A predetermined number of metadata (e.g., tags) with the highest co-occurrence index, or those with a co-occurrence index exceeding a predetermined threshold, can be displayed to the user as "related" metadata (e.g., tags), while at least one set of metadata (e.g., tags) that does not meet the predetermined criteria is displayed under "See also." The predetermined threshold can be calculated as a percentage of the maximum possible value of the index. All such displayed metadata can serve as hyperlinks to all media objects assigned the specified metadata. The relevance measure can be applied to all "public" media objects (i.e., those available to anyone on the system) or to smaller groups (e.g., those within a group or collection).

As part of the relevance calculation, the statistical engine can employ statistical cluster analysis, known in the art, to determine the statistical similarity between metadata (e.g., tags) and group the metadata and associated media objects according to corresponding clusters. For example, among 10,000 images tagged with the word "Vancouver," one statistical cluster within a threshold similarity level may include images also tagged with "Canada" and "British Columbia." Another statistical cluster within the threshold similarity level may instead be tagged with "Washington" and "space needle" along with "Vancouver." The cluster analysis allows the statistical engine to associate "Vancouver" with both the "Vancouver-Canada" cluster and the "Vancouver-Washington" cluster. The media server can present the two sets of related tags to the user for display, indicating, for example, that they belong to different clusters corresponding to different subject areas.

One embodiment of the present invention allows a user to determine the relevance of a tag to a media object, particularly to media objects posted by other users. An appropriateness setting icon or other input graphic may be displayed next to each tag. For example, the icon may include "+" and "-" buttons to indicate, respectively, that the user believes the tag is appropriate or inappropriate for the displayed media object. A statistical engine may collect the appropriateness entries for each media object to determine a measure of the object's appropriateness. For example, the measure may simply be the number of "+"s entered divided by the total number of appropriateness entries for each media object. The statistical engine associates each vote with the voting user to prevent "stuffing the ballot box," i.e., the statistical engine avoids counting multiple votes by a single user regarding the appropriateness of a tag to a media object.

The statistical engine can factor the appropriateness value into the cluster analysis to influence the relevance metric. For example, a tag with a low appropriateness value will be treated as irrelevant to other tags associated with the same media object (i.e., weighted as having a longer statistical distance).

According to embodiments of the present invention, metadata processing logic 118 may calculate an "interest" metric for each media object. Interest may be a function of user actions associated with the media object, including, for example, the amount of user-entered and/or user-edited metadata and/or access patterns for the media object. Alternatively, or in addition to those factors, interest may be a function of time, system settings, and/or the user's relationship with the publisher of the metadata.

Each of the above factors can be clamped to a maximum value set by the system designer, which is one way to weight each factor. Alternatively, or in addition, each factor can be weighted more directly by multiplying the factor by a weighting coefficient before any clamping. In any case, the factors (weighted or unweighted) can be summed together to create an interest score (i.e., a ranking). Weighting and clamping can of course be applied at a more granular level to the parameters that contribute to any of these factors (described below).

An interest score may be calculated for any media object of any grouping, e.g., a grouping from all groups that contain the media object, a grouping from one group that contains the media object, a grouping from areas of a website associated with the publisher of the media object, or a grouping from, for example, a set of media objects of the user that contains the media object being scored.

The amount of user-entered metadata may include, for example, parameters such as the number of tags, comments, and/or annotations assigned to the media object, and/or the number of users who have added the media object to their favorites/bookmarks. (Adding an audio media object to a user's favorites may include adding the media object to the user's playlist.) Alternatively, or in addition to those parameters, the amount of user-entered metadata may be user-dependent and include, for example, the number of users who have added tags, comments, and/or annotations to the media object, and/or the number of users who have added the media object to their favorites/bookmarks.

Alternatively, or in addition to those parameters, the metadata processing logic 118 can take into account the access pattern of the media object, such as the number of times the media object has been viewed (or played back) and/or clicked on, and/or the number of users who have viewed (or played back) and/or clicked on the media object or a tag associated with the media object. Whether the interest algorithm considers the user's action as a "click-through," or, conversely, the "viewing" or "viewing" of the media object can depend on the route the user took to access the media object, i.e., the access pattern. For example, a search for images assigned a particular tag can return multiple thumbnail images. The algorithm can consider the user clicking on a particular one of those thumbnails as a "click-through."

Conversely, for example, an image sent by email from one user to another user can be considered "viewed" by that user. In another example, when a user accesses a pool of image groups, the user's browser may present these images as thumbnails. The user can click on a thumbnail to "view" the image. Thus, the same action of clicking on a thumbnail can be treated as either a "view" or a "click-through," depending on the path the user took to reach the image, i.e., the access mode. Based on psychological observations, market research, or other factors, system designers may wish to treat certain access modes as indicating a higher level of user interest than other access modes and assign a higher weight to such access modes when calculating interest scores. As a perhaps more compelling example, if a user accesses an image by paying $10.00 to reach and click on a thumbnail, the system designer is more likely to assign a higher weighting factor to this access path than to an image accessed for free. Conversely, certain traffic sources, search terms, tag queries, or other precursors to displaying thumbnails may be determined to be associated with motivations inconsistent with higher interest, and thus the system designer may assign a relatively lower weighting factor to such access paths.

Additionally, metadata processing logic 118 may factor relationships between the publisher of a media object and the user (e.g., the user who entered the search query) into the score. The user may, for example, be a member of a private group (e.g., friends and family members, an interest group, or a social network) that is allowed to access the publisher's media objects, or a user listed in the publisher's contact list. Given that the similarity of interests between such a user and the publisher is likely high relative to other users, such a relationship may be weighted and added to the interest score to increase it.

The above functionality is an example of a "personalized" interest score. In general, the score can be based on the identity of the requester of the media object's interest score. (As used herein, reference to a "requester," or someone requesting a media object's interest score or ranking, refers to someone who explicitly requests the score or takes action through any access mode (e.g., entering a search query) that results in the media object being presented and the interest score being calculated by metadata processing logic 118, regardless of whether the score itself is provided to the requester.) Specifically, the score can be based on the relationship between the publisher of the media object and the user requesting the score.

In another embodiment, the personalized score of a media object associated with a user can be based on the number of media objects that have been assigned the same type of metadata (e.g., tags or favorites) by both the user and the score requester. Media objects can be associated with users by, for example, assigning metadata or making it public. For example, assume a first user and a second user store photo images in their online photo albums 100 and 200, respectively. The second user can search for images associated with a particular tag. Search engine 111 can return images stored in the first user's album. Metadata processing logic 118 can assign a score to the image that is an increasing function of the number of other images in the first and second user's albums that the first and second users have jointly assigned as favorites or tagged, based on the theory that this shared behavior serves as a predictor that the second user may be particularly interested in images in the first user's album that the second user has not yet "favorited" or tagged.

In another embodiment, the metadata processing logic 118 may calculate an interest score based on a location associated with the media object and a location associated with the user requesting the score for the media object. For example, the metadata processing logic 118 may indicate that a user is more interested in a media object if the location associated with the media object is associated with a particular user's residence (e.g., near the user's residence or in the same geographic area as the user's residence), or is associated with the residence of another user with whom the user has a predetermined relationship (e.g., a friend or family member), or is associated with a location that is associated with a threshold number of media objects that have been assigned metadata by the user (e.g., tagged or favorited).

In the latter case, for example, the metadata processing logic 118 may positively factor into the interest score of an image of the Washington Monument, for a particular user, the fact that the user has designated a large number of images related to the Washington, D.C. area as favorites. This assumes that metadata indicating a location in the Washington, D.C. area was associated with the image of the monument when the metadata was assigned to the image, e.g., by the publisher of the image or another user who entered the location via a tag field or a separate "location" field.

System designers can also set other interest score components. For example, some media objects may be treated as undesirable because they contain objectionable content, such as obscene images or promotions of competitor products. System designers can, for example, set up a score calculation that decrements the previously accumulated score by a predetermined score offset percentage, where the predetermined score offset percentage is the percentage assigned to media objects with "blacklisted" tags or other metadata. Because a media object can be associated with more than one blacklisted tag, the score offset value is selected as the maximum score offset value associated with those tags.

Another score component can also be considered. For example, the system designer can set up the score calculation to decrement the accumulated score by a predetermined percentage over the time since the media object was published. For example, this time decay can cause the score to decrease by 2% per day starting from the publication date. This and other measures can be used to prevent a "positive feedback loop" from occurring when ranking media objects by interest itself deviates from the results, causing the same media objects to be visited more frequently, thereby artificially increasing their interest scores.

The final interest score may be normalized so that, for example, the interest score always falls between 0 and 100 or 0 and 1. One way to achieve normalization is to divide the true score value by the maximum possible score value.

Search engine 111 allows users to search media objects in the database based on various metadata. For example, a user can perform a Boolean search for a tag in all media objects accessible to the user. Alternatively, a user can perform a full-text Boolean search for an item in a comment, annotation, title, or description. Media objects accessible to a user include, for example, public media objects, media objects within a group, media objects of friends/family members or members of another private group of the user, all media objects posted by the user, or media objects of the user within a user-defined collection.

The media objects returned from the search can be ranked according to interest. For example, in one embodiment, the media server can only provide the following media objects to the searching user for display: media objects with interest scores greater than a predetermined threshold, or a predetermined number of the highest-scoring media objects.

In response to a search using tags, for example, the statistics engine 109 can determine (based on a relevance metric) the tags (or other metadata) that are most relevant to one or more tags (or other metadata) in the search query. The media server 100 can return these most relevant tags (or other metadata) to the user at the client along with the retrieved media objects. If the relevance calculation results in two clusters of related tags (or other metadata), the media objects associated with these two clusters can be ranked (and displayed) in order of interest.

In the context of advertising, ads can be associated with their own metadata/keywords, such as "Rome Italy Hotels tourism travel" for an Italian hotel ad. Based on those associated keywords, ad server 122 can utilize a collection of related tags or other metadata and a relevance metric provided by the media server via web server 102 to determine which ads sponsoring the website are associated with predefined metadata/keywords that most closely match the collection of related tags. (The ad server can be a third-party server on network 112.) The ad server can then display the most closely matching ad to the user. For example, the ad server can display an ad for an Italian hotel on the user's client computer and a picture from the media server with the highly relevant tag "Rome Italy Honeymoon." In this way, the ad server utilizes the relevance metric and the collection of highly relevant tags to provide the ad that is most relevant to the displayed media object.

In another embodiment, in response to a search utilizing tags (or other metadata databases), the statistical engine may also prevent media objects assigned tags (or other metadata sets) whose relevance scores fall below a relevance threshold from being returned as search results.

It will be appreciated that the above description clearly describes embodiments of the present invention in conjunction with different functional units. However, it should be understood that any distribution of functionality between different functional units may be used without departing from the present invention. Therefore, references to specific functional units should only be considered as references to appropriate means for providing the functionality, rather than describing a strict logical or physical structure or organization.

The present invention can be realized in any suitable form, including hardware, software, firmware or any combination thereof. Different aspects of the present invention can be implemented at least in part as computer software or firmware running on one or more data processors and/or digital signal processors. The elements and components of the embodiments of the present invention can be implemented physically, functionally and logically in any suitable manner. In fact, functions can be implemented in a single unit, in multiple units, or functions can be implemented as a part for other functional units. Like this, the present invention can be implemented in a single unit, or can be physically and functionally distributed between different units and processors.

Although the present invention has been described in conjunction with certain embodiments, it is not intended to limit the invention to the specific forms set forth herein. Rather, the scope of the invention is limited solely by the claims. Furthermore, although it may appear that a feature has been described in conjunction with a specific embodiment, those skilled in the art will recognize that various features of the described embodiments may be combined in accordance with the present invention. Furthermore, various aspects of the invention described in conjunction with an embodiment may stand alone as an invention.

Furthermore, it will be appreciated that modifications and substitutions may be made by those skilled in the art without departing from the spirit and scope of the present invention.The present invention is not to be limited by the foregoing illustrative details, but is to be defined by the claims.

Claims (43)

1. An apparatus comprising: processor; A storage device for storing computer instructions, the instructions including: Logic executed by the processor for providing at least one image to be displayed to at least one user; and The processor executes logic for accepting input from multiple users regarding the at least one image, wherein the input includes at least two different types of metadata, and the metadata input by the users includes tags; The logic executed by the processor for searching an image using at least one set of metadata; The processor executes logic for ranking at least one image returned from the search after the search is executed, the ranking logic ranking the at least one image based at least in part on the amount of metadata of user input about the at least one image, and the ranking logic ranking the at least one image also based at least in part on one or both of the following: the frequency with which a first metadata set and a second metadata set are jointly assigned to the at least one image, and access patterns associated with the at least one image, the access patterns including the route taken by the user to access the at least one image; and The processor executes logic for determining the appropriateness of at least one of the tags for the at least one image based on appropriateness from at least one user input, wherein the ranking is at least partially based on the appropriateness of the at least one tag for the at least one image. 2. The apparatus of claim 1, wherein the at least two different types of metadata are selected from a group of metadata types including: tags, comments, descriptions, preferences, and annotations. 3. The apparatus of claim 2, wherein the tag includes location metadata. 4. The apparatus of claim 1, wherein the at least one image comprises a video image. 5. The apparatus of claim 1, wherein the apparatus is located in a server, and each user is associated with a corresponding client. 6. The apparatus of claim 1, wherein the user-input metadata is selected from a group of metadata types including tags, comments, and annotations. 7. The apparatus of claim 1, wherein the ranking logic further ranks the images based at least in part on the number of users who have assigned metadata to the at least one image. 8. The apparatus of claim 7, wherein the user-input metadata is at least partially based on the number of users who have designated the at least one image as a favorite. 9. The apparatus of claim 1, wherein the access mode is at least partially based on the number of clicks on the at least one image. 10. The apparatus of claim 1, wherein the access mode is at least partially based on the number of times the at least one image is viewed. 11. The apparatus of claim 1, wherein the ranking logic further ranks the at least one image based on the passage of time associated with the at least one image. 12. The apparatus of claim 11, wherein the elapsed time associated with the at least one image is the elapsed time since the at least one image was uploaded. 13. The apparatus of claim 1, wherein the ranking logic further ranks the at least one image based at least in part on the relationship between the publisher of the at least one image and the user who initiated the search. 14. The apparatus of claim 1, wherein the computer instructions further include statistical logic for determining a metric based at least in part on the frequency with which the first metadata set and the second metadata set are jointly assigned to the same image. 15. The apparatus of claim 14, wherein the statistical logic is logic for the following operations: Clusters of relevant metadata are determined based on metrics associated with multiple images; and Rank the associated multiple images within each cluster. 16. The apparatus of claim 14, wherein the first metadata and the second metadata are tags. 17. The apparatus of claim 14, wherein the statistical logic is configured to provide the first metadata and the second metadata as relevant metadata for display if the determined metric exceeds a threshold relevance value. 18. The apparatus of claim 14, wherein the computer instructions include logic for determining the suitability of the first metadata and the second metadata for the image based at least in part on suitability input from at least one user, wherein the statistical logic is logic for changing the metric based at least in part on the determined suitability. 19. The apparatus of claim 1, wherein the computer instructions include logic for determining the frequency at which at least one particular set of metadata is assigned to multiple images within a predetermined time period. 20. The apparatus of claim 19, wherein the at least one metadata set includes at least one tag. 21. The apparatus of claim 1, wherein the computer instructions include logic for determining the number of images to which at least one particular set of metadata is assigned. 22. A method comprising: At least one image is provided by a computing device to be displayed to at least one user; The computing device accepts input from multiple users regarding the at least one image, wherein the input includes at least two different types of metadata, and wherein the metadata input by the users includes tags; The computing device searches the image using at least one set of metadata; After the search is executed, the computing device ranks at least one image returned from the search, the ranking comprising: ranking the at least one image at least partially based on the amount of metadata input by the user about the at least one image, and ranking the at least one image at least partially based on one or both of the following: the frequency with which a first metadata set and a second metadata set are jointly assigned to the at least one image, and access patterns associated with the at least one image, the access patterns including the routes taken by the user to access the at least one image; and The computing device determines the appropriateness of at least one of the tags for the at least one image based on appropriateness from at least one user input, wherein the ranking is at least in part based on the appropriateness of the at least one tag for the at least one image. 23. The method of claim 22, wherein the at least two different types of metadata are selected from a group of metadata types including: tags, comments, descriptions, preferences, and annotations. 24. The method of claim 22, wherein providing at least one image and accepting input occur in the server, and each user is associated with a corresponding client. 25. The method of claim 22, wherein the user-input metadata is selected from a group of metadata types including tags, comments, and annotations. 26. The method of claim 22, wherein the ranking further includes ranking the images based at least in part on the number of users who have assigned metadata to the at least one image. 27. The method of claim 26, wherein the user-input metadata is at least partially based on the number of users who have designated the at least one image as a favorite. 28. The method of claim 22, wherein the ranking further includes ranking the at least one image based on the passage of time associated with the at least one image. 29. The method of claim 22, wherein the ranking further includes ranking the at least one image based at least in part on the relationship between the publisher of the at least one image and the user who initiated the search. 30. The method of claim 22, further comprising determining a metric based at least in part on the frequency with which the first metadata set and the second metadata set are jointly assigned to the same image. 31. The method of claim 30, wherein determining the metric further comprises: Clusters of relevant metadata are determined based on metrics of multiple metadata associated with multiple images; and Rank the associated multiple images within each cluster. 32. The method of claim 30, further comprising providing the first metadata and the second metadata as relevant metadata for display if the determined metric exceeds a threshold relevance value. 33. A non-transitory computer-readable storage medium including a computer-executable program, said computer-executable program being executed by a processor to: The processor provides at least one image to be displayed to at least one user; and The processor accepts input from multiple users regarding the at least one image, wherein the input includes at least two different types of metadata, and wherein the metadata input by the users includes tags; The processor searches the image using at least one set of metadata; After the search is executed, the processor ranks at least one image returned from the search, the ranking comprising: ranking the at least one image at least partially based on the amount of metadata input by the user about the at least one image, and ranking the at least one image at least partially based on one or both of the following: the frequency with which a first metadata set and a second metadata set are jointly assigned to the at least one image, and access patterns associated with the at least one image, the access patterns including the routes taken by the user to access the at least one image; and The processor determines the appropriateness of at least one of the tags for the at least one image based on appropriateness from at least one user input, wherein the ranking is at least in part based on the appropriateness of the at least one tag for the at least one image. 34. The non-transitory computer-readable storage medium of claim 33, wherein the at least two different types of metadata are selected from a group of metadata types including tags, comments, descriptions, preferences, and annotations. 35. The non-transitory computer-readable storage medium of claim 33, wherein providing at least one image and receiving input occur in a server, and each user is associated with a corresponding client. 36. The non-transitory computer-readable storage medium of claim 33, wherein the user-input metadata is selected from a group of metadata types including tags, comments, and annotations. 37. The non-transitory computer-readable storage medium of claim 33, wherein the ranking further includes ranking the images based at least in part on the number of users who have assigned metadata to the at least one image. 38. The non-transitory computer-readable storage medium of claim 37, wherein the user-input metadata is based at least in part on the number of users who have designated the at least one image as a favorite. 39. The non-transitory computer-readable storage medium of claim 33, wherein the ranking further includes ranking the at least one image based on the passage of time associated with the at least one image. 40. The non-transitory computer-readable storage medium of claim 33, wherein the ranking further includes ranking the at least one image based at least in part on the relationship between the publisher of the at least one image and the user who initiated the search. 41. The non-transitory computer-readable storage medium of claim 33, wherein the computer-executable program is further configured to: determine a metric based at least in part on the frequency with which the first metadata set and the second metadata set are jointly assigned to the same image. 42. The non-transitory computer-readable storage medium of claim 41, wherein determining the metric further comprises: Clusters of relevant metadata are determined based on metrics associated with multiple images; and Rank the associated multiple images within each cluster. 43. The non-transitory computer-readable storage medium of claim 41, wherein the computer-executable program is further configured to: provide the first metadata and the second metadata as relevant metadata for display if the determined metric exceeds a threshold relevance value.