Open Annotations on Multimedia Web Resources

Many Web portals allow users to associate additional information with existing multimedia resources such as images, audio, and video. However, these portals are usually closed systems and user-generated annotations are almost always kept locked up and remain inaccessible to the Web of Data. We believe that an important step to take is the integration of multimedia annotations and the Linked Data principles. We present the current state of the Open Annotation Model, explain our design rationale, and describe how the model can represent user annotations on multimedia Web resources. Applying this model in Web portals and devices, which support user annotations, should allow clients to easily publish and consume, thus exchange annotations on multimedia Web resources via common Web standards.

💡 Research Summary

The paper addresses the problem that user‑generated annotations on multimedia web resources (images, audio, video) are typically locked inside proprietary portals and therefore invisible to the Web of Data. To bridge this gap, the authors present the Open Annotation Model (OAM), describing its current beta specification, design rationale, and how it can represent annotations on multimedia resources in a Linked Data‑compatible way.

The authors begin by illustrating real‑world annotation scenarios on platforms such as Flickr and YouTube, where users add textual notes, tags, region highlights, and temporal markers. They argue that earlier standards like Annotea were web‑page centric, offered limited segment addressing, required proprietary protocols, and did not handle the ephemerality of web resources.

From a series of concrete use cases, four guiding design principles emerge: (1) An annotation is a qualified association between two first‑class web resources – a Body and a Target – each identified by a URI; (2) Both Body and Target may be of any media type, not just text; (3) Segment addressing must be supported for both Body and Target, preferably using standard Media Fragment URIs but extensible for more complex cases; (4) The model must acknowledge that web resources change over time, therefore timestamps and fixity information are required.

The baseline OAM data model consists of three primary RDF classes in the oac namespace: oac:Annotation, oac:hasBody, and oac:hasTarget. An oac:Annotation is itself an HTTP‑dereferenceable RDF document that links a Body resource (any URI, any MIME type) to one or more Target resources (also any URI). Optional metadata such as dcterms:created, dcterms:creator, dc:title, FOAF properties, etc., can be attached to any of the three entities, allowing separate creation times for the annotation, its body, and its target.

For segment identification, the model reuses the W3C Media Fragment URI specification (e.g., #t=10,20 for video time ranges, #xywh=160,120,320,240 for image rectangles). When standard fragments are insufficient—such as polygonal regions on high‑resolution maps or MPEG‑7 feature descriptors—the model permits custom constraints expressed as additional RDF resources linked via oa:hasConstraint. This extensibility ensures that even highly specialized multimedia use cases can be modeled without breaking the core specification.

Ephemerality is tackled by encouraging the inclusion of creation timestamps on each resource and by allowing references to specific versions or snapshots (e.g., via Memento or oa:hasSource). This prevents misinterpretation when a target page evolves after an annotation has been made.

Interoperability is achieved through a modular architecture: the baseline model is intentionally minimal, enabling any client to implement core functionality, while domain‑specific extensions can be added as needed. The specification deliberately avoids prescribing any particular client‑server protocol; instead, it assumes that annotations are published as dereferenceable HTTP URIs and discovered using standard web mechanisms (e.g., content negotiation, Linked Data crawling).

The paper also surveys related work, noting that various Annotea extensions (bookmark, topic, SVG‑based segment descriptions) attempted to solve parts of the problem but remained fragmented. OAM consolidates these ideas into a coherent, Linked‑Data‑first approach.

In conclusion, the Open Annotation Model provides a robust, extensible, and web‑centric framework for publishing, consuming, and sharing annotations on any multimedia web resource. By treating annotations, bodies, and targets as first‑class web resources, supporting standard and custom segment addressing, and embedding temporal metadata, OAM enables seamless integration of user‑generated commentary into the broader Linked Data ecosystem, fostering richer discovery and reuse across domains ranging from social media to scholarly digital humanities.