Aller au contenu principal

Web Annotation Policy

Purpose of This Document​

This document is intended to outline how LINCS uses the Web Annotation Data Model and for which purposes. It also serves as a reference for data stewards as they consider whether Web Annotations make sense for their data.

Web Annotations​

The Web Annotation Data Model is a standard developed by the W3C to enable the creation, sharing, and reuse of annotations across different platforms. The purpose of web annotations is to “convey information about a resource or associations between resources” (W3C 2017). For example, a comment on a webpage, a tag on an image or text, or a product review can all be represented in the form of Web Annotations.

In its simplest form, a Web Annotation is composed of a body and a target, and expresses the fact that the body is related to the target in some way. The target is the web resource that is being annotated, while the body is the referenced resource. For more details, see the Sources & Metadata Application Profile.

Diagram

In LINCS, the main purpose of Web Annotations is to annotate natural language text. For example, we could annotate the name Virginia Woolf (target) in the sentence Virginia Woolf was born in London on 25 January 1882 with the URI http://www.wikidata.org/entity/Q40909 (body), pointing to the Wikidata description of English author Virginia Woolf. Likewise, we could annotate “London” (target) with the GeoNames URI https://sws.geonames.org.2643743/ (body), etc.

Example Diagram – Person

Example Diagram – Place

Advantages of Web Annotations​

Web Annotations allow us to describe the annotation and the underlying process of its creation in a detailed way. For example, we can attach specific metadata to the annotation to keep track of who created it, when it was made, the specific purpose for which it was created. We can also attach additional metadata, e.g., to express a level of certainty.

We can use selectors to link the annotation to a specific fragment of the target object in a precise way. For example, we can select a specific string of text within a document, a portion of an image, or a specific frame within a video, and attach the annotation directly to this fragment. This makes the annotation very specific, allowing us to express complex relationships between the body and the target object.

Web Annotations are compliant with W3C standards, thereby simplifying sharing and reuse of the annotations. They are widely adopted and fully interoperable beyond the confines of a specific community (e.g., digital humanists, or the adopters of CIDOC CRM). The Web Annotation Data Model is well documented and supported by a variety of tools and platforms.

Web Annotations are machine-readable and can be used to express both annotations created by humans (including through large-scale crowdsourced projects) and those made by automated agents such as natural language processing tools.

Criteria for Usage​

Web Annotations make sense whenever most of the following conditions are met:

  1. There is a need to connect a certain resource (the body of the annotation) to a specific segment of a target object (e.g., a word in a document); for example, an entity representing a person to a mention of their name.
  2. There is enough information to describe each annotation in detail, i.e., by explicitly stating who made it, when, and for what purpose;
  3. There is a need to attach the annotation to a specific subpart of the target object, e.g., a specific word or sentence in a text;
  4. None of the alternate solutions from CIDOC CRM described below are able to express the annotation in a satisfactory way.

Based on the needs on the dataset, only some of these criteria may be met. The choice whether to use web annotations is up to the data stewards, with the support of the LINCS team.

Possible Alternatives​

Web Annotations can be used to express very specific and precise references, but at the cost of significant added complexity. Moreover, depending on the specific usage, the Web Annotation Data Model may not always be easy to integrate with the CIDOC CRM ontology adopted by LINCS. For these reasons, when deciding whether to adopt web annotations, it is important to consider the alternatives listed below, which could be more appropriate depending on the specific project and available data.

P67 Property​

The CIDOC CRM P67 refers to property expresses the fact that “an instance of E89 Propositional Object makes a statement about an instance of E1 CRM Entity” (CRM SIG, 2025). For example, the Wikipedia article titled “Virginia Woolf” refers to English author Virginia Woolf, the city of London, King's College, and the book Orlando, among other entities. Hence it would be correct to express any of these relations through P67. This property is simple to use but quite vague, because it can be applied to any kind of reference.

P129 Property​

The CIDOC CRM P129 is about property expresses the fact that “an instance of E89 Propositional Object has as subject an instance of E1 CRM Entity” (CRM SIG, 2025). P129 differs from P67 in that it describes the primary subject (or subjects) of the Propositional Object. For example, the Wikipedia article titled “Virginia Woolf” is (primarily) about Virginia Woolf. It is not (primarily) about London, King's College, or the book Orlando. P129 is less generic than P67, but it is limited to a specific subset of references (those about primary subjects).

Attribute Assignments​

CIDOC CRM allows the use of E13 Attribute Assignment to describe “the actions of making assertions about one property of an object or any single relation between two items or concepts” (CRM SIG, 2024).

Attribute assignments can be used to express who made a certain assertion and provide additional information about it. For example, it is possible to express the fact that the statement “the Wikipedia article titled Virginia Woolf is about the English author Virginia Woolf” was made by the authors of the present document in August 2025.

By combining attribute assignments with P67 or P129, we can support multiple and potentially conflicting references from the same text (or media object) identified by different people (or automated tools) at different points in time. It is also possible to express the type of each P67 reference by using the property P67.1 has type.

Similarly to Web Annotations, attribute assignments add some complexity to the way the data is modelled, but they have the advantages of being fully supported by CIDOC CRM and not introducing any additional dependencies.

Use Case Examples​

This section provides examples of use cases for web annotations, along with the recommended solutions based on the specific needs of each dataset.

Use CaseRecommended Solution
My dataset includes a set of text documents tagged with the primary subject of each document. I care about references at the document level, without going deeper into the document structure or contents.CIDOC CRM property P129
My dataset includes a set of text documents tagged with the names of people and places that are mentioned in each document. I care about references at the document level, without going deeper into the document structure or contents.CIDOC CRM property P67
My dataset includes a set of text documents. I want to keep track of the primary subject of each document, and also any other entity that has been identified in the document. I care about references at the document level, without going deeper into the document structure or contents.P129 for the primary subject, P67 for the other entities
My dataset includes a set of text documents which have been annotated with their primary subject, but multiple annotations have been made by my institution in previous decades, and these are sometimes conflicting with each other. I want to keep track of the different annotations, but I don’t have much information about them (just author and date). I care about references at the document level, without going deeper into the document structure or contents.P129 with attribute assignments
My dataset includes a set of text documents which are being annotated through an ongoing crowdsourced project, generating tags that are attached to individual words or sentences. I want to store the full contents of the documents along with any tags added by individual users to each subpart of the document. I want to keep track of who made each annotation, when, and for what purpose. I intend to share the annotations with other projects for later reuse.Web Annotations
My dataset involves systematically identifying named entities such as people and places and marking them up according to the guidelines of the Text Encoding Initiative by using a tool such as LEAF Writer. I would like to share these annotations with others.Web Annotations
My dataset includes a set of text documents which have been annotated by a natural language processing tool, such as NERVE, on a large scale. I want to store the full contents of the documents along with any tags generated by the tool. I want to keep track of who made each annotation and when, and also the level of annotation confidence indicated by the tool. I intend to share the annotations with other projects for later reuse.Web Annotations

Current Usage in LINCS​

The following table lists the current usage of web annotations in LINCS datasets. Exploring the datasets can be useful when deciding whether to adopt web annotations. Current data can be accessed by running this query on the Fuseki triple store.

DatasetUsage
Historical Canadian Persons70,000+ annotations
Orlando40,000+ annotations
Heresies1,400+ annotations
Cabinet Conclusions400+ annotations

References and Resources​

Document Details​

Version: 1.0

Authors: Dani Metilli (University of Guelph), Natalie Hervieux (University of Alberta), Susan Brown (University of Guelph)

Contributors: the LINCS Project team

Last Updated: 2025-08-13

Released: 2025-08-13