Annotations are coming to the web
- 2. Annotations are coming to the web, and ... well, everything else. Annotating Everything Published on December 14, 2016 Kurt Cagle Founder and Chief Ontologist at Semantical, LLC Life is lived in the margins. Three thousand years ago, priests and scholars were adding comments about the validity of poems about the gods in copies of clay tablets along the edges, along with comments such as Hath-ur wrote this part. Most extant copies of the Illiad from the fifth century BCE have more marginal content than actual text, and creators in the various manuscripts of the bible (leading up to the masterpiece of the Book of Kells) deliberately kept wide margins not only to assist in binding the large, wide pages, but also to provide room for scholars, transcriptionists, and illuminators to add their own last words (or sketches) about a given paragraph or poem. With the advent of printing (and consequently the significant reduction of the costof producing a book), annotations exploded into multiple different forms - commentaries, citations, corrections, glossary definitions and similar notations. These were often collected after the publication of one edition of a work and appended to the next, either within the bodyof the text (typically in footnotes either on the page or in chapter or bookappendices). The Internet was, ultimately, built around the conceptof an annotation. A typical web page link has all the key pieces one needs for a (very simple) annotation: <ahref="/link/to/annotation"title="My comments about this annotation"> some inline text (the body)</a> As the web shifted from being a read-only medium to being read-write oriented, one of the first things to happen was the emergence of comments to content that was published, whether articles or images. Most such systems provided a body(the comment), the author (the person who posted the comment) and when the comment was made. Additionally, summary and title
- 3. metadata could be added. As mail moved away from being a stand-alone application to being a mostly web-based app, it also became more annotation oriented. Facebookand Twitter are, in the main, nothing but annotations annotating other annotations. A tweet or facebookpostis submitted, typically containing a bodythat might include a link (which gets transformed from an inline piece of text to a web URL), and subsequent comments made to that tweet or postare annotations to that post, or to comments on that post. Facebook content itself has a specific address, but comments generally do not. Twitter's retweet facility does much the same thing, but no distinction is made between a primary tweet and any retweets - they are each distinct objects. Indeed, at their core, most social networks follow one of those two models - Pinterest, Tumblr, Instagram, Snapchat, Wordpress. All of these are effectively annotations of annotations. Given this realization, the W3C is moving towards final release of the Open Annotation standard (OA). The core idea behind OA is that any resourcethat has a representation on the web can be the target of an annotation - a web site, a web page, a blog post, an image, a PDF document or anything else that is accessible on the web. The structure of an annotation is straightforward, though its implementation is perhaps a bit more complex (Figure 1):
- 4. Figure 1. Core structure of an annotation. The annotation has two distinct parts: its target (what the annotate is referring to) and its body (the content of the annotation itself). The target is represented by a URL (or web address) though it could also be conceptualIRI (see Annotations and Semantics, below). The body could be text (plain text or html) but could also be a link to an external document such as a blog post, data definition, picture, video, fragment from another web source, in short, anything that has an address on the web. Annotations usually have a motivation or purpose. An annotation may be a general comment, may be a formal or informal description or definition, may illustrate, indicate a change in status, or similar action. An annotation may also have multiple bodies or targets, For instance, an annotation may serve to identify the resumes of people in contention for a job: <http://www.annotationserver.com/anno1234> rdfs:label "Data Scientist Contenders"; a oa:Annotation; oa:hasBody "Joe, these are the candidates I think we should be looking at for the data scientist position."; dc:subject <http://www.example.com/posting/jobDataScientist1>; oa:hasTarget <http://www.example.com/resume/janeMarple>, <http://www.example.com/resume/deweyCheatem>, <http://www.example.com/resume/sherlockHolmes>, <http://www.example.com/resume/herculesPoroit>; oa:motivation oa:identifying; dc:creator foaf:tomSwift; .
- 5. In this case, the bodyis the content of the annotation, the subject provides a reference to an external position, the motivation for the annotation is to identify individuals, and the targets are the four individuals listed. The oa: is a namespace prefix identifying the various terms as being in the Open Annotation namespace, while dc:is Dublin Core, a common standard for identifying publishing metadata. An annotation can be thought of as a link on steroids. The above annotation creates associations between four different individuals through an external agency, something that can't easily be done on the web. Similarly, an annotation can also link to another annotation: <http://www.annotationserver.com/anno1235> a oa:Annotation; oa:hasBody "I like Jane Marple.She seems well qualified."; dc:subject <http://www.example.com/resume/janeMarple>; dc:subject <http://www.example.com/posting/jobDataScientist1>; oa:hasTarget <http://www.annotationserver.com/anno1234>; oa:motivation oa:assessing; dc:creator foaf:janeDoe; . <http://www.annotationserver.com/anno1236> a oa:Annotation; oa:hasBody "I have problems with Dewey Cheatem. His resume looks dubious."; dc:subject <http://www.example.com/resume/deweyCheatem>; dc:subject <http://www.example.com/posting/jobDataScientist1>; oa:hasTarget <http://www.annotationserver.com/anno1234>; oa:motivation oa:assessing; dc:creator foaf:mattSmith; . Here, there are two annotations, the first endorsing Jane Marple, the second raising questions about Dewey Cheatem. Both point back to the original annotation, and the subjects, in turn, are what were targets in the previous annotation. How Web Annotations Work One key difference between a link in an HTML document and an annotation is that a link is embedded within a document, and is auto-selected by what's within the associated <a> tag.
- 7. With annotations, on the other hand, the target - the thing being annotated - is indicated by a document located on a server, called an annotation server. Correspondingto this is an annotation client, which can be thought of as being like a news reader - it takes all of the annotations either associated with a given target or that corresponds to a given keyword or query. In the former case, what comes back are all of the available "comments" for a given document, while in the latter case, what comes back are suggestions about targets that satisfy a search query ("give me all annotations about cats and show what they are annotating"). You can also get conversations back - annotations that point to other annotations much like a layered comment feed looks at the end of most articles ( or Facebookresponses). Theclient app would then be responsible for presenting this information in a user-friendly manner. Conceivably there could be thousands or even millions of such servers. Public annotation servers (perhaps hosted by Google or Facebookor Twitter - or the next incarnations of these kinds of companies) would be able to aggregate commentary about the web in general, while private annotation servers might be run by investment banks or financial analysts, health care or life sciences organizations, political activists and so forth, that would be subscription only. Since concepts - things - can be referenced as targets, it's not hard to imagine an investment company annotating stocks, bondsorfinancial instruments (as well as regulatory structures), news organizations tagging specific watch topics or movie fans (and studios)tagging specific movies, TV shows or games. IMDB does some of this last one now (as does Rotten Tomatoes) but you could log into both the (hypothetical) IMDB or Tomato annotation servers and be able to plug directly into the information spacethat either of these services would provide. Since such annotation servers can also add tagging into established taxonomies, with enough such servers in place, you can create a semantic profile of a person, organization, thing, work of art or topic that makes it possible to query across all such servers universally, and can go a long way of providing a provenance trail to both ensure the legitimacy of specific information and how that information has changed over time.
- 8. Now, none of this is here yet. The Web Annotation Working Group has pushed the annotation model itself, its vocabulary, and the transport protocol to Candidate Recommendation status, meaning that you should start to see early prototypes ofannotation servers and clients begin to appear by mid-2017, likely with universities, research organizations and banks taking the lead here for internal applications. However, such services will likely go commercial within the next two or three years, and by 2020 we could see a new annotation explosion as more and more things (and even people) begin to develop their own annotation spaces. So, get out the sticky notes. Kurt Cagle is a writer, blogger, and owner of SemanticalLLC, a smart data company. He's covered with sticky notes.