After clarifying what I’m trying to do with a friend, he pointed out that a baseline version can be explained without resorting to web annotation. The end goal is to facilitate navigating information on the web by leveraging metadata to extract meaning from large hypertexts. As mentioned in the section Information Space and the Wiki Way [1,2], the nonlinear nature of hypertext turns readers into co-authors because they construct contours (perceived patterns of meaning) as they explore the hyperbase.
These contours constitute a semantic space in the reader’s mind, and “the” information space is a mesh network of the contours each reader constructs. It is not derivable from the hyperlink topology alone (see the essay linked above for more).
The thesis would contribute 1) a novel way to combine existing web standards, and 2) a method that demonstrates the value of such a combination. Because contours (meaning) will differ for different readers, the method would have to be capable of being personalized.
As an example, consider Wikipedia and Wikidata. The well-known “Wikipedia rabbit hole” is a case of readers constructing contours as they explore. Suppose that the reader in our example is a student exploring a rabbit hole related to a research project for which they have a citation graph. This graph could be obtained in RDF (using, for example, OpenCitations [3]) and connected to Wikidata to predict probable contours across Wikipedia. This would add value particularly where the reader is venturing into unfamiliar territory.
This problem arose in my last independent study about complex systems. It’s an interdisciplinary subject, and though my study began with work in economics and philosophy, I spent most of the semester wrestling with a body of work in evolutionary biology. I was lost in a forest of new concepts, sifting through many dead ends to find meaningful connections. If I had a tool to help me navigate, I could’ve achieved much more in the semester.
In this example, all the data mentioned is available and would be usable for a thesis. Think of this as the baseline version of what I’m trying to do. The sections below connect it to what we’ve been talking about all semester.
The previous example used Wikipedia, but it could have been any other website with metadata available in RDF (e.g. Project Gutenberg). Personal websites are useful both as places to store one’s own metadata (e.g. the citation graph from earlier) and as a hyperbase to share roadmaps with others. Thinking in terms of personal websites conveys the same social paradigm for online collaboration without having to concern ourselves with Solid Protocol or FedWiki.
Unlike a wiki, you have total control over your own website. If you wanted to connect my ideas to your own but lacked the time to read through my website, imagine the previous example but with an additional citation graph.
Web annotation lets anyone comment on your personal website. Like semantic web standards, the annotation layer spans the entire web. Plain text comments are useful for visitors of a specific webpage, but using annotation to add metadata lets people leave trails of breadcrumbs across the internet.
While working on the examples below, I discovered that public annotations are public domain and include metadata added by bots. It’s not in RDF, but bot annotations are uniform, so they could be converted to RDF and worked with as a dataset. User annotations could be anonymized and everything but the comments would be usable as RDF (after converting the hypothes.is json format to json-ld, which I am working on with a friend). This solves so many problems with regard to data availability.
Consider two examples on this website:
On The Beginnings Of Ownership, I have two annotations expressing the same thing. One is doing so in plaintext while the other is expressed as an RDF statement. Technically, it’s a plaintext comment that could be read as an RDF statement.
Here I have pairs of highlights for Age of Reason and Session 2.2 of Coining Reason. I want to link each pair and state in RDF that the two pairs are related. These texts are on the same website, but keep in mind they don’t have to be.
I’m working on doing this manually. By the time we meet on Wednesday, this paragraph should be updated accordingly.
For the remainder of the semester, I propose the following (or some subset, up to you):
What is out of scope for this semester:
[1] Dalton, Sean. “Navigating Information Space”, Unpublished literature review, 2023, https://illumenaturale.neocities.org/Navigating%20Information%20Space
[2] Bernstein, Mark, Michael Joyce, and David Levine. “Contours of constructive hypertexts.” Proceedings of the ACM Conference on Hypertext. 1993.
[3] https://opencitations.net/