Enabling Accessible Knowledge

The Berlin Declaration on Open Access to Knowledge in the Sciences and Humanities is one of the milestones of the OA (Open Access) movement [2]. Today, the push for more OA publishing is widely embraced and celebrated by many, including; researchers, governments, universities, libraries, archives, and funding agencies. The OA strategy is intended to address issues on standards, quality, stability, and transition of scholarly research. While there political and social mandates with best practices, there are no explicit technical mandates for systems and technologies (e.g., Web semantics) to carry out the OA mission.

In order to further open up the market for services based on public-sector information [3], the European Commission PSI Directive — a digital agenda for Europe — recommends high-demand datasets from libraries and archives be published in machine-readable and open formats (CSV, JSON, XML, RDF, etc.) to enhance accessibility, and described in rich metadata formats and classified according to standard vocabularies [4], and to facilitate re-use, public sector bodies should [...] make documents available [...] at the best level of precision and granularity, in a format that ensures interoperability [5]. Here again we observe well-intended recommendations without concrete mandates on technology or systems on precisely how research documents should be shared.

There are notable initiatives by public sector bodies which take on holistic approaches to build a sustainable ecosystem around data that supports social, economic and political impact [6]. The recommendations lean towards making reusable and machine-friendly formats available, in parallel to their PDF equivalents.

In 1999 article, Practical Knowledge Representation for the Web stated that the lack of semantic markup is a major barrier to the development of more intelligent document processing on the Web, and meta-data annotation of Web sources is essential for applying AI techniques on a large and successful scale [7], which at large, remain as open issues today.

The Web Science community continues to conduct research tracks, workshops, and challenges, in a manner that even the call for papers of top-level Web conferences request research documents that are neither machine nor Web-friendly. There remains strong demands for submissions to be solely in Word or PDF formats, meanwhile requesting strict adherence to the dimensions and limitations of a printable page. Unfortunately, research articles which discuss the discoveries about the Semantic Web, semantic publications or linked science are locked inside well-known data-silos: Word and PDF, and thus unnecessarily challenging to extract information, or conduct pattern recognition from within. The community is however becoming more aware of the public discussions around these issues in recent years, and considering to make efforts towards improving the state of research publishing within its own field.

For academic reviews, the Semantic Web Journal adapts an open and transparent review process with publicly available responses on the site Assigned editors and reviewers are known by name, and are published together with accepted manuscript [8].

It should be noted that, there are important achievements in Web Science that contribute to the future of scholarly communication, for example, vocabulary and ontology engineering (e.g., PROV-O, OPMW, Wf4Ever, SPAR, SIO, DDI-RDF, QB), tool-building, conceptual and architectural designs. The application of these discoveries and efforts are however absent in research articles.

In the publishing and journal space, there are different business models in place: from traditional publishers charging the readers and institutions per article, open-access publishers charging per publishing researcher, to one-time membership fees. While these systems provide varying options for research publications, the commonality is that the research documents are disjoint from one another at the lower data level, e.g., variables, hypothesis (as opposed to basic metadata). By in large, research knowledge is packaged with the intention of being printed on paper. Hence, the requirements of the publishers and journals from (Web) conferences are that, submissions foremost comply with printability. There exists digital journals, which have an open policy for the submission format, but those are few and far between.

Our strategy is as follows (adopted from StratML):

We use the definition of the term accessibility as the availability of knowledge to as many people as possible by applying universal design principles. Where applicable, the Web Content Accessibility Guidelines (WCAG) 2.0 [9] should be applied to make the content more accessible.

The second dimension to accessibility we adapt is open access: publications and accompanying research data should be discoverable, accessible, and reusable by humans, as well as machines, as the content becomes available. This is preferably without having to go through third-party proxies or services.

We use the definition of the term usability which incorporates usefulness and learnability of available knowledge. Here we seek a system in which the consumer can efficiently and with satisfaction use and interact with the tools. For example, having only the options visible that are applicable in context of a task, the system properly communicates with the author as to what is going on. In the case of learnability, the document should aim to provide interactions to help the reader get a better understanding of the content. The typography of the article matters a great deal under different devices, and therefore should be handled accordingly for optimal readability and legibility.

It is important to have a system that adapts user-centred design processes in order to allow users (both content producers and consumers) to intuitively interact with the system. This is so that researchers do not focus on the user interface, but that they are seamlessly engaged in their work. While actual experiences may not be completely designed, they can be supported with affordances.

It also follows that, building a ubiquitous user experience across different devices (i.e., customizing the user interface and interactions based on the characteristics of user’s media device) would be preferable to accommodate different access points, and learning styles.

Creating a responsive design and interactions between the user and the system can provide an opportunity for the user to learn by doing e.g., changing the parameters of an algorithm’s input and observing the corresponding graph output.

Evolvability can be seen as an attribute or a quality of a system, such that the system can survive in the long run. This is especially important when we want to have different publishing systems to be interoperable over time based on new needs. To be more specific, we are seeking to employ technology stacks that are likely to handle extensions and accommodate diverse scenarios for knowledge acquisition in the future. While it is not always clear or easy to predict which technologies will persevere in the future, we can pursue features like simplicity, flexibility, decentralization, interoperability, and tolerance. We are in agreement with Tim Berners-Lee's Web architecture commentary on the Evolution of the Web [10].

We propose an acid test in order to verify, approve, or test the openness, accessibility and flexibility of the approaches for research publication. This is along the lines of a test to distinguish gold from base metals. This test does not mandate a specific technology, therefore the challenge can be met by different solutions. It is intended to test the design philosophies so that different approaches may be closer to passing the independent invention test.

Challenge

1. Alice makes her research document available on the Web with the research objects available at fine granularity e.g., variables of a hypothesis.
2. Bob wants to precisely refer to and discuss Alice’s research objects from his own research document.
3. Carol is interested in discovering research documents in the wild, which contain a pattern of reusable research objects like the ones in Alice, Bob or others’ documents e.g., using the follow your nose type of exploration, searching, or querying against a service.
4. Dan learns by interacting with the components inside the research document e.g., changing the parameters, rerunning an experiment.
5. Eve visits Frank, and displays Alice’s article on one of his devices, and prints out a copy.

There are some assumptions (adapted from the Social Web Acid Test - Level 0):

Assumptions

• The interactions have at least two different tool stacks.
• The interactions are both human and machine-friendly.
• All interactions are according to published open standards, with 1) no dependency on proprietary APIs, protocols, or formats, and 2) no commercial dependency or priori relationship between the groups using the workflows and tools involved.
• All interactions are possible without prior knowledge of the user’s environment or configuration (within reason).
• There is no spoon.

Historically speaking, format wars occur between incompatible formats which compete for the same market e.g., AC versus DC. In that sense, we consider PDF and HTML formats to be competing in a system for knowledge discovery and acquisition. We consider PDF as the predominant format perceived based on historic precedent, and HTML as the challenger with disruptive potential. Let us have a look at some of their characteristics.

Donald E. Knuth, the inventor of TeX, considers it to be a program written using the literate programming approach. LaTeX, which is widely used in scholarly articles is a collection of TeX macros. PDF is usually generated from LaTeX.

PDF is intended for displaying and storing, and generally self-contained. However, it is not intended to be editable as the formatting instructions are no longer available. PDF is inherently layout oriented, and it is an optimal format for printing. Publishers tend to require a fixed layout with typographical guidelines for research documents, hence research articles in PDF are normally destined and designed for printing.

In order to facilitate metadata interchange, an XMP (Extensible Metadata Platform) package in the form of XML (most commonly serialized as RDF/XML) may be embedded in PDF (ISO 16684-1:2012). Otherwise, semantically useful information is not preserved when PDFs are generated, and makes it difficult to go back to source content format.

HTML is both a content format for encoding information, and a document format for storing information. HTML is the most used and open international standard (ISO 15445:2000). In our discussion, we consider XHTML as part of HTML (barring the differences in the W3C specifications).

HTML is easy to reuse, exploit, process, and extend. HTML does not have a hard page limit, so it can reflow to fit different displays. HTML has ubiquitous support across devices.

If HTML is used as the original format to record the information, a PDF copy can be generated on demand, using the specifications of the user. With an accompanying CSS for the HTML, desired layout guidelines can be achieved. Converting from PDF to HTML+CSS while possible, the precision rests on rendering the view, as opposed to creating a structurally and semantically meaningful document.

Offloading fundamental research knowledge via PDF may be preferable or convenient for some, however it comes only at the cost of losing access to granular information. Thus, PDF’s primary focus; the presentation layer leads to legacy documents that are not reusable in the long run.

In the survey conducted by Elsevier, PDF versus HTML — which do researchers prefer? [14], authors report more than 65 percent said they thought there would be a shift towards HTML use in the future.

In order to preserve and make information more accessible from different devices, the storage format choice should be based on flexibility and evolvability.

Table [Comparison TeX HTML] compares the TeX and HTML stacks.

Linked Research [15] is a proposal for Web researchers to use the technologies in the native Web stack to access, share and discover knowledge. The Call for Linked Research [16] aims to encourage the “do it yourself” behaviour for sharing and reusing research knowledge. The workflow template for Linked Research is intended to enable better discovery of research objects, and improved user experience and access to research knowledge.

Linked Research Workflow

1. URI Ownership: Publish your research and findings at a Web space that you control. This is a fundamental step to control and have responsibility over ones own research. It is inline with the self-archiving act. This serves the purpose of maximizing its accessibility, usage and citation impact.
2. Web Standards: Publish your progress and work following the Web standards as well as Linked Data design principles. Create a URI for everything that is of some value to you and may be to others e.g., hypothesis, workflow steps, variables, provenance, results. Ensure that the research is cost-free and available via open access. Use a liberal license to promote reuse.
3. Knowledge Acquisition: Reuse and link to other researchers URIs of value, so nothing goes to waste or reinvented without good reason. It also fosters a navigable network of information, as well as information to flow between research documents.
4. User Experience: Build towards a strong user experience. Use screen and print stylesheets. Create a copy of a view for the research community to fulfil organisational requirements. Design interactive user-interfaces for improved communication and education.
5. Announce: Announce your work publicly so that people and machines can discover it.
6. Open Feedback: Have an open comment system policy for your document so that any person or machine can give feedback.
7. Enable: Help, encourage, and motivate others to do the same.

We believe that our Linked Research approach (see Linked Researcher for partial implementation) qualifies under enhanced research communication, and invite the community to verify against the proposed framework for article evaluation in The Five Stars of Online Journal Articles [17] or other evaluations as appropriate.

One of the virtues of the Web is that it enables people and societies to communicate and transfer knowledge with everyone. In the case of scholarly communication, anyone in the world who can access the Web can consume the available knowledge, directly from the researcher or a Web space in which they control and trust. This lowers the barrier for knowledge dissemination — Web for All — without limiting its acquisition only to the privileged, regardless of location, or their ability. This is in fact at the heart of W3C’s mission to lead the Web to its full potential.

By enabling researchers to explore a body of knowledge in different ways, we cultivate and accommodate diverse learning.

One way of accomplishing this would be empowering the consumers to know which, and to what extent the ideas influenced a scholarly article i.e., making it possible to conduct a follow your nose type of exploration to further traverse the contributing ideas. The effect of linking ideas is that we can trace their history, understand their evolution, predict their direction, and identify missing areas. This is a view of ideas, discoveries, and innovation in motion, and it is in contrast to observing the ideas in isolation without accrediting sufficient context as to how they may have emerged. The exploration may be anywhere from navigating to related research objects, interactive analysis of the corresponding ideas, to academic references.

This article was influenced by the Connections TV Series and book [18] by James Burke.

A fundamental requirement for authors is to be able to create documents which will work on different layers i.e., content/structure, presentation, behaviour. The tooling should pre-emptively enable the progressive enhancement strategy to emphasize on the creation of research documents which are accessible, semantically rich, accommodating different views, and interactive. This is in contrast to singular or binary documents which pre-package all the material predominantly into a single view layer.

The modular approach allows different layers to co-exist, be extended and evolve over time. For example, if a research fragment is globally identifiable on the Web and its description is looked up, it can be discussed in a future document with different presentation and behaviours than the original document.

There exists a diverse set of vocabularies and ontologies, which are well suited for use in research articles. Where applicable, exploiting their use can foster better discovery and reuse. Researchers would need to have the means to select and apply the terms, which they find appropriate to describe their work. The baseline may include: SKOS (Simple Knowledge Organization System), FOAF (Friend of a Friend), DC Terms (Dublin Core Terms), SIO (Semanticscience Integrated Ontology), SPAR (Semantic Publishing and Referencing), PROV-O (Provenance Ontology), OPMW (Open Provenance Model for Workflows), RO (Wf4Ever Research Object), Disco (DDI-RDF Discovery), QB (RDF Data Cube), SIOC (Semantically-Interlinked Online Communities), and OA (Open Annotation).

Generic external resources like images, graphics and browsing contexts, as well as video, audio, and mathematical expressions should be embeddable in documents to advance the reuse of independently composed objects.

The W3C Web Components are reuse-based approach to defining, implementing and composing independent components. The component model allows for encapsulation and interoperability of individual services, resources or modules in HTML.

With the application of embedded content and components, indivisible research objects can be purposed to 1) exist on their own to 2) be referenced (or imported) from external resources. For example, a research article may want to show a regression analysis figure by simply referencing and including that foreign resource, as opposed to making a copy of it.

The need to provide components for interactive education arises to fulfil the need for better communication and knowledge transfer between the producer and consumer. By providing interactive elements in relation to the discussion, the reader can get a deeper understanding of the work than the written form alone.

Up and Down the Ladder of Abstraction presents an excellent demonstration of using interactive controls to teach about an algorithm to keep the car on the road.

Implementing these type of interactions are essential to user experiences in which the consumer feels good using the system since they are learning and accomplishing their tasks.

Citations in scholarly articles play an important role as they make a statement, which associates one document or its fragment with what is referred. As citations are vital to creating awareness, interlinking, dissemination, exploring and evaluating research, the mechanism to accomplish that should take into account at fine granularity. It follows that, citations are not merely connections at the document level, but can serve to refer to individual research fragments, from concepts, hypothesis, workflow steps, claims, evidence, arguments, to any annotation declared in other research articles.

We hypothesize that the reusable bits of information that are publicly accessible will naturally give birth to quality metrics which can be directly measured towards the research impact factor. This is akin to the Web economics to hyperlinking documents, where the quality and quantity of backlinks tend to indicate the importance of the cited work. Linking is the nuts and bolts of information for centuries, and it holds true for information shared on the Web.

A feedback system where individual responses to the research article can be referenced and archived is pivotal to strengthening the body of knowledge, its interpretation, and the questions it raises. In order to foster diverse readers, the functionality should welcome academic reviews, as well as comments and annotations by anyone in the spirit of open science, and machine-generated responses. Depending on the context, it may be appropriate to integrate a consensus system (e.g., up or down) to encourage well-written or approved responses to be more visible democratically. With the adoption of attributed peer reviews, responses can be rigorous and objective as they can be, while retaining transparency and identity of the reviewers, e.g., using decentralized authentication. This additionally creates an opportunity for the research community to carry out an open dialogue which will then retain substantial knowledge for the future simply by micro-contributions. Thus, efforts should be made to capture information from the readers in a machine-friendly way to allow data-mining operations. Being able to refer to responses would also mean that, a response in one document may refer to a response in another document, irrespective of their location on the Web.

Where applicable, the scientific process should be captured that is machine-readable, encompassing workflow and provenance level data. This type of information plays an important role in reproducibility, and foster trust towards the steps detailing the whole process.

It is a given that, publishing a Web document does not stop the authors from continuing with updates even after its announcement or submission to conferences. One way of handling academic reviews (or general feedback) for a specific version of the paper is to retain each changed copy and have sufficient annotation as to what was changed, and have links between the copies. This can also be complimented with RFC 7089 (HTTP Memento), reflecting a frozen prior state of the original resource. By submitting a digitally signed state of a document, it can fulfil submission requirements.

While complete reproducibility may not be possible, efforts should be taken to indicate or demonstrate the essential components of the experiments. Where applicable, this can be anywhere from allowing a (human or machine) consumer to discover the resources that are integral for reproduction, to creating an executable paper environment where; some code can be rerun, observing the effects of changing parameters, allowing sample data uploads, or other explorations which could be conducted inside lightweight and portable VM environments.

If the community shifts its focus towards soundness, and completeness (where possible) in papers, the underlying elements that make up the workflows can have automated tests and verifications. Thus enabling machines to do some of the mundane checks on behalf of the reviewers.

Demonstrating reproducibility inside research articles also helps the reader to experiment and learn in context of the material. For example, the Executable Papers initiative at Elsevier Connect [20] demonstrates a Web based system to encourage reproduction of computational results in research articles, in order to help readers gain deeper insight.

In order to accommodate for content availability which depends on external services, internal and local caching methods can be used to retain the last retrieved version of the information in the research article or viewing interface. In the event that remote information can not be embedded in the research article, the last cached response can be used as a fall back option. Otherwise, a new copy or reference can be called dynamically. Where applicable, caching can substantially improve the document’s load performance.

It is evident that Cool URIs don’t change. There is a social expectation to ensure identifier persistence. For the integrity of the Web, it is important that the information persists as identified well into the future. This is irrespective of the identifier scheme that is used to refer to a piece of research publication or data. Thus, the individuals or groups which are the owners and maintainers of the identifiers need to be committed and responsible to their persistence. This includes adopting policies to ensure that the representations of the work may be preserved by another management, for example, if the institution dissolves, or if the individual(s) pass away. Existing identifier schemes and services can be employed for persistence, e.g., URI, DOI, PURL, w3id.

In order to ensure the permanent or long-term preservation of research articles, there should be a mechanism to commit and submit the work to a trusted location. If the location of the archival copy is somewhere other than the identifier that is used to reference the canonical copy (i.e., on the author’s site), then well-known archival institutions should be used e.g., the Internet Archive, arXiv.org, ZENODO (which is in same cloud infrastructure as research data from CERN’s Large Hadron Collider).

Updated 2018-12-29: Linked Researcher is now called dokieli (Article: Decentralised Authoring, Annotations and Notifications for a Read-Write Web with dokieli, Source code: github:dokieli/dokieli).

Linked Researcher is a POSH editor. It contains a template and themes for well-known layouts for research documents in Computer and Information Science, e.g., LNCS and ACM. Documents which use the provided themes can be viewed on screen devices, as well as printed to paper, or output to PDF or PS files. For instance, the canonical reference to this document (i.e., https://csarven.ca/enabling-accessible-knowledge) while semantically annotated and accessible via screen devices, can also be printed.

The canonical copy is annotated using the RDFa notation, where all important concepts, claims, arguments, evidences, and along with others, have an URI assigned to them. For example, the argument of our discussion (from earlier: adopting native Web technologies from ground up) can be precisely cited using the following URI: https://csarven.ca/enabling-accessible-knowledge#argument.

It is possible to include or exclude sections for specific media types. For example, researchers can decide to omit a paragraph in the camera-ready version for the conferences, while keeping it visible for screen devices.

Inline-charts or sparkline are intended to be concise and located where they are discussed in the text. They are treated as datawords carrying dense information. Edward Tufte describes sparkline as small intense, simple, word-sized graphic with typographic resolution [21]. For example, the GDP of Canada claimed by the World Bank Linked Dataspace is on the rise despite the market crash in 2008. This brief demonstration of Linked Statistical Data Sparkline, 1) compliments the supporting text without breaking the reader’s flow, and 2) provides an opportunity for the reader to investigate further by clicking on the data-line to access the source.

Statistical analysis from remote locations can be embedded in local documents, with the possibility to trace its source for further information to foster trust between publishers and consumers. Linked Statistical Data Analysis [22] offers a way to run analysis and explore distributed linked statistics, with accompanying provenance data. Figure [1] is an SVG scatter plot of mortality rate and CPI from the World Bank and Transparency International Linked Dataspaces respectively. The figure is embedded in this document by referencing its external object:

The only requirement here is to use HTML’s generic external content element, and refer to an existing resource.

LSD (Linked Statistical Data) Cube Designer is a Web service with an user-interface for researchers to design their own statistical data cubes. The statistical objects that are part of the cube model are derived from and refer to Linked Statistical Dataspaces. It aims to lower the barrier for searching and reusing existing linked statistical components. LSD Cube Designer offers a way for researchers to search for existing cube component concepts from existing statistical linked dataspaces, select most suitable dimensions, measures, and attributes for their data structure and components, and then export the cube’s structural information (in RDF Turtle format). We can interact with this remote application in Figure [2]: