Linked Research: An Approach for Scholarly Communication

Open Access: To better enable scientific progress, research work - from raw results to completed analysis - should be widely available and accessible. Maximising reach in this way increases the likelihood of further scientific progress and new discoveries [6]. Ideally this is achieved by way of free-of-charge access and reuse of all material, including code, media, source text, data and metadata, so knowledge may be distributed regardless of wealth, social status or geographical location of either the researcher or the consumer. This requirement is to some extent borrowed from the well known Open Access [7] movement (though aiming to avoid the additional restrictions often imposed by publishers such as Article Processing Charges or Gold Access fees).

Support for decentralised authoring and publishing: Research is commonly published through centralised services today. Publishers of journals or conference proceedings act as gatekeepers to work, which may only be available behind a paywall, and is required to meet criteria specific of the publishing house before it is published. Repository services allow researchers to upload their work directly to a database, but may require proprietary processing, account creation, and may distribute work only to other users of that service. If such a service shuts down, links to the work and any additional value extracted from it by the service are lost if there are no other copies. In both cases, control over access and distribution of the work is taken out of the hands of the original author, and the author may additionally be required to give up their rights to distribute their work by other means. We propose that it must be possible for authors to opt out of such functional information silos [8] and publish work on their own terms, in a storage location controlled by themselves or someone they trust, without being excluded from the wider community [9].

Access control and attribution: In many fields, researchers are discouraged from publishing unfinished, inconclusive or ongoing work, or publishing outside of ‘official’ venues by the risk that their work will be plagiarised or reused by another without permission or due credit. To facilitate open scientific discourse, technologies for publishing must afford a trust that sharing work will not disadvantage the author. Thus the author must have means to unambiguously attach their identifying information to their own work, and selectively grant and revoke access to others during the research process. Presently there are many services offering identifiers to creators of scholarly work: the Library of Congress Linked Data Service for personal and family names and corporate bodies, Virtual International Authority File (VIAF), International Standard Name Identifier (ISNI) and the Getty’s Union List of Artist Names (ULAN), ResearchGate, Google Scholar, Scopus Author Identifier, ResearcherID, and Open Researcher and Contributor ID (ORCID). These systems are minimally, if at all, interoperable without human intervention, so creating multiple profiles is common practice. Some of these systems interoperate with – not via open standards but by proprietary solutions – certain of the others, e.g., ORCID can import data from ResearcherID. It is part of our vision to empower researchers to choose an identity provider they trust or participate with their own Web identifier, which is globally recognisable and unique.

Provenance and accountability of information: Trust and confidence in data and results can be fostered by enabling reproducibility of experiments required to validate, and demonstrating a coherent explanation of the derivation of conclusions, and the agents responsible; however this tends to be a difficult challenge under current practices. Capturing scientific processes, workflows and data origins in a machine-readable way, and exposing Research Objects [10] will further improve the utility of the scholarly contributions.

Persistence and long term preservation: Research results must be made available in such a way that access to it is reliable and consistent over time, and irrespective of place. Studies show that significant number of referenced Web resources are not preserved or are at risk of disappearing [11]. Broken links, missing data, authors who move institution, organisations which close or merge or rename, and regimes which censor can all hamper access to and reuse of knowledge. A Linked Research ecosystem strives for persistent, long-term access to information as a technical and social norm by incorporating repositories in the preservation process.

Sharing and social interactions: Authors largely rely on third parties, including publishers and the media, to distribute and raise awareness their work. However, increasing numbers of academics are making use of social media to this end, and sharing on the Web has become a normal part of daily life for many [12]. This is valuable to make use of as part of Linked Research.

Commentary and feedback: The academic peer-review process has been used to judge whether work makes contributions to the field. Studies are beginning to show the unreliability and elitist nature of this process [13, 14] and thus we call for transparency and broader participation in research discourse. It should be possible and expected for anyone to review academic work, and for this review to be attributed and recognised as a contribution of the reviewer in its own right. Moreover, research has shown that transparency in the peer-review process may be an indicator of the quality of peer-review [15].

Human and machine-readability: Scholarly content, mostly published in binary form (e.g., PDF, Word), is currently not machine-readable to the extent that automatic post-processing is possible. Desirable modes of post-processing include sorting, aggregation and more accessible display of information, to improve the ease of use by humans. Many aggregation and indexing systems require manual input of metadata (or careful scraping by proprietary software), and authors undertake this labour, often repeatedly across different systems, for the pay-off of improving findability of their work. This labour can be avoided by generating metadata from article content and publishing it according to a standard syntax and semantics.

Integration of rich semantics: Within and around the prose, scholarly communication comprises rich semantic structures:

• the structure of the content, e.g. in sections, examples, definitions, theorems, notations;
• the argumentation, e.g. premises, assumptions, deductions, conclusions;
• facts and relationships between entities expressed in sentences;
• all kinds of data (examples, experiment results).

The Joint Declaration of Data Citation Principles [16], endorsed by over 100 organisations, explicitly calls for data to be treated the same as papers in the scholarly ecosystem, and for machine readability of the data. Similarly, the FAIR Data principles for findable, accessible, interoperable and reusable data encourage the semantically interoperable reuse of data.

Making this level of semantic structure available to machines for querying (and ultimately automated reasoning) facilitates development of intelligent support services for scholars, such as recommendation, search and comparison for related work.

Globally unique identification of entities: Our premise: Any resource of significance should be given a URI [17] and The global scope of URIs promotes large-scale "network effects": the value of an identifier increases the more it is used consistently [18]

A research document contains references to things, from abstract concepts to real-world objects; anything which may be referenced from a publication (people, methods, diseases, species, organisations, and so on). For scholars it must be possible to unambiguously indicate the subjects of discussion by creating their own identifiers for these entities or by reusing existing identifiers coined by others. This is (arguably) a more accessible way to make associations between articles and data referring to the same entities than alternative natural language processing based techniques. For instance, the progressive adoption of the Digital Object Identifier (DOI) or HTTP URIs for things, along with Nanopublications [19], makes it possible to declare and represent scientific knowledge claims that are globally identifiable on the Web [10, 20]. Such initiatives for citation of research artefacts beyond the document are a strong step towards this area of Linked Research.

Integration of data: In many cases some form of structured data is a direct subject of scholarly communication. In social sciences, for example, statistical data plays a key role as ground truth for validating or falsifying theories. Similarly, in engineering and natural sciences, measurement or observation data is of crucial importance. In the life sciences, statistics, anonymised patient data as well as taxonomic and ontological data can be central. Ideally, such data is directly integrated into the digital representation of a scholarly article in both human- and machine-readable forms. This permits both live updates to the article (including charts and figures) as new data emerges with no extra effort to the authors, and the ability for closer exploration of the data for better understanding by consumers.

Integrated authoring and publication workflow: Researchers should be enabled as far as possible to have a seamless experience when making their work available to their community and to the general public. This spans from basic authoring of text content according to the desired layout, to embedding data and interactive elements, to managing citations, footnotes and metadata, and finally to publishing the article in the preferred location.

Feedback and interactions: When others interact with published work, for example to cite, comment, review, share, recommend or annotate, the original authors should be notified and able to view and react to any new content produced in response to their work. Conversations and commentary may be stored across different systems, however there must be a mechanism for aggregating and displaying scientific discourse around the subject resources to benefit authors and consumers alike.

Support for different views: Content is consumed in a wide variety of formats online and offline, on devices of different sizes and capabilities, and with different interaction techniques (touch screens, speech input/output). Scholarly communication should exploit these different media, presentation and interaction techniques to allow scholarly content to be consumed in different situations [21]. The content should adapt to the available presentation capabilities, and alter its appearance according to some combination of the author's and consumer's preferences.

Adaptation to audiences: Depending on the audience, presentation of scholarly communication should happen on different levels of granularity. For experts in the field, explanations can be shorter, examples are not required and only few illustrations are needed. For newcomers on the other hand, more detailed explanations, illustrations and examples are helpful. Ideally, scholarly content can adapt to the audience. This should happen from a single source of content to minimise the burden on the author, where content elements are marked in such a way the adaptation to the audience can be performed largely automatically.

Integration of interactive content: Where possible, scholarly content should provide dynamic and interactive content. Examples include:

• executable software source code snippets, which demonstrate algorithms, queries, workflows etc. ([22] calls for these to be considered as a critical research output).
• interactive data visualisations with different diagram types, which allow users to zoom in, filter etc.
• small games or self assessment tests, which allow readers to interact with the content in a playful way or to test their comprehension of the knowledge
• widgets or applications, which provide interactive domain specific interactive content (e.g. exploring a large phenotype taxonomy)

Support for multimedia: Multimedia content such as videos, audio, and 3D simulations can dramatically improve the comprehension of scholarly content compared to the traditional static 2D illustrations currently found in papers. In particular, fields dealing a lot with multimedia data like engineering, arts, audio and video analysis, and medicine, would benefit.

Impact metrics: As the volume of scholastic contributions increase, investigators rely on services and filters to choose the most relevant and significant sources. What constitutes relevant and significant may be at the discretion of the service in use, and can be diverse. For instance, traditional author-level metrics while widely employed, check only the bibliographic impact. The rise and adoption of altmetrics as an alternative, aims to incorporate a wide range of information on scholarly contributions in order to derive more immediate and contextually relevant metrics e.g., webometrics [23]. Based on this, reward structures can then be re-engineered if reuse or repurposing of all scholarly or other contributions (e.g., data, methods, social activities and engagement, as well as non-technical implications of science) are taken into account. The availability of machine-readable altmetrics is well aligned to further facilitate Linked Research.

Frontend Web technologies – e.g., HTML, CSS and JavaScript – can be purposed towards authoring tools that are straightforward to use for authors and that require no proprietary software or special configurations; all that is needed is a Web browser. Such tools can be used to edit local files when the author is in offline mode or without an internet connection, but can support writing directly to a personal data store or server when the author is online to facilitate collaboration. Authors can easily go beyond textual content, and embed data, interactive or multimedia elements into their work with native Web components. This is a well-established practice in the social Web, e.g., blogging. Authoring tools built on the Web can be extended or built upon with little overhead to meet specific needs of researchers.

Semantic Web technologies have been widely embraced by the librarian community for organising traditional scholarly content as well as exposing the rich terminological systems underlying such content [24]. For example, the BIBFRAME initiative aims to ease the transition from MARC metadata records to web enabled metadata systems, and both CrossRef and DataCite DOIs support Content Negotiation allowing multiple representations including machine readable representations to be returned.

We propose to use RDFa embedded in HTML together with established vocabularies to add fine-grained semantics to prose content, including metadata (e.g. authors, keywords), document structure (e.g. abstract, conclusion), components and concepts being described (e.g. methodological steps, hypotheses, specific instances of things). A browser-based authoring tool should both automatically generate RDFa where possible, and provide a user-friendly interface to allow the author to add semantics themselves. We argue that researchers should not be expected to write source code, including any form of HTML.

The Linked Research initiative encourages researchers to work transparently, publish early and often and without asking for permission. The Web provides the perfect platform for this. Ideally authors should acquire their own domain name and Web hosting; we acknowledge that this is not always possible or practical, though many institutions provide basic Web hosting and a URL space for all of their staff and students. Hence, employing DOI, PURL, w3id, or alike are possible ways to have some level of permanence for the information.

Publishing work online in static HTML allows wide accessibility for consumers and for added value services. Using CSS and JavaScript makes it possible to refine, adapt or even provide alternative views on a research document according to the consumer’s needs or viewing context.

Distributing work with URLs allows sharing equally between academics and specialists, and interested citizens. Linking between articles enhances discoverability, and coupling this with social notifications (ActivityStreams 2.0 is an appropriate emerging W3C standard which would enable interoperability with other social systems) allows authors to track citations, the reach of their work, and receive and display responses and feedback.

It is important to include licensing information with published work in order to promote fair use and remixing for content creators. We do not presume to dictate the most appropriate licenses for academic work to be released under, however it is worth emphasising that establishing common practices around open licenses such as those from the Creative Commons family is important to encourage a culture of openness and reuse, which, as we previously discussed, is important for a healthy scientific progress.

Ensuring the persistence of scientific results through time is of fundamental importance: readers should be able to access articles published in the past, authors should be able to point to articles and the statements within with confidence that they will be accessible in the future. As of today, archiving and preserving scientific documents are commonly considered the responsibilities of publishers. In the context of Linked Research, where decentralised authoring and publishing is promoted, responsibilities are to be re-evaluated: authors can publish their results online at a service of their choice, such as their own website. However, persisting academic publishing on long-term on the web has two main challenges which may be difficult for individuals to solve when hosting content themselves: 1) availability on the Internet (keeping the document/data server available, maintaining it, and defending it from attacks), and 2) presence on the Web; namely, maintaining a URL that points to the document/data and that does not change. Long-term obstacles to this include URL expiration, revocation (in case of someone taking legal action) and censorship [26].

We already see many efforts at centralised archives - Internet Archive, Archive-It, arXiv, CLOCKSS, Zenodo - however these are each a single point of failure for long-term persistence. In the case of self-archiving, the author is responsible for maintaining their service and domain name. This strategy clearly has non-trivial disadvantages: technical expertise is required, there are maintenance costs, and content may disappear if the author loses the ability to keep this up. In the case of third-party archival, although the responsibilities are delegated a service, this may cease to exist long term, or may act against the interest of the user.

An architecture in which documents are replicated on multiple servers can work towards solving these problems. If the URL is not reachable, the reader should be able to get a copy the content from elsewhere. Lots of Copies Keep Stuff Safe is the idea behind LOCKSS, a project that aims to create a network of libraries and publishers that replicate scientific articles amongst the others [27], based on this idea there is other work that uses existing distributed networks such as BitTorrent [28].

When content is distributed across a decentralised network, documents must have pointers that the network agrees upon. This can be simply done by naming documents via their hashes, as in the InterPlanetary File System (IPFS). This also allows proof that the file hasn't been altered by nodes in the network, since verification would mean re-hashing the file. Recent works show how we can use Trusty URIs to employ hashing techniques whilst remaining compatible on the Web [29].

Ultimately in a Linked Research ecosystem, the role of content creators is also to distribute the content, and a role of public institutions in supporting preservation is to provide reliable hosting, addressing and backup in a decentralised manner. We envision a two-way system. Organisations such as libraries and universities can actively create collections by selecting, sorting and storing published articles. Additionally, authors can submit work themselves to archives, for either consideration by curators or automatic inclusion for backup. Currently in order to do this scholars submit their work manually into different systems, which can result in inconsistent or infrequent updates. To improve on this, the same open Web protocols as described for calls and proceedings can be used to automatically notify chosen archives whenever a new piece of work or dataset is ready.

Just as authors can choose domains they trust to identify their work, they can also create URIs to identify themselves. By connecting this identifier to all of their publications, results and impact of the work they are involved with can be traced directly back to them. Researchers can build a profile which automatically includes their published work and their feedback on the work of others.

Currently, many academics choose to create professional profiles on centralised systems such as ImpactStory, Google Scholar, ResearchGate, Zenodo or Academia.edu. Such systems are poorly integrated with each other, and users frequently must enter their details repeatedly or choose one system and miss out on potentially connecting with members of others. Similarly from within one system, users cannot see the activities of users in other systems, even though many of the types of social activities afforded by the systems (reading or recommending a paper for example) are the same.

dokieli [31] is an open source, progressively enhanced client application for decentralised authoring, with specialisations for writing academic documents. This includes a rich text editor for writing and arranging prose content, as well as embedding data, runnable code, and media (see This ‘Paper’ is a Demo [21] for a demonstration) which generates an HTML+RDFa source automatically. The UI also allows users to add fine-grained semantics by hand, including assigning URIs to concepts at any level and relationships between them, without the need to edit source code. Concepts within a single document can be related to each other with the appropriate RDF properties, as well as creating specific relations with external resources.

In addition, metadata that does not easily fit into prose (for example, Nanopublications) can be embedded as a single block using formats RDF Turtle, JSON-LD, or TriG.

dokieli does not mandate the use of any particular RDF vocabularies, as the subject and content of an article determine how it is best described. By default, dokieli documents make use of schema.org for common terms for Web documents, the SPAR Ontologies for publishing and referencing [32], Web Annotations for comments and peer-reviews, ActivityStreams for social inbox notifications, the Linked Data Platform and Solid vocabularies for personal storage and user preferences, and the ACL vocabulary for access control.

Articles may be stored on the user’s local filesystem or hosted on ordinary Web servers which can serve static HTML files, for example: university user pages, code repositories, personal or company web spaces, or any file hosting service.

Additionally, dokieli is compliant with Solid, a set of protocols and conventions based on the Linked Data Platform recommendation. This incorporates authentication, access control and read and write access to a personal data store. Hosting a dokieli document on a Solid server allows the author to edit the document directly in-browser, and save back to a space they control, and grant permissions to collaborators to do so as well. There are a number of open source Solid server implementations already, and users may host their own or choose a provider. There also exist Solid tools and libraries to help developers to build their own Solid-compliant data store.

People identify themselves to a dokieli document through WebID-TLS authentication. This allows users to maintain a profile at a location of their choosing, and the dokieli instance to write to any Solid-compliant data space which the user is authorised to write to. This is how users can, for example, save changes to a document, or store interactions such as comments.

We support the rights of authors to own their data, and store and publish it where they feel most comfortable. This includes the social interactions and feedback users make around existing publications. Rather than centralising these interactions around the subject document, we took the decision to default to decentralisation of all content by allowing users to authenticate with their personal data space, and choose the location for their interactions at the point of making them. This gives rise to the need for a mechanism to notify the original author that their document has received some interaction. We achieve this by allowing document authors to specify an inbox for either their article as a whole, or any subsection with its own URI using the Solid inbox predicate. Inboxes are containers in a data space which may be appended to by anyone, and do not need to be on the same server as the document itself; if one article has multiple inboxes they can be distributed across as many data spaces as is convenient for the author(s). When an annotation is made, dokieli follows the appropriate inbox link and writes a notification there (see figure 3 for this process and listing 1 for notification contents). When the document is loaded, links are followed to all inboxes in order to retrieve interactions there have been notifications about, so that these can be displayed along with the document (figure 4 and figure 5).

@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix schema: <https://schema.org/> .
@prefix solid: <http://www.w3.org/ns/solid/terms#> .
@prefix as: <https://www.w3.org/ns/activitystreams#> .
@prefix oa: <http://www.w3.org/ns/oa#> .
@prefix c: <https://creativecommons.org/licenses/by/4.0/> .
<> a as:Announce ;
  as:object <http://example.net/foo/abc123> ;
  as:target <http://example.org/article#conclusions> ;
  as:updated "2016-01-24T00:00:00Z"^^xsd:dateTime ;
  as:actor <https://csarven.ca/#i> ;
  schema:license c: .
                                        

An author can also opt to allow anonymous interactions with their documents by pointing to a publicly writeable storage location in their own space, and store interactions on behalf of their audience. Conversely, an author who does not want to display interactions on their article can omit the inbox, but readers can still save notes on an article for their own use, using the same mechanism.

A dokieli research article can contain a relationship e.g., as:inReplyTo, to the call for contributions. That is, every scholarly article can be written in context of the conferences’ or journals’ call for contributions This current article on Linked Research is a reply to ACM Hypertext 2016 Calls for Contributions, and contains an RDF statement as follows:

@prefix sioc: <http://rdfs.org/sioc/ns#> .
<https://csarven.ca/linked-research-scholarly-communication>
  as:inReplyTo <http://ht.acm.org/ht2016/calls> .
                                        

SemStats is a workshop that explores and strengthens the relationship between the Semantic Web and statistical communities. We have published all material on the site for the years 2013, 2014, and 2015 in HTML+RDFa using dokieli’s tooling. It contains interlinks like:

• The SemStats 2013 workshop is a as:inReplyTo to ISWC 2013 Call for Workshops
• The SemStats 2013 workshop also schema:hasPart SemStats 2013 Call for Papers
• SemStats 2013 Call for Papers then contains statements like: description, topic of interest, motivation, keywords, event data, related links, organising committee and affiliations.

In response to the SemStats 2013–2015 calls for papers, several papers have been submitted to the workshop. Those that were accepted for publication became part of the proceedings. Still, if one such paper had previously been published at its own URI, the proceedings version points there. We have published the SemStats proceedings at CEUR-WS.org (see, e.g., http://ceur-ws.org/Vol-1549/) and at the same time introduced a modernised version of the CEUR-WS.org proceedings template based on dokieli’s HTML and CSS (but without JavaScript, which is not permitted by the CEUR-WS.org policy). Figure 6 demonstrates the interlinks between the peer-reviewed research article; Linked Statistical Data Analysis, SemStats 2013 call for contributions, ISWC 2013 call for workshops, and the proceedings of the SemStats workshop at CEUR-WS.org.

There remains a number of features to add to our implementation to meet more of the requirements of the acid test.

Archiving and preservation: Organisations (or indeed individuals) who wish to make their resources available for archiving work can advertise this fact through linking to an inbox where requests to archive can be sent, optionally along with criteria (human- and machine-readable) which submitted material must meet. Requesting inclusion in an archive involves the same underlying mechanism as any kind of notification to an inbox (replying to a call, creating an annotation), though there are user interface optimisations we can make to streamline the process. Similarly for the managers of an archival collections, the technical mechanism for duplicating a document and creating a new URI exist in current ‘Save As’ functionality, however we can automate this for submissions to archives, whilst generating new semantic relations between the original documents and archival copies such as sameAs or derivedFrom.

Collections: Personal collections of work are similar to archives, proceedings and journals, except that one user is adding third-party documents to a collection over which they have control, so there is no request/accept process for something to be included. This is therefore similar to bookmarking, and it is likely that a reference to the original rather than a direct copy will be stored in the collection owner’s space. dokieli functionality for managing collections of all kinds will also facilitate these kinds of personal collections, including the option to send a notification to the original author to let them know their work has been bookmarked. A particular use for this is for when a collection owner is writing their own academic article, and is able to auto-populate references and citations from their collections.

Profiles: An extension to dokieli which allows authors to pre-fill their affiliations and contact details permits automatic population of this data in new articles. Additionally, dokieli can automatically update an academic profile at the author’s own domain whenever new publications are authored or feedback left on the work of others.

Notifications and feedback: We do not currently have good tooling for management of notifications and feedback on articles, particularly if feedback is able to be of different types (e.g. suggestions, questions, disagreements, notes) that maybe subsequently resolved by article authors. This may be in the form of an extension or separate, generalisable, application outside of the dokieli core.

Versioning: Much research is evolving, ongoing work, and we require a solution for better management and identification of snapshots of articles; in particular to associate feedback with a specific version of work which may be subsequently addressed, or to show which iteration of an article was accepted to a conference or journal.