Scienti c publications point to many associated resources, including videos, prototypes, slides, and datasets. However, discovering and accessing these resources is not always straightforward: links could be broken, readers may be o ine, or the number of associated resources might make it di cult to keep track of the viewing order. In this paper, we explore potential integration of such resources into the digital version of a scienti c publication. Speci cally, we evaluate the most common scienti c publication formats in terms of their capability to implement the desirable attributes of an enhanced publication and to meet the functional goals of an enhanced publication information system: PDF, HTML, EPUB2, and EPUB3. In addition, we present an EPUB3 version of an exemplary publication in the eld of computer science, integrating and interlinking an explanatory video and an interactive prototype. Finally, we introduce a demonstrator that is capable of outputting customized scienti c publications in EPUB3. By making use of EPUB3 to create an integrated and customizable representation of a scienti c publication and its associated resources, we believe that we are able to augment the reading experience of scholarly publications, and thus the e ectiveness of scienti c communication.
Scienti c publications consist of more than only text: they may also point to many associated (binary) resources, including videos, prototypes, slides, and datasets. Yet today, only the access to the text of a scienti c publication is straightforward; the associated resources are often more di cult to access. For instance, readers may not always have an Internet connection at their disposal to download related materials, and even when this is the case, links might become broken after a while. Furthermore, given their diverse nature, related materials often need to be accessed in a di erent reading environment like a standalone media player, causing readers to lose track of the scienti c narrative.
The 2007 Brussels Declaration3 by the International Association of Scienti c, Technical and Medical (STM) Publishers states that \raw research data should be made freely available" and that \one size ts all solutions will not work". In this paper, we illustrate that the ability to (adaptively) create an integrated representation of a scienti c publication and its associated resources contributes to these goals. Speci cally, we evaluate the most common scienti c publication formats in terms of their capability to implement the desirable attributes of an enhanced publication and to meet the functional goals of an enhanced publication information system: PDF, HTML, EPUB2, and EPUB3. In addition, we present an EPUB3 version of an exemplary publication in the eld of computer science, integrating and interlinking an explanatory video and an interactive prototype. Finally, we introduce a demonstrator that is capable of outputting customized scienti c publications in EPUB3.
The rest of this paper is structured as follows. In Section 2, we discuss a number of current best practices among three scienti c publishers, focusing on the way open formats and their features are used to enhance scienti c publications. Next, in Section 3, we investigate to what extent PDF, HTML, EPUB2, and EPUB3 facilitate the use of enhanced scienti c publications and corresponding information systems. In Section 4, we present an exemplary scienti c publication in EPUB3 that integrates an explanatory video and an interactive prototype. In Section 5, we introduce our demonstrator for creating customized scienti c publications in EPUB3. Finally, in Section 6, we present our conclusions and a number of directions for future work. 2
In this section, we brie y discuss a number of current best practices among three scienti c publishers, focusing on the way open formats are used to make available scienti c publications that have been enhanced with multimedia, interactivity, and/or Semantic Web features.
BioMed Central and Hindawi Publishing Corporation: These publishers make scienti c publications available in several formats: PDF, HTML, and EPUB2. The HTML version of the publications can for instance be enhanced with reusable data (e.g., supplementary datasets), while the EPUB2 version of the publications just uses links to cited publications in EPUB2 format. However, the publications in question do not contain any embedded interactive multimedia content.
Elsevier: Elsevier makes available di erent versions of a scienti c
publication: PDF, HTML, MOBI, and EPUB2. In addition, authors are able to
deposit their datasets, making it possible for readers to access and download these
datasets [
In summary, we can conclude that none of the aforementioned EPUB2 versions { as currently made available by BioMed Central, Hindawi Publishing Corporation, and Elsevier { embed interactive multimedia content for o ine usage (i.e., readers need to have network connectivity in order to be able to access all linked resources), nor do they contain Semantic Web features. 3
In recent years, a new open format for distribution and interchange of digital
publications has emerged, called EPUB3 [
Thoma et al. [
Besides the attributes of enhanced publications, we also considered data
model and information system aspects. Bardi et al. [
Finally, according to Liu [
Ideally, an enhanced publication information system should be able to
support all the desirable attributes mentioned above. Considering the desirable
attributes of enhanced publications and the functional goals of enhanced
publication information systems, we mapped the attributes identi ed in [
Next, we investigated what scienti c publication formats are the most promising to cover both the desirable attributes of an enhanced publication and the functional goals of an enhanced publication information system. We have summarized our ndings in the four rightmost columns of Table 1. Corresponding explanatory notes can be found below.
Packaging with supplementary material: This functional goal states that
it should be possible to add supplementary material to a scienti c
publication. PDF can embed audio and video but it does not support rich media
(e.g., media overlays). As such, it is not a suitable format for embedding
various types of associated resources (e.g., interactive content and standalone
applications). Consequently, PDF has limited support for this functional goal
and its underlying attributes. Note that extensions exist, such as export to a
PDF Portfolio in Adobe Acrobat5, that make it possible to combine related
materials. However, to the best of our knowledge, none of these extensions for
instance allow embedding interactive content and standalone applications.
Furthermore, the embedded resources are not reusable, unlike the EPUB3
format, which lets users reuse embedded resources. In order to package
research data within an HTML le, all the dependencies need to be packaged
as well. While this is possible (e.g., using a zipped folder), there is no
standardized approach to do this, as opposed to EPUB2 and EPUB3. Therefore,
we do not consider HTML to be suitable for meeting this functional goal.
According to the EPUB2 speci cation [
Attributes Enabling repetition of experiments: This functional goal aims at enabling researchers to (re-)execute experiments and/or demonstrators from within a scienti c publication. PDF has limited support for scripting and code execution. However, the support available is not su cient for building small standalone applications that can act as interactive content (e.g., self-contained widgets). As a result, PDF is not suitable for meeting this functional goal. HTML is able to embed code (e.g., JavaScript). Moreover, thanks to the inline frame element (that is, the iframe element), HTML can also be used as an interface to other experiments. As EPUB2 does not support JavaScript, it is not suited for repetition of experiments. However, similar to HTML, EPUB3 supports JavaScript, and thus the aforementioned functional goal Format PDF HTML EPUB2 EPUB3
D D D D
D D*
D*
D D D
D* D* (unless experiments are involved that for instance use complex algorithms on clusters to obtain their results).
Improving readability and understanding: PDF is a speci c format for
print, and not for screen readers. While still undeniably the most suitable
format for print layout, in digital form, it does not have device
independence [
Apart from a suitable format, interlinking supplementary materials requires suitable ontologies. Fortunately, many suitable candidates for general and speci c interlinking purposes are already available. For example, schema.org is an ontology that is suitable for use in a variety of domains, including the description of events and creative works. It can thus be used to semantically enhance publications, and it can also be extended by other ontologies. Furthermore, Standard Analytics6 aims at turning scholarly publications into an interface to a web of data, making use of already existing web ontologies. Moreover, Structural, Descriptive, and Referential (SDR)7 is an ontology for representing academic publications, related artifacts (e.g., videos, slides, and datasets), and referential metadata. This ontology can generically de ne all possible interactive and multimedia resources. In addition, any publication can use general ontologies such as the Citation Typing Ontology (CiTO)8, the Bibliographic Ontology (BIBO)9, and the Common European Research Information Format (CERIF)10. Finally, publications may also need to make use of ontologies that are speci c for their research domains (e.g., in the medical domain, the Infectious Disease Ontology (IDO)11 could be used). 6 https://standardanalytics.io/ 7 http://onlinelibrary.wiley.com/doi/10.1002/asi.23007/full 8 http://www.essepuntato.it/lode/http://purl.org/spar/cito 9 http://bibliontology.com/ 10 http://helios-eie.ekt.gr/EIE/bitstream/10442/13864/1/IJMSO_2014_CERIF_ authorFinalVersion.pdf 11 http://infectiousdiseaseontology.org/page/Main_Page Portable packaged le: PDF has limited support for packaging interactive content and standalone applications. Furthermore, it cannot bridge the gap between online and o ine usage. Indeed, PDF is an o ine format for print, and any interactive parts will not remain after printing a publication. As mentioned before, HTML lacks a proper packaging structure, making this format not a suitable candidate for meeting this functional goal. A similar remark holds regarding EPUB2, as this format does not have support for embedding interactive multimedia resources. As EPUB3 has extensive support for embedding interactive multimedia resources, it can be considered a suitable format for creating portable packaged les. Ideally, users expect that all types of resources can be embedded in a packaged le, regardless of their size. This is one of the shortcomings of EPUB3. Embedding large datasets makes the size of an EPUB3 le potentially very large, causing portability and readability issues. We discuss a possible solution to this issue in Section 5.
Suitable for print: Currently, PDF is the only format suitable for print. Although HTML, EPUB2, and EPUB3 can also be used for the purpose of print, they have been designed for screen readers and can currently not match the high typesetting demands for print publications.
As can be seen in Table 1, EPUB3 is the format that supports most desirable attributes of an enhanced publication and most functional goals of an enhanced publication information system. Only PDF is suitable for print output, given that HTML and EPUB(2/3) have been primarily designed for screen output, typically resulting in a layout that is suboptimal for print. Note that, as a workaround for this problem, the EPUB(2/3) and HTML versions of a publication can embed or link to the PDF version of a publication. 4
In this section, we demonstrate how EPUB3 can be used to create an integrated
representation of a scienti c publication and its associated resources. To that
end, we enhanced the \Everything is Connected" publication [
epub 13 http://readium.org/
Furthermore, we semantically enhanced our exemplary EPUB3 publication by making use of schema.org, a general ontology that allows describing books and articles, among other creative works. Thanks to properties such as embedUrl, description, and contentUrl, schema.org makes it possible to indicate how a resource is related to the target EPUB3 publication in a straightforward way. We illustrate this in Figure 2. Note that schema.org is supported by major search engines such as Bing, Google, Yahoo!, and Yandex. However, at the time of writing this paper, the aforementioned search engines did not have support yet for indexing EPUB3 publications (and reading the metadata available within these publications).
Fig. 1: Exemplary scienti c publication enhanced with an explanatory video (left) and an interactive prototype (right). Both the video and the prototype have been embedded into the EPUB3 version of the scienti c publication.
In the previous sections, we explained how supplementary materials can be embedded into a scienti c publication. As mentioned before, embedding all relevant supplementary materials in a portable packaged le is not always cost-e ective and/or desirable for a user. Since the size of an EPUB3 le is dependent on the size of all embedded resources, it will not be lightweight in all use cases, e.g., when embedding large datasets. The problem is that, on the one hand, a packaged le should not face portability and other usage issues relevant to its size. On the other hand, the advantages of having a portable packaged publication are overthrown with the disadvantage of not being able to distribute the entire publication properly. Users may not need all embedded supplementary materials and instead, wish to have their own customized lightweight publication. For instance, we can refer to big datasets or high-resolution images which can be located in a remote repository instead of embedding them in the portable packaged le. An environment for outputting customized publications allows users to select and embed the supplementary materials to the extent that they choose. Hence, they can determine the size of the EPUB3 le themselves. That way, the problem of distributing overly large publications is solved, and only the content that the user needs is distributed. The only disadvantage of this approach is the added complexity at the distribution side (i.e., at the platform of the publisher). However, most publishers already have an extensive online distribution infrastructure, which could easily be expanded with an interface such as the one we propose. For example, publishers such as Elsevier o er di erent formats of a publication to users. In particular, on the ScienceDirect website of Elsevier, there is an option for the user to select his/her preferred format.
To illustrate this concept of customizable publications, we implemented a basic demonstrator in which a user can rst select the relevant supplementary material using a web interface, after which a customized EPUB3 publication is outputted. Figure 3 shows the user interface of our online demonstrator. Content selection is entirely done at the client side, based on the HTML representation of a publication. The selected content is then packaged as an EPUB3 le on the server side. The resulting demonstrator is available online14. Note that the author of a publication can determine which elements are customizable, simply by adding the class customizable to the desired HTML elements.
Ideally, the implemented functionality for outputting customized publications
in EPUB3 would be integrated into an authoring environment, where authors
and publishers could indicate which elements of a publication are customizable.
In previous work, we have implemented such an authoring environment for the
collaborative creation of enriched e-Books using EPUB3 [
We can identify a number of directions for future research. First, user-friendly authoring tools are needed that allow easily creating enhanced scienti c publications, and where these scienti c publications can act as an interface to di erent research outputs. We have already started taking steps in this direction. Second, these authoring tools need to support di erent output formats, in order to meet the needs of both readers that are reading on paper and readers that are reading digitally. Third, these authoring tools also need to make it possible to easily add metadata to EPUB3 versions of scienti c publications, such that EPUB3 versions of scienti c papers may have the same degree of discoverability as PDF and HTML versions. Finally, it would be interesting to investigate the good practices of novel publication repositories such as PLOS ONE, Figshare, and ResearchGate.
The research activities described in this paper were funded by Ghent University, iMinds (a research institute founded by the Flemish Government), the Institute for Promotion of Innovation by Science and Technology in Flanders (IWT), the FWO-Flanders, and the European Union.