Communicating research results in most fields involves managing the content of multiple media types, large datasets or databases, and computational processing and emulation, among others, and thus makes the results of almost any research an interlinked and interactive collection of content resources. The current presentation of paper results are still accomplished largely via text or power-point files, making it difficult to demonstrate the results in most suitable media and to track the supporting data and related processes.
With the rapid development of information technology, the content organization and presentation of research papers are changing. Paper content is becoming increasingly more structured, semantically tagged, interlinked, and embedded with rich media, making a paper more playable and smart. STM Tech Trends 2014
A dissertation is usually represented as a simple document such as a PDF or Word file. In this form its content is static, non-executable, and not linked to its supporting primary sources and processing tools. The content is generally not structured and has no semantic tagging. This format thus inhibits in-depth exploration and verification of background context and supporting content.
An extractable dissertation, on the other hand, is meant to be an interactive system, in which data can be extracted for reuse such as in computation, reinterpretation, and reanalysis, and the content can be reassembled, processed, and represented in various formats. An extractable document should have the following characteristics:
Be composed of digital objects of various media types, representing primary materials, processed results, and even processing modules to produce desired results; Be structured and semantically tagged; Be executable at various levels down to the level of each object; and Contain modules for processing and presentation, so the various objects of the paper can be custom-organized to meet different needs. The Metadata Encording & Transmission Standard (METS) schema is a standard for encoding descriptive, administrative, and structural metadata regarding objects within a digital library, expressed using the XML schema language of the World Wide Web Consortium. More details are available at
An electronic thesis or dissertation (ETD) tool was recently designed as a data carrier and research gateway (Schopfel, Chaudiron, & Jacquemin, 2014), where studies on digital objects management and utilization were one of its active topics. A typical tool is OpenETD
Typical models for extractable digital papers are the modular paper model and the semantic publishing model. The modular paper model was proposed by Kircz (1998; 2002), where a paper is composed of multiple modules defined as information units with unique characteristics and representation. Datasets, images, music, videos, and so on are regarded as independent interactive objects or modules that can be connected to a fixed framework for exchange. A modular structure thus gives better user experience for reading as well as publishing.
With the semantic publishing model, Hunter (2006) put forward a new information format called the Scientific Publication Package (SPP). This tool is used to package raw data, provenance products, algorithms, software, text, context, and metadata, wherein scientists are able to capture, index, store, share, exchange, reuse, compare, and integrate scientific results. SPP is a compound digital object based on a number of scientific models and represented as a Resource Description Framework (RDF) package. Relations among internal digital objects in compound objects are either explicitly defined by the inference of ontology in the stage of metadata capture, or defined by the scientist in the SPP specifications. The workflow technology is emphasized as part of the research process to capture the processing chains of generating scientific data and provenance products. This tool thus allows scientists to describe and perform their experiments, or track the source of errors and defects in a repeatable, verifiable, and distributed way (Woutersen-Windhouwer & Brandsma, 2009).
In terms of author editing tools, Project BioLit (Fink & Bourne, 2007) and the Scientific Compound Object Publishing and Editing System (SCOPE) (Cheung et al. 2008) offer practical data explorations made from the perspective of semantic markup and compound digital objects, respectively. Project BioLit uses the National Library of Medicine’s Document Type Definition (NLM DTD) to store standardized and machine-readable publications. The NLM DTD format enables articles to be archived as XML files, which include some semantic markup of the content and unique identifiers for the article and its objects (e.g. graphs and tables). The tool is supposed to facilitate the integration of the open literature and biological data (Fink & Bourne, 2007).
Ideally, the researchers themselves are the best persons to create the extractable digital objects, because they know the complete research processes, methods, experimental materials, data, and results. SCOPE has made some attempts to tap this potential. It is designed to enable scientists to easily author, publish, and edit scientific compound objects and to package scientific experiments or datasets and resources (Cheung et al., 2008). In this system, the digital objects can be published and exchanged individually. However, examples of SCOPE have shown that the generation of compound digitals needs the help of a semantic relation network, thus making the tool impractical.
In the proposed Dpaper, a research paper is organized with a modular paper model that uses standardized digital paper templates to structure the content. Content and presentation are independent of each other during the authoring process. The whole paper is composed of a set of digital objects that can be tagged, executed, and assembled in a personalized way, and presented as Web pages or converted to an MS Word document. Digital objects are described with open metadata specifications (such as METS and Dublin Core), and a series of operation interfaces are defined to act as micro-services for certain complex digital objects.
Figure 1 illustrates the processing framework of Dpaper. The system consists of five modules: (1) structured documents representation, (2) rich media objects making, (3) digital documents editing, (4) Web page presentation, and (5) structured documents storage. Their functions are as follows:
Structured documents representation: A general structured paper (dissertations or journals) template that is designed with descriptive specifications. Rich media objects making: Responsible for the management of digital objects including generation, interaction, encapsulation, and conversion. Digital documents editing: Various kinds of content (texts and objects) are created, edited, modified, and combined together as a paper content unit that is defined as a new structured object. Each object is automatically tagged with structural elements to make the paper smartly structure-aware. The various digital objects are assembled in the general paper template, in which objects can be associated with data. Web page presentation: The paper can be presented as Web pages to be browsed and published, or converted to an MS Word document. Structured documents storage: All the data and digital objects are stored as Web packages, while the structured information is stored in an XML file for digital document exchange or third-party reuse.
The content of the digital dissertation is described according to its granularity levels, where the description specifications are adopted from METS, Dublin Core, and Collection Tag Library version 3.0
As illustrated in Figure 2, the description framework consists of metadata objects, organizational objects and content objects. The metadata utilizes elements in Dublin Core, including title, author(s), institutions, abstract (Chinese and English), keywords, etc. The organizational objects consist of cover, declaration statements, table of contents, list of charts/figures, references, acknowledgments, appendices, and so on. The content objects include chapters and sections, and rich media objects that cover images, atlases, charts, audios, videos, animation, maps, data, network diagrams, algorithms, software, etc. Each digital object is uniformly identified by a Universally Unique Identifier (UUID). For instance, the section object that records the structure and content of a section in a chapter is described by Sec-id, Sec-title, and Sec-content. Content such as images, atlases, charts, audios, videos, and so on are embedded in the corresponding sections in the form of compound digital objects.
Compound digital objects in Dpaper integrate the entire procedures of data loading, program processing, editing, rendering, storage, etc. Data are represented as JavaScript Object Notation (JSON) It a lightweight data-interchange format (
We define several kinds of events to drive digital objects, such as Create, BeginEdit, EndEdit, Delete, Copy, Erase, and so on. The typical event response procedure is as follows:
Document operation (button/shortcut)
With digital template technology, different digital objects defined in the digital dissertation are marked with semantic tags, and the annotation process is handled automatically by the system. To reduce the complexity of editing, two digital templates, Dissertation Template and Journal Article Template, are developed in accordance with the Master Thesis Format of the University of Chinese Academy of Sciences and with the Article Format in the
During the Dpaper creation, the dissertation becomes structured and partially semantic, which makes it easy for the machine to read and understand and to reuse for third-party systems. There are following three modes for reuse.
Internally for a digital object, the data, programs, and library files are packaged into a separate Web package that has its interface and object description metadata, and can run without Dpaper. The copied objects can be embedded in other digital papers.
The data in digital objects support reuse by means of format conversion. The data in Dtable, Dcharts, Rrelation Chart, etc. can be converted into such formats as JSON, CSV In computing, a comma-separated values (CSV) file stores tabular data (numbers and text) in plain text. RDF/XML is a syntax defined by the W3C to express (i.e. serialize) an RDF graph as an XML document.
Internally in Dpaper, there are three kinds of XML files designed to store the structure, the data, and the formatting styles of the dissertation. The structure file records the hierarchical relationships between objects, and the style file stores the format displayed in templates, such as font, size, position, color and style. The internal structure files used for data storage and documents exchange are converted into separate METS (Figure 5).
A Dpaper system consists of PC client software, an MS Word plug-in, and Web application software (Figure 6). As the main platform, PC client software is designed for digital paper authoring, and includes functions of digital objects making, content editing, data management, preview, documents conversion, etc. The MS Word plug-in constructs a toolbar in MS Word for extractable digital paper authoring, and it can open digital papers created in Dpaper PC client, embed rich media objects defined in Dpaper, and insert references automatically. So the plug-in makes it possible to author digital papers in MS Word contexts. The Web application provides the functions of data synchronization management, team collaboration, and other functions. Currently, Dpaper can run on Windows 7, Windows 8, and Windows XP, and supports Internet Explorer 8.0 or above.
Figure 7 illustrates the process on how to create an extractable digital dissertation. The screenshot shows the results of a master’s thesis created in Dpaper.
Dpaper explores a method to generate extractable ETDs, and presents the main ideas on digital representation, object making, presentation, and system framework. By using the digital template mechanism, a paper is represented as different types of objects, and stored as three kinds of XML files: data files, structure files, and style files, which are separated but interlinked. This kind of digital papers can be manipulated, reused, and represented as a Word document, an XML file that can be read by the machine, and runnable Web pages. Dpaper has changed an academic paper from an only human-readable document to a machine-readable document, and supports real-time playing of its objects to richly track and demonstrate the research and its data. So far, Dpaper has been released as an online version (idpaper.las.ac.cn) and currently provides a digital authoring platform for dissertations or journal papers. These papers are structured and partially semantically tagged, and can be run as a Web system. In the future, we will explore Dpaper’s applicability to create other extractable academic papers such as reports and books.
Dpaper is not without limitations. Compared with MS Word, there is still a large gap in terms of general editing functions, so it needs to include objects for equations, graphics, etc. in the main platform. Also, the digital object calls between the main platform and Word plug-in need to be smoother, and its stability has to be improved.