Introduction: A Description of the Application and Users

Context: Linked Data 

Although the break-neck speed of innovation in science and technology make these very exciting times, we are also facing serious bottlenecks when it comes to communicating scientific and technical knowledge. Due to their inefficient nature, the current scientific communication tools— lectures, scientific papers, and data— make research and education agonizingly slow processes \cite{Olah_2017}. The heart of the matter seems to be a general stagnancy in communicating implicit knowledge to others, or a lack of participation in the effort of making the implicit explicit, so that it can be processed faster by humans and machines alike. After all, searching for an answer to a question through querying a knowledge graph with natural language is likely to be much more efficient than using the modest text-recognition and reasoning faculties of humans, and for a machine, it would be much more efficient to reason over that same graph than scanning the texts of billions of documents with an attempt to extract knowledge from them. Indeed, this bottleneck in knowledge flow is widely recognized, and solutions such as Semantic Web and linked data has been proposed and acknowledged as the next step in World Wide Web's evolution \cite{Berners_Lee_2001}. These are relatively old propositions, however, and their adoption still seems limited after more than 20 years. Awe-inspiring initiatives such as DBpedia and adoption by giant data creators like Twitter aside, the semantic web technology remains inaccessible (or simply undesirable) to average Web citizen / developer / designer today. The Web is full of dead SPARQL endpoints, and technologies such as REST APIs is being preferred over triple stores.
One of the major reasons for slow adoption rate of semantic web technologies could be their inaccessibility. Like with any new technology, semantic web is in a technical-state and generally lacks user-friendly interfaces. For instance, the industry standard ontology editor, Protege is disliked by many but used anyway due to lack of a better alternative; and tools for semantic web are spread out throughout the web. Besides the shortcomings of existing standard tools even in the most basic aspects (e.g., both Protege and Stardog missing RDF validators, and Protege 5 not even giving an error message upon encountering a .turtle file with a faulty line),  as of today, there is no integrated development environment for semantic web IDE—something that could streamline the messy linked data workflow from data to ontology, to triple store, and to query interface. Therefore, development of tools that are better designed from both a software development and user-experience perspective (e.g., Neo4j \cite{inc2017}) are urgently needed to overcome linked data and semantic web initiatives' own bottleneck: lack of good tools. Linked Data Reactor (LD-R),  a project at Vrije Universiteit Amsterdam, is such a candidate, and it could significantly lower the threshold for using and producing linked data. The current project aims to utilize the LD-R framework and demonstrate this point on usability with a prototype, while also using the framework to create an analysis platform for a scientific problem: knowledge flows and interdisciplinarity in research.

Domain: Interdisciplinarity

Interdisciplinarity in research is generally seen as desirable, and it is likely to be an important factor that can bring about new perspectives and solutions to our increasingly sophisticated and multi-faceted research pursuits today. However, the impact of interdisciplinarity —or to put simply, the effect of diversity of research in an article,  journal, or institute— on the scientific quality and merit is a matter of debate, and there does not seem to be conclusive findings. Some authors suggest that 'distance' between disciplines may play a critical role in the effectiveness of interdisciplinarity \cite{Jensen_2013,Zhang_2015}, and some claim a 'U-shaped' relationship \cite{Wang_2015}, while the discussion also includes various other theories and findings \cite{Yegros_Yegros_2015,Barry_2008}. The ongoing debate and possible impact of results on policy making invites more studies in this direction, and due to the reliance of common research questions in this field on bibliometric data, it provides a suitable domain to investigate for the current project. 

Goal

The current project aims to create a linked-data interface that allows users to query a database of scientific publications in order to get detailed metadata concerning publication patterns . Through the interface, it is particularly aimed to allow a data-driven investigation of interdisciplinary collaboration patterns. Some research questions, alongside other possible exploratory ones that may arise in future, are as following:
This project is part of the 10-month research project 'Knowledge Flows in Interdisciplinary Research' \citep{institute}, and as the investigation progresses over the next few months, the current linked-data interface is expected to shed light to these research questions, and also motivate new ones.

Methodology

 In order to explore the effects of interdisciplinarity and discover scientific collaboration patterns, we aim to model the domain of of scientific publications using an ontology, and populate it using a bibliometric databases such as Elsevier's Pure \cite{pure}, RISIS \cite{20172017}, and Web Of Science \cite{analytics2017}. After being parsed and integrated with an ontology, this data will be made available in a triple store, which will be connected to a linked data query engine and & interface, Linked Data Reactor (LD-R) \cite{besselaar2017}. For the purposes of the current project, and bearing in mind time limitations, the Pure database of VU was implemented in the prototype detailed in this report.

Users

The intended initial user base for the linked data interface are the researchers involved in the Knowledge Flows in Interdisciplinary Research project: Dr. Ali Khalili, Dr. Sascha Friesike, Prof. Peter van den Besselaar; and Academy Assistants Frederik König and John Can Lokman. As this research trajectory continues in the following years, more researchers and students who are involved in proceeding —or similar— projects in both Vrije Universiteit Amsterdam and other universities can be expected to join the user base.

Design