3.2 Can Computational Knowledge Discovery Tools Speed up Scientific Discovery?

The inherent nature of environmental systems calls for interdisciplinary and collaborative research, which is in contrast with the traditional organisation of research around discipline-centric silos. The disconnectedness between marine biology, marine chemistry, socio-economics, etc. is the main barrier slowing down the speed of discovery of new knowledge about complex problems such as the impacts of climate change on nature and society. The main obstacle of scientific advancement in such problems has shifted from production of discipline-centric knowledge to linking together pieces of existing knowledge across disciplines. The management of such a vast body of knowledge is far beyond the capabilities of individual scientists.

Computer tools have so far focused on keyword-based document retrieval while Ocean-Certain(OC), an FP7 project, aims to develop tools for Literature-Based Discovery (LBD) of scientific knowledge about the role of oceans in the export of CO2 to the sediments. LBD is used in biomedicine but use of LBD in earth science is virtually unexplored. The principal idea is that using text mining, computers can machine-read huge amounts of literature, extract the fertile pieces of knowledge (in the form of entities, events and relations) and link them together to identify potential gaps and missing links in the published work, while inferring new knowledge. The ongoing work also investigates the coherence among the published results on a particular question, e.g., whether the growth of phytoplankton leads to better CO2 export. The tool aims to retrieve instances of positive and negative answers, whose ratio would indicate the degree of uncertainty in the collective knowledge. The online collaborative platform will ultimately allow researchers to see one another’s questions and the system responses, as well as giving feedback on which of the system inferred hypotheses seem plausible to pursue.

For maximum benefit to research communities the platform must overcome today’s legal and technical barriers. These are mostly publisher-imposed obstacles, which increase (i) the uncertainty for the final users and low uptake due to unclear licensing issues, thus influencing the ability for scientific reproducibility; (ii) technological complexity that requires going through licensing issues for constant verification; and (iii) hidden costs from the fact that many researchers need to repeat the same text mining processes in their own environment to nonshared content.

Pinar Öztürk is Associate Professor in the Department of Computer and Information Science in the Norwegian University of Science and Technology (NTNU). She received her PhD from the Department of Computer and Information Science at NTNU. Her focus is artificial intelligence methods including rule and model-based methods and machine learning. She applies AI methods for distributed decision-making, ontology building and knowledge extraction from text. She was project leader and worked on various small and larger projects funded by the Norwegian Research Council, EU and the industry. Currently, she is leading a work package in the Smart Power Grid project (funded by Utilities) and a task in the EU funded OCEAN-CERTAIN project (since November 2013) that focuses on text mining in Climate Science.