A Scientific Information Extraction Dataset for Nature Inspired Engineering

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2078–2085 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2078 A Scientific Information Extraction Dataset for Nature Inspired Engineering Ruben Kruiper∗, Julian F.V. Vincent, Jessica Chen-Burger, Marc P.Y. Desmulliez, Ioannis Konstas Heriot-Watt University, Riccarton Campus Edinburgh, United Kingdom EH14 4AS ∗Corresponding author: rk22@hw.ac.uk Abstract Nature has inspired various ground-breaking technological developments in applications ranging from robotics to aerospace engineering and the manufacturing of medical devices. However, accessing the information captured in scientific biology texts is a time-consuming and hard task that requires domain-specific knowledge. Improving access for outsiders can help interdisciplinary research like Nature Inspired Engineering. This paper describes a dataset of 1,500 manually-annotated sentences that express domain-independent relations between central concepts in a scientific biology text, such as trade-offs and correlations. The arguments of these relations can be Multi Word Expressions and have been annotated with modifying phrases to form non-projective graphs. The dataset allows for training and evaluating Relation Extraction algorithms that aim for coarse-grained typing of scientific biological documents, enabling a high-level filter for engineers. Keywords: Scientific Information Extraction, Relation Extraction, Biomimetics, Trade-Offs 1. Introduction Discovering relevant scientific literature is a hard and time- consuming task that often requires domain expertise (Alper et al., 2004; El-Arini and Guestrin, 2011; Pain, 2016). This difficulty can lead to barriers in inter-disciplinary fields of study, where an expert in a target domain is often a novice in the source domain (Carr et al., 2018). A specific ex- ample of an inter-disciplinary field of study is Nature In- spired Engineering, also known as biomimetics (Kruiper et al., 2016). A well known example of biomimetics is Velcro R ⃝ (Mestral, 1955), an ‘adhesion’ method that is in- spired by nature - specifically the burrs of burdock plants. In biomimetics an engineer is interested in learning from nature to solve a problem at hand (Hoeller et al., 2016). It has been shown that identifying, selecting and understand- ing relevant biological information is challenging for en- gineers (Vattam and Goel, 2013a). Information Extraction (IE) systems can partially ease these challenges by captur- ing the central concepts and relations in a text (Augenstein et al., 2017; G´abor et al., 2018). A major issue in inter-disciplinary research is that search terms from a source domain may not retrieve relevant re- sults in a target domain. As an example, in the biological domain the term ‘bleaching’ can refer to a separation pro- cess between the retina and opsin in vertebrate eyes (Grimm et al., 2000). A non-biologist may expect ‘bleaching’ to refer to cleaning, sterilizing, or whitening (Nagel, 2014). As a result, terminology from the engineering domain does not always provide a good starting point for the identifica- tion of relevant biological information (Fayemi et al., 2015; Kruiper et al., 2017). A more appropriate approach that allows for cross-domain search, without relying so much on domain-specific semantics, focuses on TRADE-OFF re- lations (Vincent, 2016; Kruiper et al., 2018). In …

Content truncated to 500 words