Purpose: This pilot project aimed to assess the applicability and capability of the Emerging Risk Identification Support System (ERIS), developed by The Netherlands Organisation for Applied Scientific Research (TNO), to identify emerging hazards in the atlantic salmon (Salmo salar) and pacific oyster (Crassostrea gigas). ERIS applies text mining rules to identify grammatical and contextual relationships in article titles and abstracts, between potential hazards, effects, and exposure.
Methods: During the pilot phase, ERIS’ ontology was adapted in an iterative and interactive process, according EFSA’s needs, followed by a blind trial in order to align the expert’s evaluation from EFSA and TNO. In the second phase the text mining tool was applied to abstracts from two different databases MEDLINE®/PubMed® and FSTA®, published January 2015-June 2016. The output was evaluated by experts for identification of potential emerging risks, according to an accepted quality assured protocol, based on a multi-eye principle, including a benchmark for the relationships found against a database containing scientific literature of 10 years (2005-2014).
Results: ERIS processed 1,821,576 abstracts and retrieved 707 abstracts (422 for salmon and 285 for oyster). After a first round of expert evaluation, 67 articles for salmon and 47 for oyster were pre-selected. A second round of expert evaluation, comparing the pre-selected abstracts with the benchmark, led to the identification of 28 potential emerging risks; 18 in salmon and 10 in oyster.
Significance: Significant resources were invested by EFSA and TNO to improve the precision of the ERIS tool in identifying potential emerging risks. ERIS has been proven to be a valuable tool for automatically select relevant research abstracts, allowing the identification of potential emerging risks from a trusted and manageable data set, after expert evaluation.