The increasing demand for bioenergy crops presents our society with the opportunity to design more sustainable landscapes. We have created a Biomass Location for Optimal Sustainability Model (BLOSM) to test the hypothesis that landscape design of cellulosic bioenergy crop plantings may simultaneously improve water quality (i.e. decrease concentrations of sediment, total phosphorus, and total nitrogen) and increase profi ts for farmer-producers while achieving a feedstock-production goal. BLOSM was run using six scenarios to identify switchgrass (Panicum virgatum) planting locations that might supply a commercial-scale biorefi nery planned for the Lower Little Tennessee (LLT) watershed. Each scenario sought to achieve different sustainability goals: improving water quality through reduced nitrogen, phosphorus, or sediment concentrations; maximizing profi t; a balance of these conditions; or a balance of these conditions with the additional constraint of converting no more than 25% of agricultural land. Scenario results were compared to a baseline case of no land-use conversion. BLOSM results indicate that a combined economic and environmental optimization approach can achieve multiple objectives simultaneously when a small proportion (1.3%) of the LLT watershed is planted with perennial switchgrass. The multimetric optimization approach described here can be used as a research tool to consider bioenergy plantings for other feedstocks, sustainability criteria, and regions.Published in
Named entity recognition (NER) is a key component of many scientific literature mining tasks, such as information retrieval, information extraction, and question answering; however, many modern approaches require large amounts of labeled training data in order to be effective. This severely limits the effectiveness of NER models in applications where expert annotations are difficult and expensive to obtain. In this work, we explore the effectiveness of transfer learning and semi-supervised self-training to improve the performance of NER models in biomedical settings with very limited labeled data (250-2000 labeled samples). We first pre-train a BiLSTM-CRF and a BERT model on a very large general biomedical NER corpus such as MedMentions or Semantic Medline, and then we fine-tune the model on a more specific target NER task that has very limited training data; finally, we apply semi-supervised self-training using unlabeled data to further boost model performance. We show that in NER tasks that focus on common biomedical entity types such as those in the Unified Medical Language System (UMLS), combining transfer learning with self-training enables a NER model such as a BiLSTM-CRF or BERT to obtain similar performance with the same model trained on 3x-8x the amount of labeled data. We further show that our approach can also boost performance in a low-resource application where entities types are more rare and not specifically covered in UMLS.
ABSTRACT:The application of spatiotemporal (ST) analytics to integrated data from major sources such as the World Bank, United Nations, and dozens of others holds tremendous potential for shedding new light on the evolution of cultural, health, economic, and geopolitical landscapes on a global level. Realizing this potential first requires an ST data model that addresses challenges in properly merging data from multiple authors, with evolving ontological perspectives, semantical differences, and changing attributes, as well as content that is textual, numeric, categorical, and hierarchical. Equally challenging is the development of analytical and visualization approaches that provide a serious exploration of this integrated data while remaining accessible to practitioners with varied backgrounds. The WSTAMP project at Oak Ridge National Laboratory has yielded two major results in addressing these challenges: 1) development of the WSTAMP database, a significant advance in ST data modeling that integrates 10,000+ attributes covering over 200 nation states spanning over 50 years from over 30 major sources and 2) a novel online ST exploratory and analysis tool providing an array of modern statistical and visualization techniques for analyzing these data temporally, spatially, and spatiotemporally under a standard analytic workflow. We discuss the status of this work and report on major findings.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.