Abstract
Various industrial and development projects have the potential to adversely affect threatened and endangered species and their habitats. The federal Endangered Species Act (ESA) requires preparation of a biological assessment or habitat conservation plan before federal agencies can authorize, through decision documents and permits, unintentional and otherwise prohibited “take” (i.e., harm) of listed species. These documents describe the potential effects of proposed projects on listed species and include measures to mitigate those effects. Collectively, these assessments, plans, decision documents, and permits—termed ESA documents in our study—are valuable for identifying approved mitigation options that could apply to future projects. However, owing to the volume, length, and complexity of these documents, manual review would be time- and labor-intensive. In this study, we apply three supervised machine learning algorithms, including two based on state-of-the-art transfer learning, to develop and evaluate predictive models capable of extracting mitigation-related text from ESA documents. The machine learning models were developed based on a training dataset that was created as part of this study. The best performing model showed an estimated ROC-AUC score of 0.98 and a precision recall AUC score of 0.86 during cross-validation, indicating great potential for effectively extracting mitigation-related content from existing documents. To illustrate the utility of this technology, we present a simulated case study application in which the use of pretrained machine learning models capable of recognizing mitigation measures, coupled with a large historical corpus of ESA documents and keyword filters, provided a means to rapidly assess the commonly used mitigation measures for a given species. While this technology did not eliminate the requirement for biological expertise, it did allow for rapid scoping assessments and could serve as a supporting resource even for experienced biologists.
Similar content being viewed by others
Notes
See https://www.fws.gov/endangered/laws-policies for more information.
Take means to harass, harm, pursue, hunt, shoot, wound, kill, trap, capture, or collect or attempt to engage in any such conduct.
See https://www.fws.gov/sacramento/es/overview/Documents/ESA_Basics.pdf for more information.
See https://esadocs.defenders-cci.org for more information.
References
Aphinyanaphongs Y, Tsamardinos I, Statnikov A, Hardin D, Aliferis CF (2005) Text categorization models for high-quality article retrieval in internal medicine. J Am Med Inform Assoc 12:207–216
Bekhuis T, Demner-Fushman D (2012) Screening nonrandomized studies for medical systematic reviews: a comparative study of classifiers. Artif Intell Med 55(3):197–207
Chang CC, Lin CJ (2011) LIBSVM: a library for support vector machines. ACM Trans Intell Syst Technol 2:1–39
Cohen AM, Hersh WR, Peterson K, Yen P-Y (2006) Reducing workload in systematic review preparation using automated citation classification. J Am Med Inform Assoc 13:206–219
Cohen AM, Ambert K, McDonagh M (2012) Studying the potential impact of automated document classification on scheduling a systematic review update. BMC Med Inform Decis Mak 12(1):33
Defenders of Wildlife (2020) ESAdocs search. https://esadocs.defenders-cci.org. Accessed 22 Feb 2019
Devlin J, Chang M, Lee K, Toutanova K (2018) BERT: pre-training of deep bidirectional transformers for language understanding. Preprint at https://arxiv.org/abs/quantph/1810.04805
Frunza O, Inkpen D, Matwin S, Klement W, O’blenis P (2011) Exploiting the systematic review protocol for classification of medical abstracts. Artif Intell Med 51:17–25
Horspool RN (1980) Practical fast searching in strings. Softw Pract Exp 10(6):501–506
ICF (2015) Document classification and topic extraction resource (DoCTER). https://www.icf-docter.com. Accessed 14 Mar 2020
Ingersoll GS, Morton TS, Farris AL (2013) Taming text: how to find, organize, and manipulate it. Manning Publications Co., New York
Jonnalagadda S, Petitti D (2013) A new iterative method to reduce workload in systematic review process. Int J Comput Biol Drug Des 6:5–17
Kaur H, Pannu HS, Malhi AK (2019) A systematic review on imbalanced data challenges in machine learning. ACM Comput Surv 52:1–36
Kaushik N, Chatterjee N (2016) A practical approach for term and relationship extraction for automatic ontology creation from agricultural text. ICIT 2016:241–247. https://doi.org/10.1109/ICIT.2016.056
Kowsari K, Jafari Meimandi K, Heidarysafa M, Mendu S, Barnes L, Brown D (2019) Text classification algorithms: a survey. Information 10(4):150
Kulkarni R, Minin E (2021) Automated retrieval of information on threatened species from online sources using machine learning. Methods Ecol Evol. https://doi.org/10.1111/2041-210X.13608
Lee J, Yoon W, Kim S, Kim D, Kim S, So C, Kang J (2019) BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. https://doi.org/10.1093/bioinformatics/btz682
Maiya AS (2020) ktrain: a low-code library for augmented machine learning. Preprint at https://arxiv.org/abs/quantph/2004.10703
Nadeau N, Sekine S (2007) A survey of named entity recognition and classification. Lingvist Investig 30(1):3–26. https://doi.org/10.1075/li.30.1.03nad
O’Mara-Eves A, Thomas J, McNaught J, Miwa M, Ananiadou S (2015) Using text mining for study identification in systematic reviews: a systematic review of current approaches. Syst Rev 4:5
Palmer D, Hearst M (1997) Adaptive multilingual sentence boundary disambiguation. Comput Linguist 23(2):241–267
Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V (2011) Scikit-learn: machine learning in Python. J Mach Learn Res 12:2825–2830
Platt J (1999) Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. Adv Large Margin Classif 10(3):61–74
Python Software Foundation. Python Language Reference, version 3.6. Available at http://www.python.org. Accessed 15 Mar 2020
Shemilt I et al (2014) Pinpointing needles in giant haystacks: use of text mining to reduce impractical screening workload in extremely large scoping reviews. Res Syn Methods 5(1):31–49
Varghese A, Cawley M, Hong T (2017) Supervised clustering for automated document classification and prioritization: a case study using toxicological abstracts. Environ Syst Decis. https://doi.org/10.1007/s10669-017-9670-5
Varghese A, Hong T, Hunter C, Agyeman-Badu G, Cawley M (2019) Active learning in automated text classification: a case study exploring bias in predicted model performance metrics. Environ Syst Decis. https://doi.org/10.1007/s10669-019-09717-3
Wallace BC, Trikalinos TA, Lau J, Brodley C, Schmid CH (2010) Semi-automated screening of biomedical citations for systematic reviews. BMC Bioinform 11:55
Acknowledgements
We are deeply grateful to Jacob Malcolm of Defenders of Wildlife, who supported this work by providing us the historical repository of ESA documents used in this study from his organization’s website. This work would not have been possible without this data resource. Mr. Malcolm also converted the documents to text format, which was an important preliminary step in this study.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
The authors declare that they have no conflict of interest.
Ethical approval
This study was performed by ICF under a contract with the Electric Power Resources Institute.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Varghese, A., Allen, K., Agyeman-Badu, G. et al. Extraction of mitigation-related text from Endangered Species Act documents using machine learning: a case study. Environ Syst Decis 42, 63–74 (2022). https://doi.org/10.1007/s10669-021-09830-2
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10669-021-09830-2