Natural Language Processing as a tool to evaluate emotions in conservation conflicts

• Published in: Biological Conservation
• Main Authors: Arbieu, Ugo, Helsper, Kathrin, Dadvar, Maral, Mueller, Thomas, Niamir, Aidin
• Other Authors: Senckenberg Biodiversity and Climate Research Centre, Senckenberganlage 25, 60325 Frankfurt am Main, Germany, Web-based Information Systems and Services, Stuttgart Media University, Nobelstrasse 8, 70569 Stuttgart,Germany, This study was supported by the Robert Bosch Foundation
• Format: Article in Journal/Newspaper
• Language: English
• Published: HAL CCSD 2021
• Subjects: Artificial Intelligence Conservation conflicts Emotion Detection Framing effects Humancarnivore coexistence Media Content Analysis Sentiment analysis; Artificial Intelligence; Conservation conflicts; Emotion Detection; Framing effects; Humancarnivore coexistence; Media Content Analysis; Sentiment analysis; [SDV]Life Sciences [q-bio]; [SHS]Humanities and Social Sciences; Canis lupus
• Online Access: https://hal.science/hal-04087665; https://hal.science/hal-04087665/document; https://hal.science/hal-04087665/file/Arbieu_BiolCons_2021.pdf; https://doi.org/10.1016/j.biocon.2021.109030

International audience Conservation conflicts involve an important emotional component that has been little investigated so far. In particular, the study of emotions involved in conservation conflicts have been limited in their scope (e.g. negative vs. positive sentiments) by the language studied (mostly English). Natural Language Processing (NLP) approaches combined with informed expert knowledge constitute a promising tool to analyze the occurrence of discrete emotions in textual contents. We applied NLP on the textual content of German print and online news publications featuring the return of wolves (Canis lupus) in the Saxony region of Germany (n=3,096), to investigate the occurrence of 7 basic emotions and their associations with 9 stakeholder groups. All emotions could be detected with different accuracy levels, and except for interest that was detected in all articles, negative emotions tended to dominate the articles' emotional content (anger in 74% and fear in 36% of the articles). Anger was most frequently detected in articles featuring farmers (20%) and hunters (15%), as was fear (17% with farmers, 14% with hunters). Our study of a diverse set of emotions, in local language, demonstrates the usefulness of NLP approaches to broaden the understanding of local conservation conflicts, in which the news organizations have a pivotal, yet often neglected role.