“The #COP28 deal is yet another historic failure.” Multilingual Sentiment Term Extraction on Environmental Sustainability

Anonymous

Abstract: This paper introduces EnviS, a corpus of 5k tweets annotated with sentiment terms in three languages (Italian, English, and Indonesian) for investigating the debate on environmental sustainability in Social Media. We present a framework for the automatic aggregation of span-level annotations that preserves the annotators' perspective, avoiding additional manual intervention, reducing costs, and preserving the quality of the annotations. Furthermore, we ran a battery of baseline experiments using six open-source instruction/chat-based LLMs in zero-shot and few-shot settings, showing the limits of these models in following instructions and providing correct answers for the extraction and classification of sentiment terms.

Paper Type: long

Research Area: Sentiment Analysis, Stylistic Analysis, and Argument Mining

Contribution Types: Data resources

Languages Studied: English, Indonesian, Italian