TartuNLP at EvaLatin 2024: Emotion Polarity Detection
This addresses emotion analysis for historical Latin texts, which is an incremental improvement in a niche domain.
The paper tackled emotion polarity detection in historical Latin texts by creating training data using heuristics and GPT-4, and achieved first place in the EvaLatin 2024 shared task with LLM-generated labels.
This paper presents the TartuNLP team submission to EvaLatin 2024 shared task of the emotion polarity detection for historical Latin texts. Our system relies on two distinct approaches to annotating training data for supervised learning: 1) creating heuristics-based labels by adopting the polarity lexicon provided by the organizers and 2) generating labels with GPT4. We employed parameter efficient fine-tuning using the adapters framework and experimented with both monolingual and cross-lingual knowledge transfer for training language and task adapters. Our submission with the LLM-generated labels achieved the overall first place in the emotion polarity detection task. Our results show that LLM-based annotations show promising results on texts in Latin.