Papers by Elizabeth A. Olson
Extraction of Texters’ Explicit Emotion Expressions in Crisis Conversations (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for extracting present and past personal emotion expressions from text-based crisis conversations are lacking in clinically relevant areas. |
| Approach: | They propose a method for extracting present and past personal emotion expressions from text-based crisis conversations and train a transformer-based model that captures contextual distinctions between true personal emotion and other mentions. |
| Outcome: | The proposed method outperforms a regex and a model trained on real conversation data and achieves an F1 score of 0.856. |
Assessing effective de-escalation of crisis conversations using transformer-based models and trend statistics (2025.emnlp-main)
Copied to clipboard
| Challenge: | a lack of quantitative approaches to assess emotion in crisis conversations hinders the science of crisis intervention. |
| Approach: | They propose a transformer-based emotional valence scoring model that measures emotion in crisis conversations . they compare numerical emotional vs. verbal valencies to a corpus of hand-scored social media messages . |
| Outcome: | The proposed model outperforms dictionary-based tools and an LLM in matching scores from human annotators. |