Evaluating the Capabilities of Large Language Models for Multi-label Emotion Understanding
Tadesse Destaw Belay, Israel Abebe Azime, Abinew Ali Ayele, Grigori Sidorov, Dietrich Klakow, Philipp Slusallek, Olga Kolesnikova, Seid Muhie Yimam · 2024
Read Evaluating the Capabilities of Large Language Models for Multi-label Emotion Understanding by Tadesse Destaw Belay, Israel Abebe Azime, Abinew Ali Ayele, Grigori Sidorov, Dietrich Klakow, Philipp Slusallek, Olga Kolesnikova, Seid Muhie Yimam (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-17837.txt.
Description
Tadesse Destaw Belay and a large team tested language models in two thousand twenty-four on reading emotions in four Ethiopian languages. Emotion labels often overlap, since one text can hold joy and anger at once. This multi-label task is little studied in models, above all for low-resource languages.
They built EthioEmo, a data set for Amharic, Afan Oromo, Somali, and Tigrinya, and also tested an English set. They compared large models used cold or with a few examples against smaller tuned models. Even English results fell short, and the gap to low-resource languages was large. Scores varied by language and model type. The data set is shared in public.
Catalog record
| CREATOR | Tadesse Destaw Belay, Israel Abebe Azime, Abinew Ali Ayele, Grigori Sidorov, Dietrich Klakow, Philipp Slusallek, Olga Kolesnikova, Seid Muhie Yimam |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-17837