Overview

Fine tuned VinAI's BERTweet base model on the Wiki Weasel 2.0 Corpus from the Szeged Uncertainty Corpus for hedge (linguistic uncertainty) detection in social media texts. Model was trained and optimised using Ray Tune's implementation of Deep Mind's Population Based Training with the arithmetic mean of Accuracy & F1 as its evaluation metric.

Labels

• LABEL_1 = Positive (Hedge is detected within text)
• LABEL_0 = Negative (No Hedges detected within text)

<a name="models2"></a> Model Performance