Linguistically Aware Model Routing for Financial Sentiment Analysis

Spring 2026 CSCI 5541 NLP: Class Project - University of Minnesota

TrustTheTokens

Large Model (RoBERTa-large) Results

1. Baseline RoBERTa-large (Zero-Shot) — PhraseBank (Test)

Baseline RoBERTa-large — PhraseBank confusion matrix Baseline RoBERTa-large — PhraseBank per-class report

2. Baseline RoBERTa-large (Zero-Shot) — Combined (Test)

Baseline RoBERTa-large — Combined confusion matrix Baseline RoBERTa-large — Combined per-class report

3. Fine-Tuned RoBERTa-large — PhraseBank (Validation)

Fine-tuned RoBERTa-large — PhraseBank validation confusion matrix Fine-tuned RoBERTa-large — PhraseBank validation per-class report

4. Fine-Tuned RoBERTa-large — PhraseBank (Test)

Fine-tuned RoBERTa-large — PhraseBank test confusion matrix Fine-tuned RoBERTa-large — PhraseBank test per-class report

5. Fine-Tuned RoBERTa-large — Combined Dataset (Validation)

Fine-tuned RoBERTa-large — Combined validation confusion matrix Fine-tuned RoBERTa-large — Combined validation per-class report

6. Fine-Tuned RoBERTa-large — Combined Dataset (Test)

Fine-tuned RoBERTa-large — Combined test confusion matrix Fine-tuned RoBERTa-large — Combined test per-class report