Modelpublic

Zhoyang/yelp-sentiment-roberta-large

sourceHugging Facemitupdated 4mo agoView on Hugging Face

0likes27downloads

Model Card

yelp-sentiment-roberta-large

RoBERTa-large fine-tuned for 3-class sentiment classification on Yelp reviews.

Weighted F1: 0.9389 | Accuracy: 93.76% Evaluated on 40,000 held-out samples.

Class	Precision	Recall	F1
Negative	0.9393	0.9261	0.9327
Neutral	0.6934	0.7661	0.7280
Positive	—	—	0.9731

from transformers import pipeline

classifier = pipeline( "text-classification", model="Zhoyang/yelp-sentiment-roberta-large" )

result = classifier("The food was amazing!")

—Dataset: Yelp Academic Dataset (499,229 reviews → 239,870 training samples)
—Base model: roberta-large (355M parameters)
—Hardware: NVIDIA A4000 via Paperspace Gradient
—Key decisions:
—max_len=512 (largest single-step F1 gain: +0.012)
—Class-weighted loss: Neg 3.0 / Neu 8.0 / Pos 1.5
—7 controlled iterations from BERT-base baseline

Version	Key Change	Weighted F1	Neutral F1
v1	BERT-base baseline	0.881	0.514
v3	RoBERTa-base	0.892	0.573
v4	max_len 256→512	0.904	0.604
v6	RoBERTa-large	0.912	0.612
Final	EDA-driven weights	0.9389	0.647

Developed as part of Georgia Tech OMSA Practicum, Spring 2026 — PathSynch Labs Pod C.