You can deploy this model and start earning money today!

qxswy/
distilbert-sports-sentiment-v2

Name: distilbert-sports-sentiment-v2
Availability: InStock

distilbert/distilbert-base-uncased

finetune

apache-2.0

Views

No views yet

distilbert-sports-sentiment-v2

Binary sentiment classifier for Sports & Outdoors product reviews. Fine-tuned as part of a methodology for analyzing consumer value perception — a domain transfer experiment extending v1 research from board games to fitness equipment.

Performance

Evaluated on an unbalanced held-out test set (original class distribution: ~77% positive, ~23% negative):

	Negative	Positive	Macro avg
Precision	0.94	0.99	0.97
Recall	0.96	0.98	0.97
F1	0.95	0.99	0.97
Accuracy			0.98

Training

Base model: distilbert-base-uncased (67M parameters)
Dataset: Amazon Reviews 2023, Sports & Outdoors — Fit Simplify Resistance Loop Exercise Bands (14,766 reviews → 6,000 after balanced sampling)
Labels: 4–5 stars = positive, 1–3 stars = negative
Technique: Oversampling of minority class (negative) to balance training set

Usage

python

from transformers import pipeline

pipe = pipeline(
    "text-classification",
    model="qxswy/distilbert-sports-sentiment-v2"
)

reviews = [
    "These bands are incredibly durable, love them!",
    "Snapped after two weeks, complete waste of money."
]
print(pipe(reviews))

Context

Part of a two-model NLP pipeline for value proposition analysis through consumer reviews. Paired with qxswy/modernbert-sports-ner-v2 for named entity recognition of value components.

See also: v1 models trained on board game reviews — qxswy/deberta-card-sentiment and qxswy/roberta-card-ner.