Zhoyang/yelp-sentiment-3class-eval

Yelp three-class sentiment: public 500-review reference set

I published this fixed 500-review subset of Yelp/yelp_review_full to make evaluation of my RoBERTa-large model reproducible. It preserves both three-class sentiment labels and original 1–5-star ratings. Source-row IDs refer to the upstream dataset's test split.

My original training-run evaluation used a 40,000-review held-out set and achieved 93.76% accuracy / 0.9389 weighted F1. All measured tables on this page use the 500 public reviews, with a raw baseline of 92.40% accuracy / 0.9201 weighted F1. This benchmark does not reproduce the original 40,000-review evaluation.

Composition and construction

I built the subset by scanning source test rows in order and filling fixed star quotas: 85 one-star, 30 two-star, 50 three-star, 85 four-star, and 250 five-star reviews. This provides a fixed reference mix rather than a random sample of the full test split.

Label	Meaning	Count	Share
0	Negative, 1–2 stars	115	23%
1	Neutral, 3 stars	50	10%
2	Positive, 4–5 stars	335	67%

I checked all star-to-class mappings and confirmed 500 unique source IDs and 500 distinct exact review texts. Near duplicates and overlap with the original training data have not been ruled out.

Columns

Column	Meaning
yelp_test_row_id	Source test-row index
text	Review text
label	Three-class target, 0/1/2
sentiment	Class name matching the model's configured labels
original_stars	Original rating, 1 through 5; evaluation metadata, not a model input

Public 500-row benchmark

I evaluated all 500 reviews in Zhoyang/yelp-sentiment-3class-eval. Raw predictions achieve 92.40% accuracy and 0.9201 weighted F1. OOF (out-of-fold) calibration results use five-fold cross-fitting. Temperature scaling improves NLL and ECE on this sample; temperature plus bias does not improve classification accuracy.

Policy	Accuracy	Weighted F1	Macro F1	NLL	Brier	ECE
Raw softmax	92.40%	0.9201	0.8345	0.2372	0.1200	0.0364
Inverse-weight sensitivity baseline	92.80%	0.9212	0.8300	0.2719	0.1230	0.0494
OOF temperature	92.40%	0.9201	0.8345	0.2184	0.1168	0.0147
OOF temperature + bias	92.20%	0.9176	0.8288	0.2236	0.1188	0.0159

Raw per-class performance on these 500 rows

Class	Support	Precision	Recall	F1
Negative	115	0.9310	0.9391	0.9351
Neutral	50	0.6923	0.5400	0.6067
Positive	335	0.9478	0.9761	0.9618

Raw confusion matrix

Rows are true three-class labels; columns are predictions.

True / predicted	Negative	Neutral	Positive
Negative	108	6	1
Neutral	6	27	17
Positive	2	6	327

Error breakdown by original rating

Original stars	n	Raw errors	Raw error rate	OOF temp+bias error rate
1	85	1	1.18%	1.18%
2	30	6	20.00%	20.00%
3	50	23	46.00%	48.00%
4	85	5	5.88%	5.88%
5	250	3	1.20%	1.20%

Errors are concentrated at the 2–3-star boundary: raw error rates are 20.00% and 46.00%, compared with 1.18% for 1-star and 1.20% for 5-star reviews. These are errors against the three-class sentiment target, not five-star prediction errors.

Human review capacity and held-out threshold routing

I route low-margin predictions to review, using the gap between the two highest class probabilities. Each fold learns its calibrated margin thresholds on the other four folds, then applies them to held-out reviews. The reference budget therefore need not equal the observed flagged share.

Reference budget	OOF flagged share	Auto-route n	Auto-route error rate	Errors in review queue
0%	0.0%	500	7.80%	0
5%	4.8%	476	5.46%	13
10%	9.4%	453	4.86%	17
15%	15.2%	424	3.77%	23
20%	20.0%	400	2.75%	28
25%	24.6%	377	2.12%	31
30%	30.8%	346	1.73%	33

Retrospective batch ranking with an exact review budget

For a fixed batch, ranking by ascending top-two margin selects exactly the requested number of reviews. This answers a different capacity question from reference-threshold routing. Both tables report errors only among automatically routed reviews; reviewer corrections are not measured.

Exact review budget	Review n	Raw auto-route error rate	OOF temp+bias auto-route error rate
0%	0	7.60%	7.80%
5%	25	5.26%	5.26%
10%	50	4.44%	4.67%
15%	75	3.76%	4.00%
20%	100	2.50%	3.00%
25%	125	2.13%	2.13%
30%	150	1.71%	1.71%

Final reference calibration parameters

{
  "temperature": 1.54,
  "logit_bias": [
    -0.2700000000000001,
    -0.5399999999999999,
    0.0
  ]
}

I fit these reference parameters on all 500 reviews for the Space. Their performance on new reviews has not been independently evaluated; the OOF results above evaluate the cross-fitting procedure. The Space defaults to raw scores, with experimental calibration available as an option.

Exploratory four-star review triage

This experiment tests whether an existing sentiment classifier can help prioritize four-star reviews for human inspection of issue mentions.

In the 85 four-star reviews of this reference set, one AI annotator provisionally labeled 20 reviews as containing concrete operational or product issues (Target B) and 36 as containing any explicit complaint or caveat (Target A). At a budget of 20 reviews, ranking by the lowest standalone sentence Positive score retrieved 13 labeled Target B positives, compared with 10 for whole-review scoring and 10 for a fixed negative-word lexicon. A lexicon expanded during development retrieved 12, and ranking by sentence count alone retrieved 9. The sentence method's precision and recall on Target B were both 65% at $K = 20$ on this labeled pool.

Ranking policy (85 four-star reviews)	Role	$K=10$ Target B (A)	$K=20$ Target B (A)	$K=30$ Target B (A)
Sentence minimum P(Pos) ($\ge 12$ chars)	Experimental candidate	8 (9)	13 (18)	16 (24)
Sentence minimum P(Pos) (unfiltered)	Sensitivity check	7 (9)	13 (17)	17 (24)
Whole-review P(Pos)	Comparison baseline	6 (9)	10 (17)	10 (22)
Fixed negative-word lexicon	Cheap comparison baseline	6 (9)	10 (15)	13 (21)
Expanded development lexicon	In-sample rule reference	7 (10)	12 (18)	14 (24)
Sentence count	Length-only baseline	6 (8)	9 (13)	13 (19)

These are exploratory results from the data used to develop the feature. Independent annotation and evaluation on a fresh batch are pending. Some correctly retrieved reviews were triggered by sentences about competitors or unrelated matters, so a low-scoring sentence is a ranking signal rather than a verified issue snippet.

The separate Model Sensitivity Inspector in the Space measures score changes after deleting a selected single sentence or contiguous sentence span. Those changes describe model behavior; they do not estimate customer-rating uplift or the effect of an operational intervention.

Evaluation method

I used model revision 2e39ef29d577aa84cfe70596020eb40e708b88a0 with float32 inference, a 512-token limit, and batch size 8. The public CSV is pinned to dataset revision a2aa98f96cc2e829904f47984747667d3be53663; its SHA-256 is ce6d53e65b0f229a81d807aec6e1899dd9fa8f253d8ea215c5b2873affbfb3eb.

Calibration uses five label-stratified folds and seed 20261020. Each review is excluded from fitting its own calibrator and routing thresholds. I minimize NLL with a coordinate/grid search and apply softmax((log(p_raw) + logit_bias) / temperature), fixing the Positive bias to zero for identifiability. Temperature-only calibration fixes all biases to zero; temperature plus bias shares one temperature across classes.

Class-weighted training can shift class probabilities. The inverse-weight comparison uses temperature 1 and bias [-log(3/1.5), -log(8/1.5), 0]. This is a sensitivity baseline motivated by ideal weighted cross-entropy, rather than an assumed correction for this fitted model.

NLL uses natural logs. Multiclass Brier is the mean sum of squared errors across all three classes. ECE uses ten equal-width bins of top-label confidence. Lower values are better for these three metrics.

Limitations

• The 500 reviews follow fixed star quotas drawn from early source test rows, rather than a random sample of the entire split. Their 23% / 10% / 67% class mix is a reference composition, not an estimate of production prevalence or the original 40,000-review class mix. Source test membership alone does not establish independence from my original training and evaluation splits.
• Small slices, especially the 30 two-star reviews, have substantial sampling uncertainty; ECE is also sensitive to sample size and binning. Business, customer, and timestamp identifiers are unavailable, so this benchmark cannot establish generalization across those groups. Cross-fitting reduces calibration leakage but does not turn an exploratory benchmark into an untouched final test.
• The Space uses calibration and thresholds fitted on the full reference set. Future review rates and accuracy can differ, and calibration is not guaranteed to help. Star-level error rates describe the three-class target, not complaint severity.
• Human reviewer accuracy has not been measured. A perfect-reviewer calculation would be a scenario upper bound. This benchmark measures no churn, retention uplift, revenue, or pricing outcomes.

Reproduction

Load the exact CSV used for this benchmark:

import pandas as pd

revision = "a2aa98f96cc2e829904f47984747667d3be53663"
url = (
    "https://huggingface.co/datasets/Zhoyang/yelp-sentiment-3class-eval/"
    f"resolve/{revision}/test.csv"
)
reviews = pd.read_csv(url)

Use the pinned model revision and inference example in the model card, retaining all three probabilities for each review. Compare predicted class IDs against label; use original_stars only to slice the errors. The evaluation method above defines the calibration transformation, folds, scoring conventions, and routing rule.

Source

Yelp Review Full.