The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 31 new columns ({'hand_graded_deflection_count', 'name', 'hand_graded_deflection_ambiguous', 'corpus_review_count', 'reply_rate_negative_ci95_high', 'category', 'reply_rate_all_n', 'reply_rate_negative_n', 'chain', 'hand_graded_remedy_pct', 'google_score', 'hand_graded_apology_count', 'mean_stars_corpus', 'hand_graded_n', 'reply_rate_negative_ci95_low', 'hand_graded_remedy_ambiguous', 'one_star_count', 'reply_rate_all_ci95_low', 'hand_graded_remedy_count', 'hand_graded_generic_pct', 'hand_graded_generic_ambiguous', 'hand_graded_apology_pct', 'reply_rate_all_ci95_high', 'reply_rate_all', 'hand_graded_deflection_pct', 'city', 'hand_graded_apology_ambiguous', 'google_review_count', 'two_star_count', 'hand_graded_generic_count', 'placeId'}) and 2 missing columns ({'shop_count', 'replied_count'}).
This happened while the csv dataset builder was generating data using
hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset/dataset.csv (at revision 01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5), ['hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset@01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5/county-benchmarks.csv', 'hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset@01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5/dataset.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
placeId: string
name: string
city: string
county: string
category: string
chain: bool
google_score: double
google_review_count: int64
corpus_review_count: int64
mean_stars_corpus: double
one_star_count: int64
two_star_count: int64
negative_review_count: int64
reply_rate_all: double
reply_rate_all_n: int64
reply_rate_all_ci95_low: double
reply_rate_all_ci95_high: double
reply_rate_negative: double
reply_rate_negative_n: int64
reply_rate_negative_ci95_low: double
reply_rate_negative_ci95_high: double
hand_graded_n: int64
hand_graded_apology_count: int64
hand_graded_apology_pct: double
hand_graded_apology_ambiguous: int64
hand_graded_remedy_count: int64
hand_graded_remedy_pct: double
hand_graded_remedy_ambiguous: int64
hand_graded_generic_count: int64
hand_graded_generic_pct: double
hand_graded_generic_ambiguous: int64
hand_graded_deflection_count: int64
hand_graded_deflection_pct: double
hand_graded_deflection_ambiguous: int64
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 4939
to
{'county': Value('string'), 'shop_count': Value('int64'), 'negative_review_count': Value('int64'), 'replied_count': Value('int64'), 'reply_rate_negative': Value('float64')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 31 new columns ({'hand_graded_deflection_count', 'name', 'hand_graded_deflection_ambiguous', 'corpus_review_count', 'reply_rate_negative_ci95_high', 'category', 'reply_rate_all_n', 'reply_rate_negative_n', 'chain', 'hand_graded_remedy_pct', 'google_score', 'hand_graded_apology_count', 'mean_stars_corpus', 'hand_graded_n', 'reply_rate_negative_ci95_low', 'hand_graded_remedy_ambiguous', 'one_star_count', 'reply_rate_all_ci95_low', 'hand_graded_remedy_count', 'hand_graded_generic_pct', 'hand_graded_generic_ambiguous', 'hand_graded_apology_pct', 'reply_rate_all_ci95_high', 'reply_rate_all', 'hand_graded_deflection_pct', 'city', 'hand_graded_apology_ambiguous', 'google_review_count', 'two_star_count', 'hand_graded_generic_count', 'placeId'}) and 2 missing columns ({'shop_count', 'replied_count'}).
This happened while the csv dataset builder was generating data using
hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset/dataset.csv (at revision 01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5), ['hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset@01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5/county-benchmarks.csv', 'hf://datasets/PixelLabsLLC/west-georgia-auto-shop-review-response-dataset@01bde35c6bf4ce93e8f6a00def2cd83bd5417ba5/dataset.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
county string | shop_count int64 | negative_review_count int64 | replied_count int64 | reply_rate_negative float64 |
|---|---|---|---|---|
Paulding | 19 | 172 | 96 | 0.5581 |
Haralson | 9 | 56 | 2 | 0.0357 |
Carroll | 21 | 235 | 44 | 0.1872 |
Douglas | 18 | 226 | 104 | 0.4602 |
Heard | 1 | 3 | 3 | 1 |
ALL_FIVE_COUNTIES | 68 | 692 | 249 | 0.3598 |
Carroll | null | 4 | null | 0 |
Carroll | null | 2 | null | 0 |
Carroll | null | 5 | null | 0.2 |
Carroll | null | 9 | null | 0.8889 |
Carroll | null | 49 | null | 0 |
Carroll | null | 10 | null | 0.8 |
Carroll | null | 2 | null | 1 |
Carroll | null | 5 | null | 1 |
Carroll | null | 17 | null | 0.2353 |
Carroll | null | 7 | null | 0.7143 |
Carroll | null | 40 | null | 0 |
Carroll | null | 0 | null | null |
Carroll | null | 4 | null | 0.25 |
Carroll | null | 14 | null | 0.1429 |
Carroll | null | 11 | null | 0 |
Carroll | null | 2 | null | 1 |
Carroll | null | 5 | null | 0.8 |
Carroll | null | 12 | null | 0.1667 |
Carroll | null | 11 | null | 0 |
Carroll | null | 25 | null | 0 |
Carroll | null | 1 | null | 0 |
Douglas | null | 3 | null | 0.3333 |
Douglas | null | 19 | null | 0.9474 |
Douglas | null | 4 | null | 0 |
Douglas | null | 20 | null | 0 |
Douglas | null | 3 | null | 1 |
Douglas | null | 0 | null | null |
Douglas | null | 3 | null | 1 |
Douglas | null | 2 | null | 0.5 |
Douglas | null | 3 | null | 0 |
Douglas | null | 8 | null | 1 |
Douglas | null | 17 | null | 0.8824 |
Douglas | null | 69 | null | 0 |
Douglas | null | 6 | null | 0.3333 |
Douglas | null | 1 | null | 1 |
Douglas | null | 25 | null | 0.92 |
Douglas | null | 16 | null | 1 |
Douglas | null | 13 | null | 0 |
Douglas | null | 14 | null | 0.9286 |
Haralson | null | 10 | null | 0.1 |
Haralson | null | 2 | null | 0 |
Haralson | null | 8 | null | 0 |
Haralson | null | 31 | null | 0 |
Haralson | null | 1 | null | 1 |
Haralson | null | 0 | null | null |
Haralson | null | 4 | null | 0 |
Haralson | null | 0 | null | null |
Haralson | null | 0 | null | null |
Heard | null | 3 | null | 1 |
Paulding | null | 21 | null | 0.2857 |
Paulding | null | 3 | null | 1 |
Paulding | null | 5 | null | 0.2 |
Paulding | null | 3 | null | 0.6667 |
Paulding | null | 18 | null | 1 |
Paulding | null | 1 | null | 0 |
Paulding | null | 17 | null | 0.0588 |
Paulding | null | 0 | null | null |
Paulding | null | 2 | null | 0 |
Paulding | null | 9 | null | 1 |
Paulding | null | 9 | null | 0.4444 |
Paulding | null | 10 | null | 0.9 |
Paulding | null | 3 | null | 1 |
Paulding | null | 2 | null | 0.5 |
Paulding | null | 2 | null | 1 |
Paulding | null | 3 | null | 1 |
Paulding | null | 31 | null | 0.4516 |
Paulding | null | 22 | null | 0.9091 |
Paulding | null | 11 | null | 0 |
West Georgia Auto Shop Review-Response Dataset
Reply rates and hand-graded reply quality for 68 tire and auto repair shops across five counties in West Georgia (Carroll, Paulding, Douglas, Haralson, Heard).
Dataset Summary
This dataset provides aggregated, per-shop metrics on how 68 independently owned and chain tire/auto-repair shops across five West Georgia counties respond to Google reviews. Derived from a corpus of 5,716 Google reviews collected on 2026-08-01, it reports per shop: total and negative (1-2 star) review counts, star rating, reply rate to all reviews and to negative reviews specifically (each with a 95% confidence interval), and a hand-graded breakdown of the 249 owner replies sent to negative reviews across four categories — apology, concrete remedy offered, generic/template language, and deflection. County-level benchmarks are included for the same metrics.
All owner-reply category counts come from an independent human hand-grading pass conducted
2026-08-07 that corrected an earlier automated classifier's output; the classifier's own
measured accuracy (81.83% overall) is reported in METHODOLOGY.md for transparency, but its
raw output is not used in any published figure. No verbatim review text, reviewer-identifying
information, or owner-reply text is included — this is derived/aggregate data only.
This is the same dataset already published on Zenodo (DOI, citable) and Kaggle; this Hugging Face listing exists to reach the ML/NLP practitioner audience specifically — the hand-graded owner-reply category counts and the disclosed, measured classifier-accuracy figures make it a small but clean reference point for anyone building or evaluating a customer-service-reply classifier.
Supported Tasks
The hand_graded_* columns describe ground-truth category rates (apology / concrete remedy /
generic / deflection) for a real-world customer-service-reply classification task, with a
disclosed, measured baseline classifier accuracy (81.83% overall; per-category: apology 94.4%,
remedy 79.1%, generic 78.3%, deflection 75.5%) — useful as a sanity-check reference point when
building or evaluating a similar reply-classification model. Note this dataset itself contains
no raw text (see "What is NOT included" below); it is aggregate/tabular, not a text corpus.
Languages
English (en) — the underlying reviews and replies were in English; no review or reply text is included in this dataset itself.
Dataset Structure
Data Files
| File | Contents |
|---|---|
dataset.csv |
68 rows, one per shop — the core dataset. |
county-benchmarks.csv |
6 rows — per-county and five-county aggregate reply-rate benchmarks. |
METHODOLOGY.md |
Full collection/grading methodology, correction history, validation status per figure type. |
Data Fields (dataset.csv)
- Identity/location:
placeId,name,city,county,category,chain - Google listing:
google_score,google_review_count - Corpus review counts:
corpus_review_count,mean_stars_corpus,one_star_count,two_star_count,negative_review_count - Reply-rate metrics (all reviews and negative-only), each with a 95% Wilson confidence
interval:
reply_rate_all,reply_rate_all_n,reply_rate_all_ci95_low,reply_rate_all_ci95_high,reply_rate_negative,reply_rate_negative_n,reply_rate_negative_ci95_low,reply_rate_negative_ci95_high - Hand-graded owner-reply category breakdown (count, percentage, ambiguous-judgment count
per category):
hand_graded_n,hand_graded_apology_count,hand_graded_apology_pct,hand_graded_apology_ambiguous,hand_graded_remedy_count,hand_graded_remedy_pct,hand_graded_remedy_ambiguous,hand_graded_generic_count,hand_graded_generic_pct,hand_graded_generic_ambiguous,hand_graded_deflection_count,hand_graded_deflection_pct,hand_graded_deflection_ambiguous
Data Fields (county-benchmarks.csv)
county, shop_count, negative_review_count, replied_count, reply_rate_negative — one
row per county (Paulding, Haralson, Carroll, Douglas, Heard) plus one ALL_FIVE_COUNTIES
aggregate row.
Dataset Creation
Source Data
Google Maps business listings and their public reviews, collected via a paid Apify Google-Maps-reviews scraper run on 2026-08-01. 92 candidate listings were scraped; 68 met all inclusion criteria (auto/tire service category, inside the five-county footprint, not permanently closed, not parts-retail-only, at least 10 reviews in the scraped sample) and form the full population in this dataset.
Annotations
The 249 owner replies to 1-2 star reviews were first tagged by an automated classifier, then
every one of the 249 was individually hand-graded by a human (Claude Code, under human
direction) on 2026-08-07 against four category definitions (apology, concrete remedy, generic/
template, deflection), reading each reply's text before consulting the classifier's label. 27
of the 249 judgments were flagged ambiguous. The hand_graded_*_count and hand_graded_*_pct
columns in dataset.csv are this hand-graded ground truth, not the classifier's raw output.
Full annotation methodology, including two disclosed public self-corrections, is in
METHODOLOGY.md.
What is NOT included
- Verbatim review text (any star rating).
- Verbatim owner-reply text.
- Reviewer names or any reviewer-identifying profile information.
- Any output from a separate, unrelated complaint-topic classifier (wait time, rudeness, workmanship, upselling, etc.) — that classifier is flagged internally as under-validated and is barred from external distribution project-wide.
Licensing
CC-BY 4.0 — permits reuse with attribution.
Citation
If this dataset is useful, please cite it via its Zenodo DOI:
Williams, Cole (Pixel Labs LLC). West Georgia Auto Shop Review-Response Dataset. Zenodo. https://doi.org/10.5281/zenodo.21899331
Related Links
- Zenodo (citable DOI): https://doi.org/10.5281/zenodo.21899331
- Kaggle mirror: https://www.kaggle.com/datasets/pixellabsllc/west-georgia-auto-shop-review-response-dataset
- Source study 1 — We Read 5,716 Google Reviews of West Georgia Auto Shops
- Source study 2 — West Georgia Owner-Reply Study
- Shop comparison page (rimsandtires.net, the West Georgia tire/auto shop directory this dataset was built from)
Provenance / Disclosure
Built by Cole Williams / Pixel Labs LLC (operator of rimsandtires.net, a West Georgia tire and auto shop directory) from data already collected and published in aggregate form on the site above. The site has a disclosed commercial interest (it lists many of the shops in this dataset) — stated here for the same transparency reason it is stated on the source pages, on Zenodo, and on Kaggle.
- Downloads last month
- 31