Dataset Viewer
Auto-converted to Parquet Duplicate
asin
large_stringlengths
10
10
facet
large_stringlengths
2
42
pos
int64
0
31
neg
int64
0
20
neu
int64
0
6
total
int64
1
31
0000306045
included features
1
0
0
1
0000306045
portability
4
0
0
4
0000306045
repairability
1
0
0
1
0006530818
accuracy
1
2
0
3
0006530818
balance
0
1
0
1
0006530818
content
4
0
0
4
0006530818
content depth
0
1
0
1
0006530818
depth
1
0
0
1
0006530818
objectivity
1
2
0
3
0006530818
pacing
0
1
0
1
0006530818
perspective
0
3
0
3
0006530818
relevance
0
1
0
1
0006530818
writing quality
2
2
0
4
000714699X
atmosphere
1
0
0
1
000714699X
character development
2
3
0
5
000714699X
characters
0
1
0
1
000714699X
content
3
0
0
3
000714699X
engagement
1
0
0
1
000714699X
length
0
1
0
1
000714699X
narrative structure
1
1
0
2
000714699X
pacing
0
7
0
7
000714699X
plot
0
3
0
3
000714699X
plot realism
0
1
0
1
000714699X
premise
1
0
0
1
000714699X
readability
1
0
0
1
000714699X
research
1
0
0
1
000714699X
story substance
0
1
0
1
000714699X
supporting characters
1
0
0
1
000714699X
themes
1
1
0
2
000714699X
writing quality
7
0
0
7
000714699X
writing skills
1
0
0
1
000714699X
writing style
1
1
0
2
0007243251
character depth
0
1
0
1
0007243251
character development
0
1
0
1
0007243251
characters
0
1
0
1
0007243251
content
1
1
0
2
0007243251
editing quality
0
1
0
1
0007243251
ending
0
1
0
1
0007243251
engagement
1
0
0
1
0007243251
length
0
1
0
1
0007243251
pacing
0
3
0
3
0007243251
plot
3
2
0
5
0007243251
plot resolution
0
1
0
1
0007243251
writing quality
0
1
0
1
0007247729
author
1
0
0
1
0007247729
author credibility
1
0
0
1
0007247729
author expertise
1
0
0
1
0007247729
authorial voice
0
1
0
1
0007247729
completeness
0
1
0
1
0007247729
content
2
0
1
3
0007247729
content depth
2
1
0
3
0007247729
content focus
1
0
0
1
0007247729
depth of content
0
1
0
1
0007247729
detail
0
1
0
1
0007247729
editing
0
1
0
1
0007247729
educational value
1
0
0
1
0007247729
engagement
1
0
0
1
0007247729
historical accuracy
2
1
0
3
0007247729
illustrations
1
0
0
1
0007247729
narrative tone
0
1
0
1
0007247729
objectivity
0
1
0
1
0007247729
pacing
2
2
0
4
0007247729
readability
1
0
0
1
0007247729
research quality
2
0
0
2
0007247729
sophistication
0
1
0
1
0007247729
structure
0
1
0
1
0007247729
target audience
1
0
0
1
0007247729
thematic analysis
1
0
0
1
0007247729
tone
0
1
0
1
0007247729
writing style
6
1
0
7
0007259336
affordability
1
0
0
1
0007259336
content
1
0
0
1
0007259336
content quality
1
1
0
2
0007259336
creativity
0
1
0
1
0007259336
ease of use
1
0
0
1
0007259336
instructions
1
0
0
1
0007259336
originality
1
0
0
1
0007259336
readability
1
0
0
1
0007259336
recipe variety
2
0
0
2
0007259336
reliability
1
0
0
1
0007259336
simplicity
1
0
0
1
0007259336
suitability
1
0
0
1
0007259336
suitability for families
1
0
0
1
0007259336
utility
1
0
0
1
0007259336
variety
1
0
0
1
0007259336
visuals
2
0
0
2
0007441223
appeal
1
0
0
1
0007441223
music
1
0
0
1
0007441223
readability
1
0
0
1
0007441223
story
1
0
0
1
0007477155
appearance
1
0
0
1
0007477155
binding
0
0
2
2
0007477155
build quality
2
1
0
3
0007477155
color accuracy
0
2
0
2
0007477155
condition
2
0
0
2
0007477155
content
0
0
1
1
0007477155
content quality
2
0
0
2
0007477155
cover art
2
0
0
2
0007477155
design
1
0
0
1
0007477155
dialogue
2
0
0
2
End of preview. Expand in Data Studio

ESCI LLM Aspect Mining (US)

Code on GitHub

A product-intrinsic, polarity-labeled aspect-annotation layer over the us locale of ESCI-S, extracted with a calibrated local LLM (google/gemma-4-12B-it, guided JSON, temperature 0). 8.85M aspects over 3.95M reviews; held-out facet F1 0.604, polarity accuracy 0.9455.

⚠️ No Amazon review text is included. Rows carry derived facets + join keys (asin, review_no, review_md5). Reconstruct the original quote by joining your own copy of ESCI-S on asin+review_no and verifying review_md5.

Full methodology, cost model, and corpus-scale QA (the star cross-check) live in the code repository; this dataset's provenance and licensing are in its DATASHEET.

Configs

  • review_aspects (9.5M rows): asin, review_no, facet, polarity, review_md5.
  • product_aspects (5.4M rows): asin, facet, pos, neg, neu, total.

Scope & limitations

US locale only. Model-generated labels (calibrated, not human gold except the 598-review gold split used for tuning). Single-annotator gold set (no inter-annotator agreement). Aggregate polarity is validated (monotone in stars across all slices); no per-review guarantee.

License & provenance

The cc-by-4.0 license applies to the derived annotation layer (the facet phrases and polarity labels) — the author's contribution. Use it freely, including commercially, with attribution (see the citation below).

It does not and cannot grant rights over the underlying review text, which is the property of Amazon and its users and is not included in this dataset. The reviews were obtained via ESCI-S, an unofficial January-2023 scrape, which this dataset neither mirrors nor endorses. If you reconstruct original quotes by joining ESCI-S, you are responsible for your own lawful access to and use of that source text. Full statement: DATASHEET.

Downloads last month
34