遇见数据集

What reviews of iPhone step counters talk about

收藏
Zenodo2026-08-31 更新2026-10-01 收录
官方服务:

资源简介:

What people complain about and praise in iPhone step counter reviews, counted. 7,859 unique reviews of ten apps on the US App Store were read by a hand-written dictionary of phrasings and counted against 21 themes. The dataset holds counts only — no review text, no author, no rating of any single review. Nothing here says whether an app is good; it says what its reviewers chose to write about. What is inside. 7,859 unique reviews (8,492 collected, de-duplicated) of 10 apps from the US App Store, collected on 19 August 2026 and counted against 21 themes. The reviews split into 1,101 at 1-2 stars, 1,274 at 3-4 stars and 5,484 at 5 stars. The file holds 630 per-app rows and 63 rolled-up rows — counts only, with no review text, no author names and no per-review records. Method. Reviews were collected on 2026-08-19 from Apple's public customer-reviews feed, two sort orders per app, and de-duplicated by review id: 8,492 records collected, 7,859 unique. Themes are matched by a hand-written dictionary of phrasings, not by a model; one review can match several themes or none. Polarity compares the 1-2 star band against the 5 star band; the 3-4 star band is published but excluded from that comparison. ⚠️ This is not a random sample and the caps are Apple's, not ours — see the sampling and selection fields, which travel inside this file for that reason. The full method, including how the ten apps were chosen and one defect found and fixed after publication of the numbers, is written up at https://angelstep.app/data/review-themes and its figures are recomputed from the raw corpus by a checker rather than typed by hand. What it is not. No app is ranked, scored or recommended here. What is published under CC BY 4.0 is the compilation, its structure, the theme dictionary and the described method; not claimed is the reviews themselves, which were written by App Store customers and are not republished here in any form — this dataset contains counts only, no review text, no author names and no per-review records. Canonical home: https://angelstep.app/data/review-themes

提供机构:
Zenodo
创建时间:
2026-08-31
二维码
社区交流群
二维码
科研交流群
商业服务