Rakuten Travel Reviews Scraper

Extract structured hotel & ryokan reviews from Rakuten Travel (楽天トラベル) — Japan's largest accommodation review platform. This is a review-only actor: point it at a hotel and get every customer review as clean JSON, with ratings, sub-ratings, stay/posted dates, room & plan info, and hotel replies.

There is no other Rakuten Travel review-only actor on Apify Store. If you already use a Rakuten/Agoda hotel scraper for metadata, this is the matching review extractor.

Run on Apify

This actor is deployed on the Apify platform: console.apify.com/actors/XxwoZr8npNCXBoQI8

• Console: open the actor, fill in the Input tab, and click Start.
• API: POST https://api.apify.com/v2/acts/huggable_quote~rakuten-travel-reviews-scraper/runs?token=<APIFY_TOKEN> with the input JSON as the body.
• CLI: apify call huggable_quote/rakuten-travel-reviews-scraper --input-file=input.json

Results land in the run's default dataset (export as JSON / CSV / Excel from the Console or the API).

What it does

• Collects all reviews for each hotel (or up to a per-hotel cap you set).
• Normalizes Japanese data: full-width → half-width numbers, Japanese dates (2026年5月28日) → ISO-8601 (2026-05-28), clean UTF-8 text with no HTML residue.
• Captures per-review sub-ratings (room, service, location, bath, facilities, cleanliness — plus breakfast/dinner where available).
• Captures hotel-level aggregates (average rating, rating distribution, sub-rating averages) on the first record of each hotel.
• Captures the hotel's reply to each review.

Input

Field	Type	Description
startUrls	array	Rakuten Travel hotel/review URLs. Any URL containing the hotel number works, e.g. https://review.travel.rakuten.co.jp/hotel/voice/5521 or https://travel.rakuten.co.jp/HOTEL/5521/review.html.
hotelIds	array	Hotel numbers (hotelNo) directly, e.g. ["5521"]. Alternative to startUrls.
maxReviewsPerHotel	integer	Max reviews per hotel. 0 (default) = all available.
requestDelayMs	integer	Polite delay between requests (default 1000). Rakuten tolerates ~1 req/s.

Example:

{
  "startUrls": [{ "url": "https://review.travel.rakuten.co.jp/hotel/voice/5521" }],
  "maxReviewsPerHotel": 0,
  "requestDelayMs": 1000
}

Output

One dataset record per review:

{
  "hotelId": "5521",
  "hotelName": "神戸ポートピアホテル",
  "reviewId": "fb2cec799cd8e6a8",
  "reviewUrl": "https://review.travel.rakuten.co.jp/hotel/voice/5521?page=1",
  "reviewerName": "shigeki777",
  "reviewerType": "家族",
  "tripPurpose": "レジャー",
  "reviewerAge": "60代",
  "reviewerGender": "男性",
  "reviewerReviewCount": 1,
  "overallRating": 4.0,
  "ratingScale": 5,
  "subRatings": { "room": 4.0, "service": 4.0, "location": 4.0, "bath": null, "facilities": 4.0, "cleanliness": 4.0 },
  "title": "30年ぶりの宿泊、南館は広くて快適でした",
  "text": "今回、30年ぶりに、宿泊しました。\n南館は…",
  "stayDate": "2026-05",
  "postedDate": "2026-05-28",
  "roomType": "南館サウスリゾートフォース(4名様)【禁煙】",
  "planName": "【事前決済限定】キャンセル・返金不可プラン 11時OUT<素泊まり>",
  "likeCount": null,
  "viewCount": 245,
  "replyFromHotel": "この度はポートピアホテルをご利用いただき…",
  "replyFromHotelDate": "2026-05-30",
  "scrapedAt": "2026-06-09T00:00:00+00:00",
  "hotelStats": {
    "displayedTotalReviewCount": 8136,
    "collectedReviewCount": 311,
    "averageRating": 4.41,
    "ratingDistribution": { "5": 2314, "4": 1490, "3": 261, "2": 101, "1": 65 },
    "subRatingAverages": { "room": 4.43, "service": 4.43, "location": 4.43, "breakfast": 4.57, "dinner": 4.54, "bath": 4.06, "facilities": 4.27, "cleanliness": 4.4 }
  }
}

Notes:

• hotelStats is attached to the first record of each hotel only; it is null on the rest.
• displayedTotalReviewCount is Rakuten's all-time displayed total (legacy お客さまの声). collectedReviewCount is the number actually browsable in the current クチコミ system — this is what the actor can collect.
• stayDate is YYYY-MM (Rakuten exposes only the stay month). postedDate and replyFromHotelDate are YYYY-MM-DD.
• reviewerType = companion type (家族/一人/カップル …); tripPurpose = レジャー/ビジネス.
• Some reviews are photo-only and legitimately have an empty text.
• Reviewer info is limited to the public display nickname Rakuten shows — no personal data is reconstructed.

How it works

Pure Python: httpx fetches the server-rendered review pages and BeautifulSoup parses them (no headless browser needed). The actor stops immediately and logs clearly if it ever sees a block/CAPTCHA — it does not attempt evasion.

Local development

pip install -r requirements.txt
python -m tests.test_live      # live parse + normalization validation
apify run                      # run the actor locally (Apify CLI)
apify push                     # deploy to the Apify platform

Actor icon

The icon lives at .actor/icon.svg. Apify sets the Store icon via the Console (it is not applied by apify push): open the actor → Settings → Display information → Icon and upload .actor/icon.svg.