Traveloka Review & Sentiment Extractor

Give it a Traveloka URL, get back every review as structured JSON — ready to pipe straight into sentiment analysis, a reputation dashboard, or a spreadsheet.

Works across Traveloka's product families from the same input box: hotels, Xperience activities, bus operators, airport transfers and car rentals. The product is detected from the URL path, so you can mix them in one run.

What you get

Per review (recordType: REVIEW):

Field	Notes
text	Review in your chosen locale
originalText	What the guest actually typed, before machine translation
isMachineTranslated	True when text came from Google Translate, not the guest
score, scoreScale	Guest score, normally out of 10
reviewedAt	ISO-8601 UTC, parsed from Traveloka's millisecond timestamp
reviewerName, reviewerStatus	Partly masked by Traveloka; ANONYMOUS is common
photoUrls, videoUrls, mediaCount	Full-size guest media
subRatings	Per-aspect scores when the reviewer gave them
helpfulCount	How many people marked it helpful
propertyReplyText	The operator's public reply, when there is one

Keeping originalText separate from text matters for NLP: scoring a machine translation measures the translator as much as the guest.

Per object (recordType: OBJECT_SUMMARY), one row holding objectName, rating, ratingLabel, totalReviewCount, reviewsCollected, and ratingTags — Traveloka's aspect tags with counts, split into positiveTagIds and negativeTagIds, which is the closest thing to sentiment the platform publishes itself. Filter the dataset on recordType to separate the two shapes.

Filtering and sorting

• Sort newest first, most helpful, or by rating in either direction. Lowest rating first is the fast way to do a complaint audit.
• Rating bands — exceptional (9+), very good (8+), pleasant (6+).
• Format — text only, media only, or both. Useful for dropping score-only ratings before NLP.
• Aspect tags — run once with no tag filter, read ratingTags off the summary record, then re-run filtered to BAD_CLEANLINESS or similar to read only the reviews behind one complaint.

A note about sort

Traveloka's review API takes sort as a bare string. If you send it as an object ({"sortType": ...}) the API answers HTTP 200 and quietly ignores it, so the data looks sorted but is not. This actor always sends the string form; the trap is documented here because it bites anyone rolling their own client.

Xperience reviews work even though Xperience search does not

Traveloka's activity search and detail endpoints sit behind an AWS WAF CAPTCHA and cannot be read headlessly. The review service does not — so if you have an activity URL, this actor reads its reviews normally. What it cannot do is discover activities for you; bring the URLs.

Cost

Per object: 1 aggregate + 1 aspect-tag call + 1 count + one request per page of reviews (20 per page by default). Pulling 200 reviews is about 13 requests.

maxReviewsPerObject is the stopping point; Traveloka pages until hasNext goes false, and the actor de-duplicates by reviewId so an overlapping page cannot inflate your counts.

Output record types

• REVIEW — one guest review.
• OBJECT_SUMMARY — rating, totals and aspect tags for one object.
• NOT_FOUND — no reviews and no rating summary; usually the product type does not match the URL.
• GATED — an endpoint answered with an AWS WAF challenge.
• ERROR — a request failed; see error.

A note on robots.txt

Traveloka's robots.txt disallows /api/, which is where this actor reads from. Decide for yourself whether that fits your use of the data, your jurisdiction and your agreement with Traveloka. Reviews are personal content written by real people — keep request rates modest, and mind your obligations when you store or republish them.