Douban Reviews Scraper

Scrape public Douban (豆瓣) ratings, long-form reviews and short comments — each tagged with a positive / neutral / negative sentiment label — for movies, TV, books, music and groups. Built for media & entertainment researchers, publishers, recommendation-engine teams and AI-training-data buyers who need structured Chinese cultural-opinion data.

What you get

Paste one or more Douban subject URLs and get a clean dataset with three record types:

• subject — the title, year, overall rating, rating count and genres.
• comment — short user comments (high volume) with star rating + sentiment.
• review — long-form reviews with full body text, star rating + sentiment.

Group URLs produce group_topic records (discussion topics with reply counts).

Sentiment is derived from the author's own Douban star rating — no guesswork, no ML black box: 5–4★ = positive, 3★ = neutral, 2–1★ = negative, unrated = null.

Supported URLs

Type	Example
Movie / TV	https://movie.douban.com/subject/1292052/
Book	https://book.douban.com/subject/1084336/
Music	https://music.douban.com/subject/1407217/
Group	https://www.douban.com/group/beethoven/

Example input

{
    "startUrls": [
        { "url": "https://movie.douban.com/subject/1292052/" }
    ],
    "scrapeShortComments": true,
    "scrapeLongReviews": true,
    "maxCommentsPerSubject": 200,
    "maxReviewsPerSubject": 50,
    "fetchFullReviewText": true,
    "tagSentiment": true,
    "proxyConfiguration": {
        "useApifyProxy": true,
        "apifyProxyGroups": ["RESIDENTIAL"]
    }
}

Example output

A short comment record:

{
    "record_type": "comment",
    "id": "comment:1234567890",
    "subject_id": "1292052",
    "subject_type": "movie",
    "subject_title": "肖申克的救赎 The Shawshank Redemption",
    "author": "影迷小王",
    "author_url": "https://www.douban.com/people/12345/",
    "rating_stars": 5,
    "rating_label": "力荐",
    "sentiment": "positive",
    "content": "希望让人自由。每看一次都有新的感动。",
    "useful_count": 1842,
    "created_at": "2021-03-14T21:05:00+08:00",
    "source_url": "https://movie.douban.com/subject/1292052/comments?status=P&start=0&limit=20&sort=new_score",
    "scraped_at": "2026-06-15T09:12:00.000Z"
}

Field notes:

• id is stable across runs (built from Douban's comment/review id) — use it to deduplicate on your side.
• rating_stars / sentiment are null when the user left a comment without a rating.
• created_at is the original Douban timestamp in China Standard Time (UTC+8); scraped_at is ISO-8601 UTC.
• For long reviews, content_truncated: false means the full essay body was captured (fetchFullReviewText enabled).

Pricing

Pay-per-event — you pay per item actually extracted, so cost scales with value:

Event	Price
Long review (full body + sentiment)	$0.025
Short comment	$0.005
Subject info	$0.006
Group topic	$0.012
Actor start	$0.00005

(Final prices are shown on the Apify Store listing.)

Limitations

• Douban serves a JavaScript proof-of-work anti-bot challenge, so this actor runs a real headless browser (Playwright + Chromium) to clear it. Recommended run settings: 4 GB memory and residential Apify Proxy. The browser solves the challenge automatically and retries on a fresh session if it doesn't clear.
• Douban limits how deep short-comment pagination goes for logged-out access (typically the first few hundred). Set maxCommentsPerSubject realistically.
• A minority of titles (often sensitive ones) gate their short comments behind login entirely for logged-out visitors; for those, the actor still returns the subject info and long reviews, but short comments come back empty. Long reviews and ratings are not gated.
• Keep maxConcurrency modest (default 3). Under heavy concurrency Douban occasionally soft-throttles a page, which can make a single comment page come back empty; lower concurrency avoids this.
• Some music/book subject pages expose fewer fields (e.g. no genres); those come back null.
• Public data only — the actor never logs in or scrapes login-walled content.

FAQ

Which content should I scrape? Toggle scrapeShortComments, scrapeLongReviews and scrapeSubjectInfo independently. Short comments are cheapest and highest-volume; long reviews are richer for sentiment / NLP work.

Can I run this on a schedule? Yes — use Apify Schedules. Reviews are evergreen, so weekly is usually plenty.

How do I export to my DB / Google Sheets? Use Apify Integrations or the Dataset API — every field above is available via /items?format=json|csv|xlsx. The dataset also ships pre-built Short comments and Long reviews table views.

Why is sentiment sometimes null? The user rated nothing, so there's no star signal to map. The raw content is still captured.