Spanish Review Classifier — sentiment & complaints (beta)

Part of LeoWorks review analysis tools (same labels, any language): Korean · Japanese · Spanish · Amazon · AliExpress · Google Maps · social comments.

For sellers, dropshippers and AI agents that need to classify Spanish reviews — MercadoLibre reviews from a review scraper, any other Apify dataset, or pasted as text — and need every review labelled with complaint type, sentiment and purchase motive, for $0.50 per 1,000 reviews, with no scraping, prompt writing or LLM key. It is Spanish sentiment analysis built for Mercado Libre reviews (opiniones).

Beta: complaint-type accuracy is below our 85% bar (79% on reviews not used for tuning) — the numbers in Accuracy are measured; rely mainly on sentiment (92–96%).

• Complaint types — delivery, quality/defect, size/fit, price/value, customer service, packaging, no effect, skin/body reaction, other (multi-label, each with a probability)
• Sentiment — positive / neutral / negative
• Purchase motive — price/discount, reviews & reputation, brand, gift, repurchase, unknown
• Your own labels — up to 10 yes/no criteria in plain English or Spanish (e.g. "mentions sound quality", "menciona el olor")

Labels come with English keys and Spanish names (e.g. quality_defect / calidad / defecto, field labelEs).

Use it to: see why MercadoLibre buyers leave 1-star reviews (análisis de reseñas) · compare complaint mix against competitors · track sentiment of your listings · tag Spanish reviews for a dashboard.

Output sample

Real rows from run FGhAkWt9siPPL0VwL (2026-10-12): iPhone 15 reviews from MercadoLibre Mexico — the 41-review check set in Accuracy — with one custom label ("says the phone is refurbished or not new"). English translations are added here for readers; the Actor returns the original text.

text (Spanish)	English (added)	stars	complaint (probability)	sentiment	custom: refurbished / not new
Vino con defectos en las dos cámaras traseras	Both rear cameras came with defects	1	quality_defect (0.96)	negative (1.00)	no (0.04)
Son reconstruidos, obvio me di cuenta hasta que se tuvo que abrir…	They are rebuilt — I only noticed when it had to be opened…	1	quality_defect (0.89)	negative (1.00)	yes (0.85)
Pantalla: la pantalla no es la original está cambiada…	Screen: it is not the original, it was replaced…	1	quality_defect (0.96)	negative (1.00)	yes (0.69)
Se ha bloqueado y no me responden los mensajes.	It got locked and they don't answer my messages.	1	customer_service (0.90)	negative (1.00)	no (0.01)
Batería: el nivel de batería se le baja demasiado rápido 😑…	Battery: the charge drops far too fast…	1	quality_defect (0.90)	negative (1.00)	no (0.11)
2 años ya y todo fino.	2 years now and all good.	5	none (0.98)	positive (1.00)	no (0.04)

Spanish AliExpress reviews work the same way — or use our AliExpress Reviews Scraper, which collects and labels them in one run.

Each row also has labelEs names, the rating and the ID fields you choose, and in full mode complaintScores for every complaint type.

Input example

The form default — two pasted reviews, no dataset needed (about $0.001, 4 seconds):

{
  "texts": [
    "El paquete llegó aplastado y uno de los auriculares no funciona.",
    "Siempre lo compro, excelente calidad. ¡Lo volvería a comprar!"
  ]
}

Beta: accuracy below our 85% bar — see Accuracy. To classify a scraper's output, pass its dataset instead — the text, rating and ID fields are detected automatically:

{
  "datasetId": "YOUR_MERCADOLIBRE_REVIEW_DATASET_ID",
  "customLabels": ["mentions sound quality"]
}

Pricing

Pay only for classified reviews — no subscription.

Event	Price	When
review-judged	$0.0005	One review classified (complaint types, sentiment, purchase motive and any custom labels).

That is $0.50 per 1,000 reviews. First run with the form defaults: about $0.001 (2 reviews, 4 seconds).

Cost examples

Reviews	Cost
100	$0.05
1,000	$0.50
10,000	$5.00
100,000	$50.00

With the free $5 monthly Apify credit you can classify about 10,000 reviews.

Items without review text are skipped and not charged. Reviews that fail after retries are reported with an error field and not charged. If you set a maximum cost per run, the Actor stops cleanly when it is reached.

Works with

Source	Scraper on Apify Store	Text field	Rating	IDs kept
MercadoLibre reviews	MercadoLibre Reviews Scraper (saswave)	content	rating	id, sku
AliExpress reviews	AliExpress Reviews Scraper (leoworks) — already labels reviews itself; use this Actor only for Spanish-specific custom labels	text	rating	reviewId, productId

Field detection was checked against real output of both. Also: the Apify API and JavaScript/Python clients · Apify Schedules · Claude, Cursor and Claude Code through the Apify MCP server (next section).

Use with Claude, Cursor or Claude Code (MCP)

Ask your AI agent. Add the Apify MCP server with this Actor as a tool, then say, for example:

• "Classify these Spanish reviews and tell me the top complaint types: “El paquete llegó aplastado y uno de los auriculares no funciona.” / “Siempre lo compro, excelente calidad. ¡Lo volvería a comprar!”"
• "Are these MercadoLibre reviews about delivery, quality or price? “Tardó 15 días en llegar.” / “Es más caro que en otras tiendas pero la calidad es buena.”"
• "Run a Spanish review sentiment analysis on dataset YOUR_DATASET_ID from my MercadoLibre reviews scraper run and summarize the complaints."

The agent calls the tool leoworks--spanish-review-classifier and reads the labels with get-dataset-items. Replace YOUR_DATASET_ID with the ID of your own dataset.

Claude Desktop or Cursor (mcp.json):

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=leoworks/spanish-review-classifier",
      "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" }
    }
  }
}

Claude Code: claude mcp add --transport http apify "https://mcp.apify.com?tools=leoworks/spanish-review-classifier" --header "Authorization: Bearer YOUR_APIFY_TOKEN". Leave out the header to sign in with OAuth in the browser instead. Your Apify token is in Console → Settings → API & Integrations. We verified this setup with the Apify MCP server (v0.17.3) on 2026-10-08: the agent classified a pasted review in 5.5 seconds (run 9J1ddgbOq59Fl9B7f).

Output (one row per review)

{
  "index": 0,
  "text": "la caja viene quebrada , solo carga 1 audífono, es lamentable, solo porque lo gane en el huerto mágico no debiera venir malo. he canjeado varias cosas y esta es la primera decepción. que pena que no funciona.",
  "labels": {
    "complaint": [
      {
        "label": "quality_defect",
        "labelEs": "calidad / defecto",
        "probability": 0.95
      },
      {
        "label": "packaging",
        "labelEs": "embalaje",
        "probability": 0.91
      }
    ],
    "sentiment": {
      "label": "negative",
      "labelEs": "negativo",
      "probability": 1,
      "confidence": 1
    },
    "motive": {
      "label": "unknown",
      "labelEs": "desconocido",
      "probability": 0.59,
      "confidence": 0.51
    }
  }
}

When no complaint type passes the threshold, complaint is [{ "label": "none", "labelEs": "sin queja" }]. Minimal mode returns label keys only.

Summary by product (REPORT)

Each run also saves a REPORT record (Output tab → Summary by product) at no extra charge: for every product, the complaint rate and complaint mix, sentiment shares, purchase motives, average rating and the 3 strongest complaint reviews — plus the same for all reviews together. Products are grouped by summaryGroupField (detected automatically from fields such as productId, productName or placeId when empty).

{
  "groupField": "productId",
  "groups": [{
    "group": "A",
    "reviews": 3,
    "averageRating": 2.67,
    "complaintRate": 0.667,
    "complaints": [{ "label": "delivery", "count": 1, "share": 0.333 }, { "label": "quality_defect", "count": 1, "share": 0.333 }],
    "sentiment": { "negative": 0.667, "positive": 0.333 },
    "motive": { "unknown": 0.667, "price": 0.333 },
    "exampleComplaints": [{ "complaint": "delivery", "rating": 2, "text": "El envío tardó una semana, demasiado lento" }]
  }]
}

Accuracy

Measured on hand-labelled Spanish reviews (2026-10-07/08). The questions were adjusted on the first set, then checked on two sets of different products.

Set	Products	Reviews (1–2 stars)	Complaint type	Sentiment
Check sets — not used for adjusting	iPhone (MercadoLibre MX) + earbuds (AliExpress ES/LatAm)	91 (74)	79%	93%
— of which	iPhone 15, MercadoLibre MX	41	80%	95%
— of which	Earbuds, AliExpress	50	78%	92%
First set (used for adjusting)	Panettone, MercadoLibre AR	45 (15)	98%	96%

Most misses add a second, broader label ("other") next to the right one, or miss "not as described" cases (used or refurbished item sold as new). Beta until it passes 85% on a new check set. Automated labels can be wrong; check samples before making big decisions.

Limits

Item	Limit
Review length	First 4,000 characters are used
Custom labels	Up to 10, each up to 200 characters
Dataset size	Any — datasets are read in pages of 1,000
Language	Spanish (beta, measured above). Other languages: see our Korean and Japanese classifiers
Speed	About 100 reviews in 5 seconds
Data	Only the text, rating and the ID fields you choose are sent for classification; reviewer names are not output

Run it on a schedule

Daily or weekly, following your scraper: schedule your MercadoLibre or other review scraper with an Apify Schedule and connect this Actor to it with an Actor-to-Actor integration so every finished scrape is classified. This Actor has no date filter, so set the scraper to collect recent reviews only. To start it from other tools, use the Apify n8n or Make integration, or the Apify API.

Related Actors

Scrapers that collect the input (Apify Store, other publishers — this Actor reads their datasets):

Scraper	Collects
MercadoLibre Reviews Scraper (saswave)	MercadoLibre product reviews

More from Leoworks:

Need	Actor
Korean reviews (Coupang, Olive Young, Naver)	Korean Review Classifier
Japanese reviews (Rakuten and others)	Japanese Review Classifier
Amazon reviews	Amazon Review Classifier

FAQ

Why not use an LLM API like ChatGPT directly?

You can, for a handful. For thousands you would write and test prompts, handle rate limits, parse free text and keep the labels consistent between runs. This Actor uses fixed labels with a probability for each, measured accuracy on hand-labelled data, and needs no LLM key: $0.50 per 1,000 reviews.

What are the alternatives to Spanish Review Classifier?

A general sentiment API (usually positive / negative only), an LLM with your own prompts, or reading reviews by hand. This Actor adds complaint type and purchase motive, with the same labels as our Korean, Japanese and Amazon classifiers.

How do I analyze MercadoLibre reviews?

Run a MercadoLibre reviews scraper (see Works with), then pick its dataset in Reviews dataset and start. Every review gets complaint type, sentiment and purchase motive with probabilities; you can also paste review texts instead.

Which AI makes the judgments?

Jev, TypeSafe's decision model (version jev-1.13.0, pinned). Jev answers each label with a calibrated probability instead of generated text, so a 0.9 means the same thing for every item. Answers are stable but not fixed: re-running the same input can change a few borderline answers (under 1% of answers in our repeat tests) — filter on probability when an item must never flip. Only the review text, its rating and your custom labels are sent to Jev; reviewer names and other fields are not.

Why beta?

Complaint-type accuracy on reviews not used for tuning is 79%, below the 85% we require for a measured language. Sentiment is reliable (92–96%).

Does it scrape MercadoLibre?

No. It classifies reviews you already have — run a review scraper first (see Works with) or paste texts.

Disclaimer

Independent tool — not affiliated with, endorsed by or sponsored by MercadoLibre or AliExpress, or by the authors of the scrapers listed above. Names are used only to describe compatible data sources.

Reviews and support

If this Actor saved you time, a short review on Apify Store helps others find it. Questions or a dataset whose fields are not detected? Open an issue in the Issues tab — we answer within a day.

Changelog

See the Changelog tab.