Ekşi Sözlük Turkish Social Opinion Dataset — Four Times a Day
Ekşi Sözlük is Turkey's best-known collaborative forum, where people write short entries under shared topics. This file captures every new entry on your tracked topics, four times a day — 16 columns covering topic, entry text, timestamp, author handle, page number and the keyword that captured it.
It is where Turkey reacts first. In the supplied sample, seven entries about a wildfire near Antalya arrived within three hours of each other, alongside Vestel's slide in quality and Apple account-security worries.
The file already exists, because this pipeline runs four times a day whether you buy it or not. Most vendors start collecting after you order — which is why they quote a lead time. Here the most recent file is in your account within minutes of payment, with an API key issued at the same time.
What you get, in plain terms
Four things. Unprompted Turkish opinion on brands, places and events — the reaction layer that news coverage misses.
Topic and keyword on every entry
Forum Name is the Ekşi topic — a city, a brand, an event — and Keyword the tracking term that captured it. They matched on 31 of 34 rows.
Timestamped to the minute
Every entry carries its posting time, so you can see reaction build — the Antalya fire went from first report to evacuation posts in under three hours.
Short, opinionated text
A median of 19 words per entry, up to 365 — the format that makes Ekşi fast to read and easy to classify.
However you want it
Direct download, REST API, Amazon S3, Google Cloud, Snowflake or SFTP. The API key comes with the dataset.
Fields included in this dataset
All 16 columns, in the exact order they appear in the file — taken straight from the sample, not from a brochure. The free sample ships with a data dictionary giving an example value for each one.
Sample rows from the real file
Real rows from the sample file, not an illustration. The supplied sample has 34 entries with all 16 columns.
Coverage
Captured four times a day across eksisozluk.com for the tracked topics. One row is one entry, identified by its Thread ID; Page Number is the page of the topic it sits on.
| Metric | In the supplied sample | What it tells you |
|---|---|---|
| Entries | 34 in 8 h 20 min | 7 to 8 September 2026 |
| Topics | 11 | Ankara 13 · Antalya 7 · Izmir 5 · brands and places |
| Distinct authors | 33 | Almost every entry from a different writer |
| Median entry length | 19 words | Up to 365 |
| Keyword matches topic | 31 of 34 | The rest differ only in spelling or case |
| Default avatar | 5 of 34 | The rest link a profile picture |
Tracking your own brands, products or competitors? Send the list of topics and we will price the file for them.
Historical data
The last 30 days come with the dataset. Beyond that we hold Ekşi Sözlük records from February 2025 onwards.
Ask about historical dataNeed more data points?
We can extend this dataset beyond the standard 16 columns — a sentiment score and an English translation are the most requested additions, along with pseudonymised author IDs.
Request custom fields“Ekşi told us how Turkish customers felt about our brand weeks before our survey did — in their own words.”
What people use this dataset for
Brand and CX teams
See every public complaint or mention of your brand, and what exactly went wrong, the day it is posted.
Competitive intelligence
Compare complaint themes and volumes across competitors in the same category.
Market research
Read unprompted Turkish consumer opinion on products, places and services at scale.
Data and AI teams
Build Turkish-language sentiment, topic and intent models on real, timestamped posts.
About Ekşi Sözlük data
Ekşi Sözlük — “sour dictionary” — has been one of Turkey's most influential online communities for more than two decades. Writers post short entries under shared topics, from cities and brands to breaking news, and its topic list is widely read as a barometer of Turkish public opinion.
What was in the sample
Thirteen entries on Ankara as the weather turned cold overnight; seven on a wildfire near Antalya, from “ash is falling on the city” to neighbourhoods being evacuated; five on everyday life in Izmir; and brand entries on Vestel — one writer blaming its decline squarely on quality — and on an Apple account-security scare.
Why four times a day
Ekşi reacts in hours. Capturing every six hours means a brand problem or a local event is in your file the same day, with the timestamp to show how quickly it spread.
Two things to plan for
Entries are in informal Turkish, often without Turkish characters, with slang and profanity — build or buy a model trained on it. Hash Id identifies the topic rather than the entry (11 values across 34 rows); use Thread ID as the unique key. The file carries public usernames and avatars; ask for the pseudonymised version if you do not need them.
Is collecting this data legal?
We collect only publicly visible posts, respect robots.txt and platform terms, and align with GDPR, CCPA and Turkey's KVKK. Posts carry the public username, profile link and avatar their authors chose to show; on request we deliver the file pseudonymised, with usernames hashed and profile links removed. The standard licence covers internal analysis such as brand monitoring and research, not republishing. Each row carries its permalink and timestamp as documented provenance, so your legal team can review the source before you buy.
Frequently asked questions
Forum Name is the topic, Conversation Stream the entry, and Page Number where it sits in the topic.
