Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Sub-tasks:
sentiment-classification
Languages:
Czech
Size:
1K - 10K
License:
| annotations_creators: | |
| - found | |
| language: | |
| - cs | |
| language_creators: | |
| - found | |
| license: | |
| - cc-by-nc-sa-3.0 | |
| multilinguality: | |
| - monolingual | |
| pretty_name: Czech Facebook comments | |
| size_categories: | |
| - 10K<n<100K | |
| source_datasets: | |
| - original | |
| tags: [] | |
| task_categories: | |
| - text-classification | |
| task_ids: | |
| - sentiment-classification | |
| # Dataset Card for Czech Facebook comments | |
| ## Dataset Description | |
| The dataset contains user comments from Facebook. Each comment contains text, sentiment (positive/negative/neutral). | |
| The dataset has in total (train+validation+test) 6,600 reviews. The data is balanced. | |
| ## Dataset Features | |
| Each sample contains: | |
| - `comment_id`: unique string identifier of the comment. | |
| - `sentiment_str`: string representation of the rating - "pozitivní" / "neutrální" / "negativní" | |
| - `sentiment_int`: integer representation of the rating (1=positive, 0=neutral, -1=negative) | |
| - `comment`: the string of the comment | |
| ## Dataset Source | |
| The data is a processed adaptation of [Facebook CZ Corpus](https://liks.fav.zcu.cz/sentiment/). | |
| This adaptation is label-balanced. | |