README.md

---
language:
- de
license: cc-by-nc-4.0
tags: 
- Sentiment Analysis
size_categories:
- 10K<n<100K
task_categories:
- text-classification
pretty_name: GermEval 2017 – Shared Task on Aspect-based Sentiment Analysis Dataset
configs:
- config_name: default
  sep: "\t+"
  column_names: ["ID", "Text", "Relevance", "Sentiment", "Aspect:Polarity"]
  data_files:
  - split: train
    path: "train-2017-09-15.tsv"
  - split: test_dia
    path: "test_dia-2017-09-15.tsv"
  - split: test_syn
    path: "test_syn-2017-09-15.tsv"
  - split: dev
    path: "dev-2017-09-15.tsv"
---

# GermEval 2017 Shared Task on Aspect-based Sentiment in Social Media Customer Feedback
This Dataset is from the GermEval Task 2017 - Shared Task on Aspect-based Sentiment in Social Media Customer Feedback
The original website with detailed information about the Task can be found [here](https://sites.google.com/view/germeval2017-absa)

## Format
The data is provided as TSV. There are two Testsets (synchronic and diachronic)

``ID <tab> Text <tab> Relevance <tab> Sentiment <tab> Aspect:Polarity (whitespace separated)``

## Annotation Guidelines
The annotation guidelines in German can be found [here](http://ltdata1.informatik.uni-hamburg.de/germeval2017/Guidelines_DB_v4.pdf).

### Licensing Information
The data is published under a creative commons by nc (4.0) licence

### Citation Information

```
@inproceedings{germevaltask2017,
title = {{GermEval 2017: Shared Task on Aspect-based Sentiment in Social Media Customer Feedback}},
author = {Michael Wojatzki and Eugen Ruppert and Sarah Holschneider and Torsten Zesch and Chris Biemann},
year = {2017},
booktitle = {Proceedings of the GermEval 2017 – Shared Task on Aspect-based Sentiment in Social Media Customer Feedback},
address={Berlin, Germany},
pages={1--12}
}
```