GERestaurant: A German Dataset of Annotated Restaurant Reviews for Aspect-Based Sentiment Analysis

Abstract: This paper introduces GERestaurant, the first publicly available German dataset for sentiment analysis in the restaurant domain that includes annotations for Aspect Category Detection (ACD), Aspect-Specific Sentiment Analysis (ACSA), Target Aspect Sentiment Detection (TASD), and End-to-End ABSA (E2E). We provide a comprehensive baseline evaluation using state-of-the-art Transformer models (e.g., mBART, mT5) and demonstrate the dataset's utility for training and evaluating German ABSA systems.

---

🚀 Overview

This repository contains the official implementation of the baseline models for GERestaurant, a high-quality German dataset for Aspect-Based Sentiment Analysis.

Key Features

• Comprehensive Annotation: Dataset includes 2,125 reviews with fine-grained annotations for four ABSA subtasks.
• Baseline Models: Implementation of ACD, ACSA, TASD, and E2E ABSA using modern Transformer architectures.
• Unified Pipeline: Scripts for data collection, preprocessing, training, and evaluation.
• Reproducible Results: Configurations and notebooks to replicate the results reported in the paper.

📁 Repository Structure

• data/: Dataset splits and preparation notebooks.
• baseline/:
  • ACD/, ACSA/, TASD/, E2E/: Subtask-specific model implementations and evaluations.
  • train_baseline.py: Main entry point for training the models.
  • results_json/: Raw evaluation results.
• 01_collect_pages_restaurants.ipynb - 07_convert_restaurants_to_latex.ipynb: Data collection and exploratory analysis pipeline.

🛠️ Setup & Usage

Installation

1. Clone the repository:

   git clone https://github.com/nils-hellwig/GERestaurant.git
   cd GERestaurant
   

2. Install dependencies:

   pip install -r requirements.txt
   

Training Baselines

To train a specific baseline model, use the train_baseline.py script. You need to specify the task and the model scale (e.g., base or large):

# Valid tasks: aspect_category, aspect_category_sentiment, end_2_end_absa, target_aspect_sentiment_detection
python baseline/train_baseline.py --task target_aspect_sentiment_detection --model_type base

📜 Citation

@inproceedings{hellwig-etal-2024-gerestaurant,
    title = "{GER}estaurant: A {G}erman Dataset of Annotated Restaurant Reviews for Aspect-Based Sentiment Analysis",
    author = "Hellwig, Nils Constantin  and
      Fehle, Jakob  and
      Bink, Markus  and
      Wolff, Christian",
    editor = "Luz de Araujo, Pedro Henrique  and
      Baumann, Andreas  and
      Gromann, Dagmar  and
      Krenn, Brigitte  and
      Roth, Benjamin  and
      Wiegand, Michael",
    booktitle = "Proceedings of the 20th Conference on Natural Language Processing (KONVENS 2024)",
    month = sep,
    year = "2024",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.konvens-main.14/",
    pages = "123--133"
}