chkla / FrameASt

📄 FrameASt: A Framework for Second-level Agenda Setting in Parliamentary Debates through the Lense of Comparative Agenda Topics

This repository provides you with ressources of the publication 📄 FrameASt: A Framework for Second-level Agenda Setting in Parliamentary Debates through the Lense of Comparative Agenda Topics [Paper] [Slides] at the ParlaCLARIN III 2022 workshop co-located with LREC 2022

📚 01 Datasets

• TRAIN set for supervised topic classification ➡️ Comparative Agendas Project > Interpellations (Germany)
• manual annotated subset TEST of parliamentary debates in data/test_annotation

🧱 02 Segments

• segments extraction tool/graphseg-german (Glavas et al. 2016) [Code] - our adapted German segmentation tool [Code]
• automatic created word-based related SEGMENTS for parliamentary debates with distant supervised topic labels in data/segments.0.1-1

@inproceedings{glavas-etal-2016-unsupervised,
    title = "Unsupervised Text Segmentation Using Semantic Relatedness Graphs",
    author = "Glava{\v{s}}, Goran  and
      Nanni, Federico  and
      Ponzetto, Simone Paolo",
    booktitle = "Proceedings of the Fifth Joint Conference on Lexical and Computational Semantics",
    month = aug,
    year = "2016",
    address = "Berlin, Germany",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/S16-2016",
    doi = "10.18653/v1/S16-2016",
    pages = "125--130",
}

🤖 03 Models

• LM GermanBERT fine-tuned on DeuParl
• task-specific fine-tuning on parlamentary speeches with CAP topic labels from the Comparative Agendas Project on 🤗 HuggingFace parlbert-topics-german

from transformers import pipeline

pipeline_classification_topics = pipeline("text-classification", model="chkla/parlbert-topic-german", tokenizer="bert-base-german-cased", return_all_scores=False)

text = "Sachgebiet Ausschließliche Gesetzgebungskompetenz des Bundes über die Zusammenarbeit des Bundes und der Länder zum Schutze der freiheitlichen demokratischen Grundordnung, des Bestandes und der Sicherheit des Bundes oder eines Landes Wir fragen die Bundesregierung"

pipeline_classification_topics(text) # Government

Cite

@article{klamm-etal-2022-frameast,
  title={FrameASt: A Framework for Second-level Agenda Setting in Parliamentary Debates through the Lense of Comparative Agenda Topics},
  author={Klamm, Christopher and Rehbein, Ines and Ponzetto, Simone},
  journal={ParlaCLARIN III at LREC2022},
  year={2022}
}

Contact: [e-mail]