Skip to content

Latest commit

 

History

History
38 lines (27 loc) · 2.17 KB

README.md

File metadata and controls

38 lines (27 loc) · 2.17 KB

Multilingual Metaphor Detection

This notebook shows how to fine-tune the multilingual language model XLM-RoBERTa for metaphor detection on a token-level using Huggingface.

Reference

We describe model and training details in our publication:

Wachowiak, L., Gromann, D. & Xu, C. (2022) Drum Up SUPPORT: Systematic Analysis of Image-Schematic Conceptual Metaphors. In EMNLP FigLang Workshop.

@inproceedings{wachowiak-etal-2022-drum,
    title = "Drum Up {SUPPORT}: Systematic Analysis of Image-Schematic Conceptual Metaphors",
    author = "Wachowiak, Lennart  and
      Gromann, Dagmar  and
      Xu, Chao",
    booktitle = "Proceedings of the 3rd Workshop on Figurative Language Processing (FLP)",
    month = dec,
    year = "2022",
    publisher = "Association for Computational Linguistics",
    doi = "10.18653/v1/2022.flp-1.7",
    pages = "44--53",
}

PDF, with further experiment code here

Usage

You can either train your own model simply running the provided notebook or try out the already trained model here

Dataset

The dataset the model is trained on is the VU Amsterdam Metaphor Corpus that was annotated on a word-level following the metaphor identification protocol. The training corpus is restricted to English, however, XLM-R shows decent zero-shot performances when tested on other languages.

Results

Following the evaluation criteria from the 2020 Second Shared Task on Metaphor detection our model achieves a F1-Score of 0.76 for the metaphor-class when training XLM-RBase and 0.77 when training XLM-RLarge..

We train for 8 epochs loading the model with the best evaluation performance at the end and using a learning rate of 2e-5. From the allocated training data 10% are utilized for validation while the final test set is being kept seperate and only used for the final evaluation.