Paper
Title: Decoding Sign Languages: The SL-FE Framework for Phonological Analysis and Automated Annotation
Authors: Karahan Şahin, Kadir Gökgöz
Venue: SignLang 2024 (LREC-COLING 2024 workshop)
ACL Anthology: https://aclanthology.org/2024.signlang-1.37/
Summary
Introduces SL-FE, a framework that derives continuous phonological feature signals (finger selection, orientation, location, movement) from MediaPipe pose, enabling automated ELAN annotation and quantitative tests of theoretical phonological claims. Uses hand-crafted continuous scores (normalized angles, axis-aligned distances, softmax over normalized differences, finite-difference complexity) plus threshold + argmax/argmin ranking. Validated on Turkish Sign Language (TID) lexical retrieval (TID Sözlük dictionary, Top-100).
Results / Conclusions
- Dominance condition: accuracy 0.90, F1 0.95.
- Symmetry condition: accuracy 0.84, F1 0.77.
- Conclusion: rule-based pose-derived phonological features can validate theoretical constraints empirically.
Code & Data
Suggested Reproduction
- Clone the repo and install dependencies.
- Run the SL-FE pipeline on TID Sözlük data.
- Reproduce Dominance and Symmetry validation accuracy / F1.
Bibtex
@inproceedings{sahin-gokgoz-2024-decoding,
title = "Decoding Sign Languages: The {SL}-{FE} Framework for Phonological Analysis and Automated Annotation",
author = {{\c{S}}ahin, Karahan and
G{\"o}kg{\"o}z, Kadir},
editor = "Efthimiou, Eleni and
Fotinea, Stavroula-Evita and
Hanke, Thomas and
Hochgesang, Julie A. and
Mesch, Johanna and
Schulder, Marc",
booktitle = "Proceedings of the LREC-COLING 2024 11th Workshop on the Representation and Processing of Sign Languages: Evaluation of Sign Language Resources",
month = may,
year = "2024",
address = "Torino, Italia",
publisher = "ELRA and ICCL",
url = "https://aclanthology.org/2024.signlang-1.37/",
pages = "335--342"
}
Paper
Title: Decoding Sign Languages: The SL-FE Framework for Phonological Analysis and Automated Annotation
Authors: Karahan Şahin, Kadir Gökgöz
Venue: SignLang 2024 (LREC-COLING 2024 workshop)
ACL Anthology: https://aclanthology.org/2024.signlang-1.37/
Summary
Introduces SL-FE, a framework that derives continuous phonological feature signals (finger selection, orientation, location, movement) from MediaPipe pose, enabling automated ELAN annotation and quantitative tests of theoretical phonological claims. Uses hand-crafted continuous scores (normalized angles, axis-aligned distances, softmax over normalized differences, finite-difference complexity) plus threshold + argmax/argmin ranking. Validated on Turkish Sign Language (TID) lexical retrieval (TID Sözlük dictionary, Top-100).
Results / Conclusions
Code & Data
Suggested Reproduction
Bibtex