Logo Lanfrica

annettehjc/ComMorph_LRD

Domaine:

natural language processing

Type de record:

software
Créateur:
ann
Hôte:
Computational Morphology project: Low Resource Dialect with FSMorphological Analyzer # ComMorph_LRD Paper: Morphological Analysis on Jeju and Korean Using Transformer (pdf) Author: Hyunjoo Cho (email) ## Info This repository contains the code framework used in the study to measure similarity between Korean and Jeju using encoder-decoder transformer models. 1. **extract_data_from_original_data** contains files to process different data format to extract needed transcription sentences of Jeju and Korean. 2. **create_organise_data** contains files to process data file for cleaning, and splitting them into training, validation, and test sets. 3. **train_model_evaluate** contains files to create vocabulary files, to train models with arguments, and to evaluate on them to see the performance level. The experiment is based on data from jeju-dialect-to-standard, jeju_potato_dataset, and jejueo_interview_transcripts.