How donisumito/acehx works
AcehX is a research resource for Acehnese NLP: it provides an unlabeled corpus, labeled sentiment data, and a tokenizer vocabulary to support MLM pretraining and sentiment fine-tuning. The documented workflow extends a tokenizer, pretrains AcehXBERT, then fine-tunes it for sentiment classification. Training notebooks are present, but their implementation was not available for inspection; links below reflect README-documented stages only.
Main entry points: run_AcehXBERT_sentiment.ipynb Sentiment Fine-tuning · run_AcehXBERT_mlm.ipynb MLM Training · vocab.txt Tokenizer Vocabulary · AcehX_Corpus.txt AcehX Corpus
Components 9
Language Data
- AcehX Corpus
datasets/MLM/AcehX_Corpus.txt - Sentiment Splits
datasets/Sentiment
Tokenization
- Tokenizer Vocabulary
tokenizer_AcehX_Corpus/vocab.txt - Tokenizer Config
tokenizer_AcehX_Corpus
Model Training
- MLM Training
scripts/run_AcehXBERT_mlm.ipynb - Sentiment Fine-tuning
scripts/run_AcehXBERT_sentiment.ipynb - AcehXBERT MLM
- AcehXBERT Sentiment
Other components
- NLP Researcher
Connections 10
- NLP Researcher → AcehX Corpus: uses
- NLP Researcher → Sentiment Splits: uses
- Tokenizer Vocabulary → Tokenizer Config: configures
- MLM Training → AcehX Corpus: trains on
- MLM Training → Tokenizer Vocabulary: uses
- MLM Training → AcehXBERT MLM: pretrains
- Sentiment Fine-tuning → Sentiment Splits: fine-tunes on
- Sentiment Fine-tuning → Tokenizer Vocabulary: uses
- Sentiment Fine-tuning → AcehXBERT MLM: fine-tunes
- Sentiment Fine-tuning → AcehXBERT Sentiment: produces