Apppublic

leangsuor/Khmer_Latin-to-Khmer_script_Transliteration

sourceHugging Facemitupdated 2mo agoView on Hugging Face

0likes

App README

Khmer Latin → Khmer Script Transliteration

Compare three transliteration models that convert romanized Khmer (Latin script) into Khmer script (ខ្មែរ).

Models

Model	Architecture	Notes
GRU	Seq2Seq GRU encoder-decoder	`gru_v3_bs64_ed128_ld512`
LSTM	Seq2Seq LSTM encoder-decoder	`lstm_v3_bs32_ed64_ld512`
Fuzzy N-gram	Fuzzy matching + greedy backoff	Rule-augmented lookup

Security Note

hf_models/fuzzy_ngram_greedy_backoff/checkpoints/model.pkl is flagged by HuggingFace's scanner because it is a pickle file. This file is a serialized KhmerFuzzyContextualPredictor object trained solely on Khmer romanization data — it contains no executable payloads. You can inspect the training code in models/fuzzy_and_ngram_greedy_backoff.py.

Usage

Type any romanized Khmer word or phrase and click Transliterate:

—soksabay → សុខសប្បាយ
—krong phnom penh → ក្រុងភ្នំពេញ
—bong srey mean snaeh → បងស្រីមានស្នេហ៍