vrclc
/

transliteration

@@ -44,134 +44,100 @@ Sequence to Sequence Model for Treansliterationg Romanised Malayalam (Manglish)
 - **Repository:** https://github.com/VRCLC-DUK/ml-en-transliteration
 - **Paper:** https://arxiv.org/abs/2412.09957
 - **Demo:** https://huggingface.co/spaces/vrclc/en-ml-transliteration
 ## How to Get Started with the Model
-The model needs to have an externally defined tokenizers for source and target languages.
-[More Information Needed]
-## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]

 - **Repository:** https://github.com/VRCLC-DUK/ml-en-transliteration
 - **Paper:** https://arxiv.org/abs/2412.09957
 - **Demo:** https://huggingface.co/spaces/vrclc/en-ml-transliteration
+- Developed as a shared task submission to [INDONLP Workshop](https://indonlp-workshop.github.io/IndoNLP-Workshop/) at [COLING 2025](https://coling2025.org//), Abu Dhabi.
 ## How to Get Started with the Model
+The model needs to have an user defined tokenizers for source and target scripts. The model is trained on words. If your use case involves transliterating full sentences, split the sentences into words before passing to the model.
+### Load Dependencies
+```
+import keras
+import huggingface_hub
+import tensorflow as tf
+import numpy as np
+from tensorflow.keras.preprocessing.text import Tokenizer
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+from huggingface_hub import from_pretrained_keras
+import re
+```
+## Load Model
+```
+model = from_pretrained_keras("vrclc/transliteration")
+```
+### Define Tokens:
+```
+source_tokens = list('abcdefghijklmnopqrstuvwxyz ')
+source_tokenizer = Tokenizer(char_level=True, filters='')
+source_tokenizer.fit_on_texts(source_tokens)
+target_tokens = [
+    # Independent vowels
+    'അ', 'ആ', 'ഇ', 'ഈ', 'ഉ', 'ഊ', 'ഋ', 'ൠ', 'ഌ', 'ൡ', 'എ', 'ഏ', 'ഐ', 'ഒ', 'ഓ', 'ഔ',
+    # Consonants
+    'ക', 'ഖ', 'ഗ', 'ഘ', 'ങ', 'ച', 'ഛ', 'ജ', 'ഝ', 'ഞ',
+    'ട', 'ഠ', 'ഡ', 'ഢ', 'ണ', 'ത', 'ഥ', 'ദ', 'ധ', 'ന',
+    'പ', 'ഫ', 'ബ', 'ഭ', 'മ', 'യ', 'ര', 'ല', 'വ', 'ശ',
+    'ഷ', 'സ', 'ഹ', 'ള', 'ഴ', 'റ',
+    # Chillu letters
+    'ൺ', 'ൻ', 'ർ', 'ൽ', 'ൾ',
+    # Additional characters
+    'ം', 'ഃ', '്',
+    # Vowel modifiers / Signs
+    'ാ', 'ി', 'ീ', 'ു', 'ൂ', 'ൃ', 'ൄ', 'െ', 'േ', 'ൈ', 'ൊ', 'ോ', 'ൌ', 'ൗ', ' '
+]
+target_tokenizer = Tokenizer(char_level=True, filters='')
+target_tokenizer.fit_on_texts(target_tokens)
+```
+### Wrapper script to split input sentences to words before passing to the model
+```
+def transliterate_with_split_tokens(input_text, model, source_tokenizer, target_tokenizer, max_seq_length):
+    """
+    Transliterates input text in roman script, retains all other characters (including punctuation, spaces, etc.)
+    """
+    # Regular expression to split the text into tokens and non-tokens
+    tokens_and_non_tokens = re.findall(r"([a-zA-Z]+)|([^a-zA-Z]+)", input_text)
+    transliterated_text = ""
+    for token_or_non_token in tokens_and_non_tokens:
+        token = token_or_non_token[0]
+        non_token = token_or_non_token[1]
+        if token:
+            input_sequence = source_tokenizer.texts_to_sequences([token])[0]
+            input_sequence_padded = pad_sequences([input_sequence], maxlen=max_seq_length, padding='post')
+            predicted_sequence = model.predict(input_sequence_padded)
+            predicted_indices = np.argmax(predicted_sequence, axis=-1)[0]
+            transliterated_word = ''.join([target_tokenizer.index_word[idx] for idx in predicted_indices if idx != 0])
+            transliterated_text += transliterated_word
+        elif non_token:
+            transliterated_text += non_token
+    return transliterated_text
+```
+### Usage
+```
+input text = "ente veedu"
+transliterated_text = transliterate_with_split_tokens(input_text, model, source_tokenizer, target_tokenizer, max_seq_length)
+print(transliterated_text)
+```
+[More Information Needed]
+## Citation
+```
+@article{baiju2024romanized,
+  title={Romanized to Native Malayalam Script Transliteration Using an Encoder-Decoder Framework},
+  author={Baiju, Bajiyo and Pillai, Leena G and Manohar, Kavya and Sherly, Elizabeth},
+  journal={arXiv preprint  arXiv:2412.09957},
+  year={2024}
+}
+```