obolo-ai-translation

Obolo AI Translation System

About This Project

This is an AI-powered translation tool for English to Obolo, built to support the important work of Obolo language preservation and translation.

Important Note: This is a technology demonstration and AI tool. The authoritative Obolo translations come from OLBTO (Obolo Language and Bible Translation Organization), the recognized experts in Obolo language and Bible translation.

My Role: I am an AI technology enthusiast bringing machine learning capabilities to support Obolo language visibility in the AI age.

Author

Okejonah1 (Oke Nteigbanam J.)

Acknowledgments

OLBTO (Obolo Language and Bible Translation Organization)

The recognized authority on Obolo Bible translation and language preservation.

Obolo Linguists and Researchers

The dedicated linguists who documented Obolo grammar rules, vocabulary, and language structures.

The Obolo Community

The speakers, elders, and community members who preserve the Obolo language.

Meta (M2M-100)

For providing the multilingual translation model.

How Their Work Was Used

Dictionary Examples (English to Obolo)

Bible Translation Examples (English to Obolo)

What This Tool Does

• AI Translation Model: Fine-tuned M2M-100 with 418M parameters trained on 40k pairs
• Dictionary Lookup: 6,383+ Obolo words with variants
• Hybrid System: Dictionary first, AI fallback for unknowns
• Meaning-Based: Cultural adaptations and natural Obolo phrasing

How to Use

from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer
import json

model = M2M100ForConditionalGeneration.from_pretrained("Okejonah1/obolo-ai-translation")
tokenizer = M2M100Tokenizer.from_pretrained("Okejonah1/obolo-ai-translation")

# Load dictionary
with open("obolo_dictionary_master.json", "r") as f:
    dictionary = json.load(f)

def translate(text):
    # 1. Check dictionary first
    text_lower = text.lower().strip()
    if text_lower in dictionary:
        return dictionary[text_lower]
    # 2. Use AI model
    tokenizer.src_lang = "en"
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)
    outputs = model.generate(**inputs, max_length=128, num_beams=5)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

print(translate("God"))        # Awaji
print(translate("thank you"))   # oniro
print(translate("please"))      # soso

Model Details

• Base Model: facebook/m2m100_418M
• Training Data: 40,000+ English-Obolo pairs
• Training Steps: 14,250
• Final Validation Loss: 0.82
• Format: safetensors

Features

• AI-assisted translation
• Dictionary lookup for known words
• Multiple variants support
• Cultural adaptation awareness

Limitations

• This is an AI tool, not an expert translator
• Always verify translations with native speakers
• OLBTO and expert translations remain the authoritative source

License

MIT License

Author

Okejonah1 (Oke Nteigbanam J.)

Contact

• Hugging Face: Okejonah1
• GitHub: Okejonah1

Thank You

To all who have worked to preserve and promote the Obolo language: Your work is the foundation upon which this AI tool is built. This project exists because of your dedication to the Obolo language.