Kobart Jeju Translation

🚀 🌊 Jeju-Standard Bidirectional Translation Model

This is a bidirectional translation model between Jeju dialect and Standard Korean, which can effectively perform translation tasks between the two languages.

🚀 Quick Start

You can use this model with transformers to perform inference. Below is an example of how to load the model and generate translations:

Basic Usage

import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

## Set up the device (GPU or CPU)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

## Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("Junhoee/Kobart-Jeju-translation")
model = AutoModelForSeq2SeqLM.from_pretrained("Junhoee/Kobart-Jeju-translation").to(device)

## Set up the input text
## Before entering a sentence, input the [Jeju] or [Standard] token according to the direction and then enter the sentence
input_text = "[표준] 안녕하세요"

## Tokenize the input text
input_ids = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True).input_ids.to(device)

## Generate the translation
outputs = model.generate(input_ids, max_length=64)

## Decode and print the output
decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Model Output:", decoded_output)

Model Output: 안녕하수꽈

Advanced Usage

## Set up the input text
## Before entering a sentence, input the [Jeju] or [Standard] token according to the direction and then enter the sentence
input_text = "[제주] 안녕하수꽈"

## Tokenize the input text
input_ids = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True).input_ids.to(device)

## Generate the translation
outputs = model.generate(input_ids, max_length=64)

## Decode and print the output
decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Model Output:", decoded_output)

Model Output: 안녕하세요

✨ Features

• This is a bidirectional translation model for Jeju dialect and Standard Korean.
• It can perform translation tasks between Jeju dialect and Standard Korean.

📦 Installation

No specific installation steps are provided in the original document.

📚 Documentation

Member

• Bitamin 12th cohort: Koo Junhoe, Lee Seohyun, Lee Yerin
• Bitamin 13th cohort: Kim Yoonyoung, Kim Jaegeom, Lee Hyeongseok

Github Link

• https://github.com/junhoeKu/Jeju_Translation.github.io

Parent Model

• gogamza/kobart-base-v2
• https://huggingface.co/gogamza/kobart-base-v2

Dataset - Approximately 930,000 rows

• AI-Hub (Jeju dialect utterance data + Middle-aged dialect utterance data)
• Github (KakaoBrain JIT data)
• Others
  • Jeju dialect dictionary data (Crawled from the Jeju Provincial Government homepage)
  • Lyrics translation data (Collected one by one from the "Mworanghaman" YouTube channel)
  • Book data (Collected one by one from books such as "The Taste and Charm of Jeju Dialect" and "Even if You're in Buenos Aires")
  • 2018 Jeju dialect oral data collection (Collected one by one - Used as evaluation data)

Hyper Parameters

• Epoch: 3 epochs
• Learning Rate: 2e-5
• Weight Decay = 0.01
• Batch Size: 32

Bleu Score

• Based on the 2018 Jeju dialect oral data collection:
  • Jeju dialect -> Standard Korean: 0.76
  • Standard Korean -> Jeju dialect: 0.5
• Based on the validation data of the AI-Hub Jeju dialect utterance data:
  • Jeju dialect -> Standard Korean: 0.89
  • Standard Korean -> Jeju dialect: 0.77

CREDIT

• Koo Junhoe: kujoon13413@gmail.com
• Kim Yoonyoung: 202000872@hufs.ac.kr
• Kim Jaegeom: worua5667@inha.edu
• Lee Seohyun: rlaorrn0123@sookmyung.ac.kr
• Lee Yerin: i75631928@gmail.com
• Lee Hyeongseok: gudtjr3638@gmail.com