Abstract
We evaluates machine translation models (GPT-3.5-turbo, GPT-4-turbo, Google Translator API, DeepL API, Papago API), focusing on cost, latency, and translation quality, alongside a novel application of LLMs for translation evaluation. Utilizing the dataset of the Korean-English and English-Korean pairs, the study reveals distinct performance attributes: GPT-3.5-turho as the most cost-efficient, Google Translator API as the fastest, and GPT-4-turbo as superior in quality despite higher costs and latency. This approach highlights the critical need for strategic model selection based on specific project requirements and paves the way for future research in LLM-enhanced evaluation.
| Original language | English |
|---|---|
| Title of host publication | ICUFN 2024 - 15th International Conference on Ubiquitous and Future Networks |
| Publisher | IEEE Computer Society |
| Pages | 551-553 |
| Number of pages | 3 |
| ISBN (Electronic) | 9798350385298 |
| DOIs | |
| Publication status | Published - 2024 |
| Event | 15th International Conference on Ubiquitous and Future Networks, ICUFN 2024 - Hybrid, Hungary, Hungary Duration: 2024 Jul 2 → 2024 Jul 5 |
Publication series
| Name | International Conference on Ubiquitous and Future Networks, ICUFN |
|---|---|
| ISSN (Print) | 2165-8528 |
| ISSN (Electronic) | 2165-8536 |
Conference
| Conference | 15th International Conference on Ubiquitous and Future Networks, ICUFN 2024 |
|---|---|
| Country/Territory | Hungary |
| City | Hybrid, Hungary |
| Period | 24/7/2 → 24/7/5 |
Bibliographical note
Publisher Copyright:© 2024 IEEE.
Keywords
- Evaluation
- Large Language Model
- Machine Translation
- Natural Language Processing
ASJC Scopus subject areas
- Computer Networks and Communications
- Computer Science Applications
- Hardware and Architecture