Constructing Multimodal Datasets from Scratch for Rapid Development of a Japanese Visual Language Model
Sasagawa, Keito, Maeda, Koki, Sugiura, Issa, Kurita, Shuhei, Okazaki, Naoaki, Kawahara, Daisuke
–arXiv.org Artificial Intelligence
To develop high-performing Visual Language Models (VLMs), it is essential to prepare multimodal resources, such as image-text pairs, interleaved data, and instruction data. While multimodal resources for English are abundant, there is a significant lack of corresponding resources for non-English languages, such as Japanese. To address this problem, we take Japanese as a non-English language and propose a method for rapidly creating Japanese multimodal datasets from scratch. We collect Japanese image-text pairs and interleaved data from web archives and generate Japanese instruction data directly from images using an existing VLM. Our experimental results show that a VLM trained on these native datasets outperforms those relying on machine-translated content.
arXiv.org Artificial Intelligence
Oct-30-2024
- Country:
- North America > United States
- Texas > Travis County > Austin (0.04)
- Europe
- Spain > Catalonia
- Barcelona Province > Barcelona (0.04)
- Italy > Piedmont
- Turin Province > Turin (0.04)
- Spain > Catalonia
- Asia > Japan
- Honshū
- Kantō > Tokyo Metropolis Prefecture
- Tokyo (0.04)
- Kansai > Kyoto Prefecture
- Kyoto (0.04)
- Kantō > Tokyo Metropolis Prefecture
- Honshū
- North America > United States
- Genre:
- Research Report (0.70)
- Industry:
- Education (0.68)
- Technology: