Constructing Multimodal Datasets from Scratch for Rapid Development of a Japanese Visual Language Model

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sasagawa, Keito, Maeda, Koki, Sugiura, Issa, Kurita, Shuhei, Okazaki, Naoaki, Kawahara, Daisuke
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909371864186880
author Sasagawa, Keito
Maeda, Koki
Sugiura, Issa
Kurita, Shuhei
Okazaki, Naoaki
Kawahara, Daisuke
author_facet Sasagawa, Keito
Maeda, Koki
Sugiura, Issa
Kurita, Shuhei
Okazaki, Naoaki
Kawahara, Daisuke
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2410_22736
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle 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
Computation and Language
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.
title Constructing Multimodal Datasets from Scratch for Rapid Development of a Japanese Visual Language Model
topic Computation and Language
url https://arxiv.org/abs/2410.22736