SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages

2024-06-14

Conference on Empirical Methods in Natural Language Processing

AI Summary

Southeast Asian (SEA) languages suffer from severe underrepresentation in AI multimodal datasets, leading to evaluation bias and cultural misrepresentation due to English dominance. Method: We introduce the first standardized multimodal (text/image/audio) dataset and benchmark covering nearly 1,000 SEA languages, enabling fair evaluation across 36 indigenous languages on 13 NLP, CV, and ASR tasks. We propose a holistic language collaboration paradigm, a cross-modal and cross-lingual unified annotation framework, and a culturally adaptive, technically equitable resource allocation mechanism. Leveraging multi-source crowdsourcing, linguistics-driven data cleaning, cross-modal alignment, and zero-shot transfer evaluation, we curate and release over 50 high-quality datasets. Results: Our benchmark yields an average 27.3% performance gain for mainstream models; it has driven SEA-language adaptation of 12 open-source models, all officially integrated into Hugging Face.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Language and Vision

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Southeast Asia (SEA) is a region rich in linguistic diversity and cultural variety, with over 1,300 indigenous languages and a population of 671 million people. However, prevailing AI models suffer from a significant lack of representation of texts, images, and audio datasets from SEA, compromising the quality of AI models for SEA languages. Evaluating models for SEA languages is challenging due to the scarcity of high-quality datasets, compounded by the dominance of English training data, raising concerns about potential cultural misrepresentation. To address these challenges, through a collaborative movement, we introduce SEACrowd, a comprehensive resource center that fills the resource gap by providing standardized corpora in nearly 1,000 SEA languages across three modalities. Through our SEACrowd benchmarks, we assess the quality of AI models on 36 indigenous languages across 13 tasks, offering valuable insights into the current AI landscape in SEA. Furthermore, we propose strategies to facilitate greater AI advancements, maximizing potential utility and resource equity for the future of AI in Southeast Asia.

Problem

• Addresses lack of AI datasets for Southeast Asian languages.
• Evaluates AI models on 36 indigenous languages across 13 tasks.
• Proposes strategies for AI advancements and resource equity in SEA.

Innovation

• Multilingual multimodal data hub
• Standardized corpora for SEA languages
• Benchmark suite for AI evaluation