JAPAGEN
收藏资源简介:
JAPAGEN是一个基于大型语言模型(LLM)生成的日语训练数据集,旨在支持少样本和零样本学习场景下的日语自然语言处理任务。该数据集由横滨国立大学、野村综合研究所和Retrieva公司合作创建,涵盖了文本分类、自然语言推理、语义文本相似性等多个任务。数据集的生成过程利用了LLM的强大生成能力,结合任务特定的知识辅助生成(KADG)方法,以提高生成数据的质量和多样性。JAPAGEN的应用领域主要集中在日语NLP任务中,旨在解决数据稀缺和高成本标注的问题,提升模型在低资源语言环境下的表现。
JAPAGEN is a Japanese training dataset generated using Large Language Models (LLMs), designed to support Japanese natural language processing (NLP) tasks under few-shot and zero-shot learning scenarios. This dataset was collaboratively developed by Yokohama National University, Nomura Research Institute, and Retrieva Inc., covering multiple tasks including text classification, natural language inference (NLI), and semantic text similarity (STS). The dataset generation process leverages the robust generative capabilities of LLMs, combined with the Task-specific Knowledge-augmented Generation (KADG) method, to improve the quality and diversity of the generated data. Primarily applied in Japanese NLP tasks, JAPAGEN aims to address the challenges of data scarcity and high-cost annotation, and boost the performance of models in low-resource language environments.




