arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2409.07088cs.CLcs.AI
收藏

基于大语言模型的无本体通用领域知识图谱到文本生成数据集合成

Ontology-Free General-Domain Knowledge Graph-to-Text Generation Dataset Synthesis using Large Language Model

Graduate School of Artificial Intelligence, POSTECH(浦项科技大学人工智能研究生院);Department of Computer Science and Engineering, POSTECH(浦项科技大学计算机科学与工程系)

Daehee Kim, Deokhyung Kang, Sangwon Ryu, Gary Geunbae Lee

AI总结:

针对通用领域高质量G2T数据集稀缺的问题,研究提出利用LLM和Data-QuestEval生成含585万对数据的无本体WikiOFGraph数据集,实验证明在其上微调的PLM性能更优,为G2T数据生成提供了可扩展的有效方案。

摘要:

知识图谱到文本(G2T)生成指将结构化知识图谱转化为自然语言文本的过程。预训练语言模型(PLMs)的最新进展提升了G2T的性能,但它们的效果依赖于具备精确图谱-文本对齐的数据集。然而,高质量通用领域G2T生成数据集的稀缺性限制了通用领域G2T生成研究的进展。为解决这一问题,我们提出了Wikipedia Ontology-Free Graph-text数据集(WikiOFGraph,维基百科无本体图谱-文本数据集),这是一个利用大语言模型(LLM)和Data-QuestEval的新方法生成的大规模G2T数据集。该数据集包含585万条通用领域图谱-文本对,无需依赖外部本体即可实现高图谱-文本一致性。实验结果表明,在WikiOFGraph上微调的PLM在各类评估指标上均优于在其他数据集上训练的模型。我们的方法被证明是生成高质量G2T数据的可扩展且有效的解决方案,显著推动了G2T生成领域的发展。

英文摘要:

Knowledge Graph-to-Text (G2T) generation involves verbalizing structured knowledge graphs into natural language text. Recent advancements in Pretrained Language Models (PLMs) have improved G2T performance, but their effectiveness depends on datasets with precise graph-text alignment. However, the scarcity of high-quality, general-domain G2T generation datasets restricts progress in the general-domain G2T generation research. To address this issue, we introduce Wikipedia Ontology-Free Graph-text dataset (WikiOFGraph), a new large-scale G2T dataset generated using a novel method that leverages Large Language Model (LLM) and Data-QuestEval. Our new dataset, which contains 5.85M general-domain graph-text pairs, offers high graph-text consistency without relying on external ontologies. Experimental results demonstrate that PLM fine-tuned on WikiOFGraph outperforms those trained on other datasets across various evaluation metrics. Our method proves to be a scalable and effective solution for generating high-quality G2T data, significantly advancing the field of G2T generation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑