arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1803.09189cs.CLcs.CV
收藏

作为依存句法分析的场景图分析

Scene Graph Parsing as Dependency Parsing

National Taiwan University(台湾大学);Johns Hopkins University(约翰斯·霍普金斯大学);Hong Kong University of Science and Technology(香港科技大学)

Yu-Siang Wang, Chenxi Liu, Xiaohui Zeng, Alan Yuille

摘要:

本文研究了从文本描述中解析结构化知识图谱的问题。特别是,我们考虑场景图表示,认为对象与它们的属性和关系:这种表示已被证明是有用的各种视觉和语言应用。我们首先介绍一个替代的,但等效的边缘为中心的视图场景图连接到依赖关系解析。再加上标签和动作空间的仔细重新设计,我们将先前工作中使用的两阶段管道(通用依存句法分析,然后是简单的后处理)合并为一个,从而实现端到端的训练。我们学习的神经依赖解析器生成的场景图与我们评估集上的地面真值图的F分数相似度为49.67%,超过了之前最好的方法5%。我们进一步证明了我们学习的解析器在图像检索应用中的有效性。

英文摘要:

In this paper, we study the problem of parsing structured knowledge graphs from textual descriptions. In particular, we consider the scene graph representation that considers objects together with their attributes and relations: this representation has been proved useful across a variety of vision and language applications. We begin by introducing an alternative but equivalent edge-centric view of scene graphs that connect to dependency parses. Together with a careful redesign of label and action space, we combine the two-stage pipeline used in prior work (generic dependency parsing followed by simple post-processing) into one, enabling end-to-end training. The scene graphs generated by our learned neural dependency parser achieve an F-score similarity of 49.67% to ground truth graphs on our evaluation set, surpassing best previous approaches by 5%. We further demonstrate the effectiveness of our learned parser on image retrieval applications.

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。

↑