Act Annotation and Identification in a Japanese Multi-party Conversation Corpus

Act Annotation and Identification in a Japanese Multi-party Conversation Corpus
复制标题

日语多方对话语料库中的行为标注与识别

DOI:
--
复制
发表时间:
2018
期刊:
--
影响因子:
--
通讯作者:
Kazutaka Shimada
Kazutaka Shimada
中科院分区:
--
文献类型:
--
作者:
Takashi Yamamura;M. Hino;Kazutaka Shimada

文献摘要

参考文献

被引文献

相似文献

话语对话行为的识别在自然语言处理的各种任务中都具有重要的作用,如对话系统和摘要。本文基于现有的日语多方对话语料库标签集,描述了对话行为的标注任务,并评估了标注者之间的一致性分数。然后,我们提出了一种方法,通过使用分类器对每个对话行为标签的输出来识别输入话语的对话行为。我们运用话语特征和视听特征对对话行为进行分类。我们还通过摘要任务评估了对话行为注释的有效性。
Identifying dialogue acts of an utterance has an important role in various tasks of natural language processing, such as dialogue systems and summarization. In this paper, we describe an annotation task of dialogue acts based on an existing tag set for a Japanese multi-party conversation corpus and evaluate agreement scores between annotators. Then, we propose a method that identifies dialogue acts for an input utterance by using the classifier’s outputs for each dialogue act tag. We apply utterance features and audio-visual features for the dialogue act classification. We also evaluate the effectiveness of our dialogue act annotation through a summarization task.
使用 Zig-zag 产品进行图扩展时几乎处处一致
DOI: --
发表时间: 2017
期刊:
影响因子: --
作者:
Hans L. Bodlaender;Hirotaka Ono;Yota Otachi;切上太希,小柴健史
通讯作者: 切上太希,小柴健史