Development of spoken language processing techniques for real-time captioning on the basis of readability

在这项研究中，我开发了以下三种口语处理技术作为生成可读字幕的基本技术，以支持对演讲和评论等口语独白的实时理解：(1)基于从句开头识别的口语增量依存句法分析技术(2)基于依存结构的口语文本格式化技术(主要是逗号插入技术)，将口语转换为可读文本(3)基于依存结构的换行插入技术，以便于阅读的方式显示字幕文本。

In this research, I developed the following three spoken language processing techniques as elemental techniques for generating readable captions to support the real-time understanding of spoken monologues such as lectures and commentaries : (1) Incremental dependency parsing technique of spoken language based on clause-start identification (2) Text formatting technique (mainly, comma insertion technique) of spoken language based on dependency structure to translate spoken language to readable texts (3) Linefeed insertion technique based on dependency structure to display caption texts in an easy-to-read manner.