Syntactic annotation under dependency scheme on Chinese spontaneous speech

Syntactic or semantic annotation is an indispensable work for understanding the intention of the spoken discourses or dialogues. As we know that dependency relation annotation is a kind of syntactic or low shallow semantic annotation, however the present annotation schemes are almost all for text rather than spoken discourses or dialogues. By analysis the online spontaneous chatting data, this paper tries to propose a dependency scheme for spoken discourses. Based on the HIT scheme(Harbin Institute of Technology), we finally proposed 26 kinds of dependencies, where four dependencies are added as “Translocation”, “Repetition”, “Duplication” and “Omission”, and three are modified as “Independent Structure”, “Independent Clause” and “Dependent Clause”. The refined dependency scheme enriches the annotation to Chinese spontaneous speech, which would benefit for speech recognition, semantic comprehension and machine translation.