Two-Pass Deterministic Dependency Parsing for Long Chinese Sentences

This paper proposes a two-pass parsing approach to improve the performance of deterministic dependency parser for long Chinese sentences. In the first pass, the sentence is divided by every comma, semicolon and colon to be parsed separately, and in the second pass, the built sub-trees are integrated to do a reparsing. The error propagation of the deterministic parser is effectively reduced through the two-pass parsing. The experimental results show that although neither punctuation classifier nor root finder is adopted, the proposed parser still achieves a desirable performance, especially the root accuracy. Keywords--dependency parsing; long Chinese sentences; punctuation-based segmentation; skeleton; reparsing

• Type: Journal
• Year: 2009
• Where: IALP
• Authors: Ping Jian, Chengqing Zong
• Keywords: Deterministic Dependency Parser, IALP 2009, Long Chinese Sentences, Natural Language Processing, Two-pass Parsing