A Semi-universal Pipelined Approach to the CoNLL 2017 UD Shared Task

摘要

This paper presents the TRL team's system submitted for the CoNLL 2017 Shared Task, "Multilingual Parsing from Raw Text to Universal Dependencies." We ran the system for all languages with our own fully pipelined components without relying on either pre-trained baseline or machine learning techniques. We used only the universal part-of-speech tags and distance between words, and applied deterministic rules to assign labels. The delexicalized models are suitable for crosslingual transfer or universal approaches. Experimental results show that our model performed well in some metrics and leads discussion on topics such as contribution of each component and on syntactic similarities among languages.