... it hasbeen the standard practice for many years, we arguethat this paradigm is inadequate both in theory andin practice, for at least the following four reasons. The first reason is that if ... Processing of the AFNLP, pages 522–530, Suntec, Singapore, Au-gust. Association for Computational Linguistics.Canasai Kruengkrai, Kiyotaka Uchimoto, Jun’ichiKazama, Yiou Wang, Kentaro Torisawa, and ... segmenta-tion accuracy based on the flat labels. This accu-racy is in fact the joint accuracy of segmentationand part-of-speech tagging. Most importantly, wecan report our parser’s accuracy in...