... tagging (Collins, 2002), Chinese word segmentation (Ng and Low, 2004; Zhang and Clark, 2007) and so on We trained a character-based perceptron for Chinese Joint S&T, and found that the perceptron ... the POS information and reported the F-measure on segmentation only, while the second performed Joint S&T using POS information and reported the F-measure both on segmentation and on Joint S&T ... higher-order word LM on a larger scale corpus Finally, the word count penalty gives improvement to the cascaded model, 0.13 points on segmentation and 0.16 points on Joint S&T In summary, the cascaded model...