Abstract
Unlike written English having spaces between words, the current Chinese orthography uses no spaces. This poses certain problems in automatic text processing, such as word segmentation and syntactic parsing. Chiang et. al. (2009) proposed Spaced (讓格) orthography that uses simple short phrases as basic writing units. With simple short phrases, the problem of word segmentation becomes much easier. Texts in Spaced orthography are also easier to read. For Taiwanese, Spaced orthography seems closely related to Taiwanese tone sandhi patterns. Three tasks are studied in this paper. The first is the mining of new words, automatic extractions of new words from newspaper texts. The second and the third are the automatic segmentations of traditional Chinese sentences into so-called generalized words (寬詞) and into simple short phrases. Using the number of characters for easy operability, we propose a set of 1234 rules for the specification of generalized words.