揭春雨_ChineseCharacters.ppt

汉字作为词汇语义基元: 探略Chinese Characters as Primitives in Lexical Semantics 揭春雨 Chunyu KIT Dept. of Chinese, Translation Lingusitics City University of Hong Kong ctckit@cityu.edu.hk Outline Chinese character vs. letter Characters in Chinese morphology Morpheme：bound vs. free Compounding vs. Derivation：Stem, prifix suffix Special case: 们 Character type in morphology: a, n, v, adj, adv, … Exceptions: adv + n, licensed by meaning Negation in Chinese: morphology syntax Text meaning in characters What did Li Bai and Du Fu say in their poetry? Self introduction: our research group CWS: Supervised unsupervised word identification Achievements in recent international evaluations: Zhao Hai: No. 1 in all closed tests oif CWS in Bakeoff-4 Hai Zhao: No. 2 1 in CoNLL dependency parsing shared tasks Billy Wong: No. 7-8, in NIST MT Evel. Metrics: ATEC Xiao Chen: No. 1 in last Chinese parsing evaluation Xu Ruifeng: No. 1 2 in 2nd COAT last NTCIR (Chinese tracks) Song Yan: No. 1 in all Chinese related transliteration shared tasks New direction: Computational poetics/poetry Character vs. letter Which sequence of char/letter is meaningful? a b c d … ab norm al mean, mean ing, mean ing ful, un mean ing, un mean ing ful  一 二 三 四 五 … 正 工 木 火 水 土 金 … 水土 木工 四正 五金 五土 五一 五四 火水 … Observations Meanings are assigned to individual symbol sequences Then, this meaning-symbol association ? stable Fixed set of symbols ? unlimitedly many meanings  Meaning combination through symbol combination Mostly, tree, in almost all morpho-syntactical levels Concatenation = small / flat tree More and more symbol combinations are created for new meanings. What license a combination? Syntax! E.g., Adj + N Adv + Adj Adv + V … But, exceptions: Adv + N (in Chinese) Syntax Morphology Characters in Chinese morphology Almost every 字 can be a word (or morpheme) Exceptions: 连绵词(binding word)：蚯蚓，蜉蝣，枇杷，琵琶，… 半连绵：蝴蝶，蟾蜍，… Words = free morphemes Bound morphemes A small close set in Chinese: 们，儿，… Chin