Posted - 06/03/2001 : 12:37:05
Department of Chinese, Translation & Linguistics
Institute of Chinese Linguistics
Here we meet and share our research publications.
Here we meet and exchange our research ideas and plans, and cheer each
Here the full papers by the speakers will be distributed.
(I) Why The Blocking Effect?
By Dr. Haihua Pan (潘海華)
[Abstract] This paper discusses the blocking effect observed in
long-distance (LD) bound bare reflexive ziji in Mandarin Chinese. Unlike the symmetrical unlike-person blocking claimed in the literature (Huang & Tang 1991, Xue, Pollard, and Sag 1994), this paper argues that (a) the blocking effect of ziji is not sym-metrical: first and second person pronouns can block third person noun phrases (NP) from long distance binding ziji, though third person NPs do not necessarily block first or second person pronouns from long distance binding ziji; and (b) other grammatical functions filled by first and second person pronouns, not just subjects or NPs contained in the subject, can induce the blocking effect. The paper claims that long distance bound ziji points to the carrier of belief, and reconstructs this notion as self-ascription. The blocking effect is explained by appealing to the fact that only first and second person pronouns are obligatory self-ascribers, and thus can block long distance binding of ziji by third person NPs if they intervene between the potential third person NP and the reflexive ziji, while third person NPs do not necessarily block ziji from being long distance bound by first/second person pronouns. The paper was published in Peter Cole, James C.-T. Huang, and G. Hermon (eds.), Long Distance Reflexives, Syntax and Semantics Vol. 33, Academic Press. New York, pp. 279-316, 2001.
(II) Unsupervised Lexical Learning as Inductinve Inference
By Dr. Chunyu Kit (揭春雨)
[Abstract] To learn a language, the learners must first learn its words.
The difficulty in learning words lies in the unavailability of explicit word boundaries in speech input. The learners have to infer lexical items with some innately endowed learning mechanism(s) for regularity detection. We hypothesize a cognitive mechanism underlying language learning that seeks for the least-effort representation for input data. Accordingly, lexical learning is to infer the minimal-cost representation for the input. The main theme of this thesis is to examine how far this learning mechanism can go in unsupervised lexical learning, entirely resting on statistical induction of structural patterns for the most economic representation for the data. We formulate the description length gain (DLG) measure to evaluate the goodness of a lexical candidate in terms of its compression effect in bits, and an unsupervised lexical learning algorithm as an optimisation process to segment an utterance into word candidates with the greatest sum of DLG. Learning experiments illustrate that our learning approach reaches the state of the art of computational lexical learning.
From Chunyu Kit, Unsupervised Lexical Learning as Inductinve Inference. PhD thesis, Sheffield University, Sept., 2000. (Available online at
Time: 4:30 - 6:00 pm;
Date: Monday, 12 March 2001
Venue: CTL Conference Room B7533, CityU
_____________________ All are welcome! ______________________