Interdisziplinäres Zentrum für Kognitive Sprachforschung

Links und Funktionen



Abstract Wiechmann

Initial parsing decisions and lexical bias

Recent research in sentence comprehension suggests that lexically specific information plays a key role in online syntactic ambiguity resolution (Kawamoto 1993, MacDonald et al. 1994). On the basis of an analysis of the local NP/S-ambiguity, the present study offers a corpus-based approach to sentence processing that in principle supports this view, but suggests that the relevant information used to recover the syntactic structure of an incoming string of words is better conceived of as residing at the level of the verb sense instantiated.

The analysis proceeds in two steps: First, I determine both the form-based and sense-contingent preferences of 20 verbs on the basis of a 17 million words sample of the BNC. For the sense-contingent preferences, the semantic distinctions assumed in WordNet were used (Fellbaum 1998) and all corpus data were coded manually (Ntoken ~ 5,000). To express the individual preferences, I applied a corpus-based method termed ‘distinctive collexeme analysis’ (Gries and Stefanowitsch 2004), which is a sub-type of a family of techniques geared to measure the interaction between the interaction of two constructions of different levels of abstraction. The method usually employs association measures from statistical hypothesis testing to estimate the strength of attraction between the two (linguistic) units under investigation. The results clearly indicate that the preferences associated with the individual verb sense strongly differ form the overall preferences of a given verb, hence predicting different structural expectations contingent on the verb sense instantiated.

Second, the results were compared with the data of a self-paced moving window experiment reported in Hare et al. (2003). Specifically, a correlational analysis was performed between the the computed associations scores, and the observed ambiguity effects from Hare et al’s study. The results supports the experimental findings and thus provides additional evidence for the hypothesis that verb senses, rather than verb forms, play an important role in syntactic ambiguity resolution [tau = 0.0308; z = 0.2807; p = 0.048].

The findings can be straightforwardly described using an Embodied Construction Grammar framework (Bergen and Chang 2003). Parsing in this account (roughly) is an analysis process, which takes an input utterance in context and determines the set of constructions most likely to be responsible for it (Bryant 2003). A construction-based analyzer can incorporate semantic and syntactic constraints in parallel instead of requiring syntactic parsing to precede semantic interpretation. It can thus model sense-contingent structural expectations very naturally by assigning different biases, i.e. different association scores, to polysemous verbal constructions towards possible argument structures constructions they can unify with.



Bergen, B.K. and Chang, N.C. (2003). Embodied Construction Grammar in Simulation-Based Language Understanding. In: Ostman, J.O. and Fried, M. (eds): Construction Grammar(s): Cognitive and Cross-Language Dimensions. John Benjamin, Amsterdam.

Bryant, J. (2003). Constructional analysis. Master's thesis, UC Berkeley.

Fellbaum, C. (1998). WordNet: An electronic lexical database. Cambridge: MIT Press.

Gries, St.Th., and Stefanowitsch. A. (2004). Extending collostructional analysis: A corpus-based perspectives on 'alternations'. International Journal of Corpus Linguistics 9: 97-129.

Hare, M. L., McRae, K., and Elman, J.L. (2003). Sense and structure: Meaning as a determinant of verb subcategorization preferences. Journal of Memory and Language, 48(2): 281-303.

Kawamoto, A.H. (1993). Nonlinear dynamics in the resolution of lexical ambiguity: A parallel distributed processing account. Journal of Memory and language 32: 474-516.

MacDonald, M.C., Pearlmutter, N.J., and Seidenberg, M.S. (1994). Lexical nature of syntactic ambiguity resolution. Psychological Review 101: 676-703.