Computational Linguistics, University of the Saarland As the importance of performance issues in parsing has been increasing
during the last few years, cognitive models of sentence processing
are becoming more than just an important source of inspiration
for computational linguistics. The course addresses current topics
in human sentence processing; the phenomena to be accounted for
include ambiguity resolution, center-embedding and reanalysis.
At the level of modelling, issues such as parsing strategies,
heuristics, incrementality, underspecification, description-based
parsing, determinism, parallelism, resource limitations, adaptiveness,
etc. will be adressed. In particular, the following topics will be addressed: Sessions 1,2: Classification schemes for human parsing models Serialism, parallelism, determinsism, underspecification, minimal
commitment parsing; architectural vs. transparency-based approaches;
computational resources and adaptiveness; lexicalisation, head-driven
parsing, incrementality; modularity, interaction, integration;
these sessions will give an overview of an almost complete collection
of contemporary approaches to human sentence processing. Session 3: Selected approaches in human sentence processing We will pick some of the currently most influencial approaches
and discuss them in more detail. How does a model fit to what
is known about the general cognitive architecture? What is its
rationale? How well do the predictions meet recent empirical data? Session 4: Modifier attachment Recent research on modifier attachment will be discussed on the
basis of current models of human sentence processing. Session 5: Anaphoric aspects of ambiguity resolution In this session, the modifier attachment approach outlined in
session 4 will be generalized towards a theory of human parsing
and anaphor resolution. Anaphoric aspects of argument binding
as well as discourse processing will be highlighted and put forth
in a general account of ambiguity resolution.
COMPUTATIONAL APPROACHES TO HUMAN SENTENCE PROCESSING
lkon@coli.uni-sb.de
None
No specific recommendation