| |
Abstract:
This project aims at collecting and classifying data known
from psycholinguistic literature and providing the means to easily
access and use the data to evaluate theories and models of human
sentence processing. We are establishing a collection of core
examples which are annotated with parse trees according to the
scheme employed in the Penn Treebank (Bies et al., 1995). Local
ambiguities are preserved by providing multiple versions (overlays)
of the same annotated ambiguous substring. Although sentences are
statically classified for the type of phenomena (following the
scheme suggested by Lewis, 1993), users will also be able to submit
inividual queries at a variing level of detail, in order to keep
the use of the database as flexible as possible. For instance,
users might want to ask for all examples which include an ambiguous
NP which can be attached to a head either to its left or right. The
search result will be a collection of items, each providing a link
to the relevant annotated parse tree, a list of literature
references where the examples can be found, and a summary of the
data including the paradigms used. This database will be made
available on the web in order to establish a broad user base.
Researchers will be encouraged to submit the results of their own
studies to the database, thus making them easily accessible to the
scientific community. By presenting this poster, we hope to receive
feedback about the usability of the database and suggestions for
further desirable features.
Bies, A., Ferguson, M., Katz, K., & MacIntyre, R. (1995).
Bracketing guidelines for Treebank II style (Penn Treebank
Project). Manuscript, Department of Computer and Information
Science, University of Pennsylvania, Philadelphia, PA, USA.
Lewis, R. (1993). An architecturally-based theory of human
sentence processing. Doctoral dissertation, Carnegie Mellon
University, Pittsburgh, PA.
|
