KIT-MARKER is sponsored by the Deutsche Forschungsgemeinschaft (DFG)
within the "Language Production" framework. It is an inter-disciplinary
project combining linguistics and computer science; the project leaders
are Prof. Mahr (Computer Science) and Prof. Posner (Linguistics/Semiotics).
Goal: Natural language texts are to be automatically generated from
a formal knowledge representation; our special interest is in the connectives
that bind sentences together.
Background: Texts are characterized by the presence of certain discourse
relations, which hold between adjacent spans of text. Examples for such
relations are BACKGROUND, ELABORATION, and CAUSE. In the text, these relations
are often explicitly marked by conjunctions, adverbs, or complete phrases:
the discourse markers. The CAUSE markers, for example, include `because',
`since', `therefore', `for this reason', and many more. Depending on the
particular context of the utterance, one or another marker is to be preferred
over the other, similar ones.
In KIT-MARKER, we build a lexicon of such discourse markers, which can
be used for purposes of automatic text generation. A prototypical system
that combines existing generation software modules with a new sentence
planning module and the marker lexicon is being developed, so that the
inter-dependencies between the various decisions in text generation, with
an emphasis on marker choice, can be investigated and tested.
Information about our software environment here.
Further information in the following articles:
Email: marker@cs.tu-berlin.de
[Last update: hemsen, 26.01.99]