Hassan EL NAHAS
Towards a Long Sentences
Preprocessor
for the ATR English Grammar
Abstract:In general, natural-language parsers are extremely inaccurate on
very long sentences. In this report, a new approach will be outlined
to the problem of insuring accurate parses of long sentences by broad-coverage grammar/parsers, or at least by the ATR English Grammar
parser. In addition to an explanation of the overall strategy being
pursued, this report will include an overview on preliminary experimental results within a crucial area of this approach: reliably identifying grossly-categorized sentence types as a first, preprocessing step
in the parsing of long sentences.