An efficient incremental architecture for robust interpretation

  • Authors:
  • Carolyn P. Rosé;Antonio Roque;Dumisizwe Bhembe

  • Affiliations:
  • University of Pittsburgh, Pittsburgh, PA;University of Pittsburgh, Pittsburgh, PA;University of Pittsburgh, Pittsburgh, PA

  • Venue:
  • HLT '02 Proceedings of the second international conference on Human Language Technology Research
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper we describe ROSENHEIM, the RObust SENtence level arcHitecture for Efficient Incremental Multi-sentence interpretation. Its efficiency is the result of its novel layered scheduling algorithm. This scheduling algorithm makes it possible to ensure that when edits are made to already processed portions of text, only the minimum changes are made to the chart. Thus, it allows the parser to take advantage of the user's typing or speaking time so that it can complete the majority of the work of parsing an input text before the user enters the last word. Our evaluation demonstrates that this layered architecture makes the parser's response time almost imperceptible. We also show that our incremental parser performs more efficiently than other well known incremental parsers [34] [20]. An implementation of the ROSENHEIM architecture is currently part of the WHY Conceptual Physics Tutoring System [28], one of several recent dialogue based tutoring Systems [23] [1] [7] [11].