From NLG Systems Wiki

Jump to: navigation, search

[edit] Summary

  • System Name: CONPAS
  • Short Description: Content-to-Prosodically Adequate Speech
  • System Builders: Marsi[1][2][3][4]
  • Development Period: 1995–2000
  • Languages: Dutch
  • Domain: prosody generation


CONPAS (Content-to-Prosodically Adequate Speech) is a front-end generator for producing spoken output with appropriate prosody. It consists of three modules: SynR, a module for syntactic realization, ProsR, a module for prosodic realization and, SpeechS, a module for speech synthesis. It is driven by a planning component (PlanR) and produces spoken output in a number of domains; the text planner is specific for each domain. Particular emphasis was placed on the pitch accent placement, and this has been evaluated.


  1. Marsi, E. (2001). Intonation in spoken language generation. Utrecht: LOT (Landelijke Onderzoekschool Taalwetenschap). Bib
  2. Marsi, E. (1998). A reusable syntactic generator for Dutch. Paper presented at Computational Linguistics in the Netherlands 1997. Selected papers from the Eighth CLIN Meeting), Amsterdam. Bib
  3. Marsi, E. (1995). Intonation in a spoken language generator. Paper presented at Proceedings 1995, Nijmegen. Bib
  4. Marsi, E. (1998). Introducing maximal variation in text planning for small domains. Paper presented at Proceedings of the Ninth International Workhop on Natural Language Generation (INLG-98), Niagara-On-The-Lake, Ontario, Canada. Bib
Facts about CONPASRDF feed
DescriptionContent-to-Prosodically Adequate Speech  +
Domainprosody generation  +
Ended2000  +
LanguageDutch  +
Started1995  +
WorkerMarsi  +
Personal tools