G2PEVAL - an evaluation suite for German grapheme-to-phoneme conversion

---

The following papers

• Inducing Probabilistic Syllable Classes Using Multivariate Clustering. Karin Müller, Bernd Möbius, and Detlef Prescher.
  to appear in Proceedings of the 37th Annual Meeting of the ACL, 2000, Hongkong, 2000, Hongkong, China. (.ps/ .ps.gz)
  
  
• Inducing Probabilistic Syllable Classes Using Multivariate Clustering. - GOLD - Karin Müller, Bernd Möbius, and Detlef Prescher.
  to appear in AIMS Report 6 (2) Institute of Natural Lanugage Processing, 2000 University of Stuttgart. (.ps/ .ps.gz)
  

present an approach to automatic detection of syllable structure. We demonstrate a novel application of EM-based clustering to multivariate data, exemplified by the induction of 3- and 5-dimensional probabilistic syllable classes. The qualitative evaluation shows that the method yields phonologically meaningful syllable classes. We then propose a novel approach to grapheme-to-phoneme conversion and show that syllable structure represents valuable information for pronunciation systems. You can get our:

• German g2p evaluation suite
  
  
• 3-dimensional English model (gz)
• 5-dimensional English model (gz)
  
  
• 3-dimensional German model (gz)
• 5-dimensional German model (gz)

---

local: Demo