From NLG Systems Wiki

Jump to: navigation, search

[edit] Summary

  • System Name: OxyGen
  • Short Description: A language independent Linearization Engine
  • System Builders: Habash[1][2][3][4]
  • Development Period: 1999–2002
  • Languages: independent
  • Forerunner: Nitrogen


A language independent Linearization Engine. OxyGen takes an input expression in the form of the Abstract Meaning Representation (AMR) developed for the Nitrogen generator and works out a linear ordering for the specified lexical items. The output of the linearization process is a word lattice. A language (oxyL) is specified for writing linearization rules. The linearization process is hybrid, combining symbolic and statistical information. The system has been used in machine translation efforts such as ChinMT and others from Bonnie Dorr.


  1. Habash, N. (2000). OxyGen: A Language Independent Linearization Engine. Paper presented at AMTA. Bib
  2. Habash, N. (2001). A reference manual to the linearization engine oxyGen (Technical Report). Maryland: University of Maryland, College Park. Bib
  3. Habash, N., Dorr, B. J., & Traum, D. (2001). Efficient Language Independent Generation from Lexical Conceptual Structure (Technical ReportNo. LAMP-TR-074, CS-TR-4262, UMIACS-TR-2001-43). Maryland: University of Maryland, College Park. Bib
  4. Habash, N., & Dorr, B. J. (2001). Large Scale Language Independent Generation: Using Thematic Hierarchies. Paper presented at Proceedings of the MT-Summit, Santiago, Spain. Bib
Facts about OxygenRDF feed
DescriptionA language independent Linearization Engine  +
Ended2002  +
ForerunnerNitrogen  +
Languageindependent  +
NameOxyGen  +
Started1999  +
WorkerHabash  +
Personal tools