Compressed Storage of Sparse Finite-State Transducers

  • Authors:
  • George Anton Kiraz

  • Affiliations:
  • -

  • Venue:
  • WIA '99 Revised Papers from the 4th International Workshop on Automata Implementation
  • Year:
  • 1999

Quantified Score

Hi-index 0.01

Visualization

Abstract

This paper presents an eclectic approach for compressing weighted finite-state automata and transducers, with minimal impact on performance. The approach is eclectic in the sense that various complementary methods have been employed: row-indexed storage of sparse matrices, dictionary compression, bit manipulation, and lossless omission of data. The compression rate is over 83% with respect to the current Bell Labs finite-state library.