3-Way composition of weighted finite-state transducers

Cyril Allauzen, Mehryar Mohri

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Composition of weighted transducers is a fundamental algorithm used in many applications, including for computing complex edit-distances between automata, or string kernels in machine learning, or to combine different components of a speech recognition, speech synthesis, or information extraction system. We present a generalization of the composition of weighted transducers, 3-way composition, which is dramatically faster in practice than the standard composition algorithm when combining more than two transducers. The worst-case complexity of our algorithm for composing three transducers T 1, T 2, and T 3 resulting in T, is O(|T| Q min (d(T 1) d(T 3), d(T 2))∈+∈|T| E ), where |•| Q denotes the number of states, |•| E the number of transitions, and d(•) the maximum out-degree. As in regular composition, the use of perfect hashing requires a pre-processing step with linear-time expected complexity in the size of the input transducers. In many cases, this approach significantly improves on the complexity of standard composition. Our algorithm also leads to a dramatically faster composition in practice. Furthermore, standard composition can be obtained as a special case of our algorithm. We report the results of several experiments demonstrating this improvement. These theoretical and empirical improvements significantly enhance performance in the applications already mentioned.

Original languageEnglish (US)
Title of host publicationImplementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings
Pages262-273
Number of pages12
Volume5148 LNCS
DOIs
StatePublished - 2008
Event13th International Conference on Implementation and Application of Automata, CIAA 2008 - San Francisco, CA, United States
Duration: Jul 21 2008Jul 24 2008

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5148 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other13th International Conference on Implementation and Application of Automata, CIAA 2008
CountryUnited States
CitySan Francisco, CA
Period7/21/087/24/08

Fingerprint

Transducers
Transducer
Chemical analysis
Information Storage and Retrieval
Information Systems
Speech Synthesis
Speech synthesis
Edit Distance
Information Extraction
Hashing
Speech Recognition
Speech recognition
3D
Automata
Preprocessing
Learning systems
Linear Time
Machine Learning
Strings
kernel

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

Allauzen, C., & Mohri, M. (2008). 3-Way composition of weighted finite-state transducers. In Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings (Vol. 5148 LNCS, pp. 262-273). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5148 LNCS). https://doi.org/10.1007/978-3-540-70844-5_27

3-Way composition of weighted finite-state transducers. / Allauzen, Cyril; Mohri, Mehryar.

Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings. Vol. 5148 LNCS 2008. p. 262-273 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5148 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Allauzen, C & Mohri, M 2008, 3-Way composition of weighted finite-state transducers. in Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings. vol. 5148 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5148 LNCS, pp. 262-273, 13th International Conference on Implementation and Application of Automata, CIAA 2008, San Francisco, CA, United States, 7/21/08. https://doi.org/10.1007/978-3-540-70844-5_27
Allauzen C, Mohri M. 3-Way composition of weighted finite-state transducers. In Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings. Vol. 5148 LNCS. 2008. p. 262-273. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-540-70844-5_27
Allauzen, Cyril ; Mohri, Mehryar. / 3-Way composition of weighted finite-state transducers. Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings. Vol. 5148 LNCS 2008. pp. 262-273 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{30d4e3aa510144d891e79c84d32d3f56,
title = "3-Way composition of weighted finite-state transducers",
abstract = "Composition of weighted transducers is a fundamental algorithm used in many applications, including for computing complex edit-distances between automata, or string kernels in machine learning, or to combine different components of a speech recognition, speech synthesis, or information extraction system. We present a generalization of the composition of weighted transducers, 3-way composition, which is dramatically faster in practice than the standard composition algorithm when combining more than two transducers. The worst-case complexity of our algorithm for composing three transducers T 1, T 2, and T 3 resulting in T, is O(|T| Q min (d(T 1) d(T 3), d(T 2))∈+∈|T| E ), where |•| Q denotes the number of states, |•| E the number of transitions, and d(•) the maximum out-degree. As in regular composition, the use of perfect hashing requires a pre-processing step with linear-time expected complexity in the size of the input transducers. In many cases, this approach significantly improves on the complexity of standard composition. Our algorithm also leads to a dramatically faster composition in practice. Furthermore, standard composition can be obtained as a special case of our algorithm. We report the results of several experiments demonstrating this improvement. These theoretical and empirical improvements significantly enhance performance in the applications already mentioned.",
author = "Cyril Allauzen and Mehryar Mohri",
year = "2008",
doi = "10.1007/978-3-540-70844-5_27",
language = "English (US)",
isbn = "354070843X",
volume = "5148 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "262--273",
booktitle = "Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings",

}

TY - GEN

T1 - 3-Way composition of weighted finite-state transducers

AU - Allauzen, Cyril

AU - Mohri, Mehryar

PY - 2008

Y1 - 2008

N2 - Composition of weighted transducers is a fundamental algorithm used in many applications, including for computing complex edit-distances between automata, or string kernels in machine learning, or to combine different components of a speech recognition, speech synthesis, or information extraction system. We present a generalization of the composition of weighted transducers, 3-way composition, which is dramatically faster in practice than the standard composition algorithm when combining more than two transducers. The worst-case complexity of our algorithm for composing three transducers T 1, T 2, and T 3 resulting in T, is O(|T| Q min (d(T 1) d(T 3), d(T 2))∈+∈|T| E ), where |•| Q denotes the number of states, |•| E the number of transitions, and d(•) the maximum out-degree. As in regular composition, the use of perfect hashing requires a pre-processing step with linear-time expected complexity in the size of the input transducers. In many cases, this approach significantly improves on the complexity of standard composition. Our algorithm also leads to a dramatically faster composition in practice. Furthermore, standard composition can be obtained as a special case of our algorithm. We report the results of several experiments demonstrating this improvement. These theoretical and empirical improvements significantly enhance performance in the applications already mentioned.

AB - Composition of weighted transducers is a fundamental algorithm used in many applications, including for computing complex edit-distances between automata, or string kernels in machine learning, or to combine different components of a speech recognition, speech synthesis, or information extraction system. We present a generalization of the composition of weighted transducers, 3-way composition, which is dramatically faster in practice than the standard composition algorithm when combining more than two transducers. The worst-case complexity of our algorithm for composing three transducers T 1, T 2, and T 3 resulting in T, is O(|T| Q min (d(T 1) d(T 3), d(T 2))∈+∈|T| E ), where |•| Q denotes the number of states, |•| E the number of transitions, and d(•) the maximum out-degree. As in regular composition, the use of perfect hashing requires a pre-processing step with linear-time expected complexity in the size of the input transducers. In many cases, this approach significantly improves on the complexity of standard composition. Our algorithm also leads to a dramatically faster composition in practice. Furthermore, standard composition can be obtained as a special case of our algorithm. We report the results of several experiments demonstrating this improvement. These theoretical and empirical improvements significantly enhance performance in the applications already mentioned.

UR - http://www.scopus.com/inward/record.url?scp=50249178637&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=50249178637&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-70844-5_27

DO - 10.1007/978-3-540-70844-5_27

M3 - Conference contribution

AN - SCOPUS:50249178637

SN - 354070843X

SN - 9783540708438

VL - 5148 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 262

EP - 273

BT - Implementation and Application of Automata - 13th International Conference, CIAA 2008, Proceedings

ER -