Assigning stress to out-of-vocabulary words: Three approaches

Manex Agirrezabal; Jeffrey Heinz; Mans Hulden; Bertol Arrieta

Assigning stress to out-of-vocabulary words: Three approaches

Manex Agirrezabal
, Jeffrey Heinz
, Mans Hulden
, Bertol Arrieta

Linguistics

Stony Brook University

University of the Basque Country
University of Colorado Boulder

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

2 Scopus citations

Abstract

In this paper we address the task of automatically assigning primary stress to out-of-vocabulary words in English. This work forms a necessary component in a scansion system for English poetry. We propose three different approaches based on (1) word similarity, (2) handwritten linguistic rules and (3) machine learning. The first and last approach require stress-annotated corpora to train a model for stress assignment, while the linguistic approach relies on grapheme to phoneme conversion, a syllabification procedure and hand-written stress assignment rules. An implementation of each approach is provided and the precision of the systems is compared. The linguistic approach proves to be the most effective, but the machine learning approach is not far behind. We anticipate that including part of speech information will improve the accuracy of each system. The source code of the approaches is released¹ under the GNU GPL license.

Original language	English
Title of host publication	Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014
Editors	Hamid R. Arabnia, David de la Fuente, Elena B. Kozerenko, Peter M. LaMonica, Raymond A. Liuzzi, Jose A. Olivas, Todd Waskiewicz, George Jandieri, Ashu M.G. Solo, Fernando G. Tinetti
Publisher	CSREA Press
Pages	574-576
Number of pages	3
ISBN (Electronic)	1601322763, 9781601322760
State	Published - 2014
Event	2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014 - Las Vegas, United States Duration: Jul 21 2014 → Jul 24 2014

Publication series

Name	Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014

Conference

Conference	2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014
Country/Territory	United States
City	Las Vegas
Period	07/21/14 → 07/24/14

Cite this

Agirrezabal, M., Heinz, J., Hulden, M., & Arrieta, B. (2014). Assigning stress to out-of-vocabulary words: Three approaches. In H. R. Arabnia, D. de la Fuente, E. B. Kozerenko, P. M. LaMonica, R. A. Liuzzi, J. A. Olivas, T. Waskiewicz, G. Jandieri, A. M. G. Solo, & F. G. Tinetti (Eds.), Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014 (pp. 574-576). (Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014). CSREA Press.

Agirrezabal, Manex ; Heinz, Jeffrey ; Hulden, Mans et al. / Assigning stress to out-of-vocabulary words : Three approaches. Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014. editor / Hamid R. Arabnia ; David de la Fuente ; Elena B. Kozerenko ; Peter M. LaMonica ; Raymond A. Liuzzi ; Jose A. Olivas ; Todd Waskiewicz ; George Jandieri ; Ashu M.G. Solo ; Fernando G. Tinetti. CSREA Press, 2014. pp. 574-576 (Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014).

@inproceedings{d6bd852ed9134ba682696f469efb87e1,

title = "Assigning stress to out-of-vocabulary words: Three approaches",

abstract = "In this paper we address the task of automatically assigning primary stress to out-of-vocabulary words in English. This work forms a necessary component in a scansion system for English poetry. We propose three different approaches based on (1) word similarity, (2) handwritten linguistic rules and (3) machine learning. The first and last approach require stress-annotated corpora to train a model for stress assignment, while the linguistic approach relies on grapheme to phoneme conversion, a syllabification procedure and hand-written stress assignment rules. An implementation of each approach is provided and the precision of the systems is compared. The linguistic approach proves to be the most effective, but the machine learning approach is not far behind. We anticipate that including part of speech information will improve the accuracy of each system. The source code of the approaches is released1 under the GNU GPL license.",

author = "Manex Agirrezabal and Jeffrey Heinz and Mans Hulden and Bertol Arrieta",

note = "Publisher Copyright: {\textcopyright}2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014.All right reserved.; 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014 ; Conference date: 21-07-2014 Through 24-07-2014",

year = "2014",

language = "English",

series = "Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014",

publisher = "CSREA Press",

pages = "574--576",

editor = "Arabnia, \{Hamid R.\} and \{de la Fuente\}, David and Kozerenko, \{Elena B.\} and LaMonica, \{Peter M.\} and Liuzzi, \{Raymond A.\} and Olivas, \{Jose A.\} and Todd Waskiewicz and George Jandieri and Solo, \{Ashu M.G.\} and Tinetti, \{Fernando G.\}",

booktitle = "Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014",

}

Agirrezabal, M, Heinz, J, Hulden, M & Arrieta, B 2014, Assigning stress to out-of-vocabulary words: Three approaches. in HR Arabnia, D de la Fuente, EB Kozerenko, PM LaMonica, RA Liuzzi, JA Olivas, T Waskiewicz, G Jandieri, AMG Solo & FG Tinetti (eds), Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014. Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014, CSREA Press, pp. 574-576, 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014, Las Vegas, United States, 07/21/14.

Assigning stress to out-of-vocabulary words: Three approaches. / Agirrezabal, Manex; Heinz, Jeffrey; Hulden, Mans et al.
Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014. ed. / Hamid R. Arabnia; David de la Fuente; Elena B. Kozerenko; Peter M. LaMonica; Raymond A. Liuzzi; Jose A. Olivas; Todd Waskiewicz; George Jandieri; Ashu M.G. Solo; Fernando G. Tinetti. CSREA Press, 2014. p. 574-576 (Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Assigning stress to out-of-vocabulary words

T2 - 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014

AU - Agirrezabal, Manex

AU - Heinz, Jeffrey

AU - Hulden, Mans

AU - Arrieta, Bertol

PY - 2014

Y1 - 2014

N2 - In this paper we address the task of automatically assigning primary stress to out-of-vocabulary words in English. This work forms a necessary component in a scansion system for English poetry. We propose three different approaches based on (1) word similarity, (2) handwritten linguistic rules and (3) machine learning. The first and last approach require stress-annotated corpora to train a model for stress assignment, while the linguistic approach relies on grapheme to phoneme conversion, a syllabification procedure and hand-written stress assignment rules. An implementation of each approach is provided and the precision of the systems is compared. The linguistic approach proves to be the most effective, but the machine learning approach is not far behind. We anticipate that including part of speech information will improve the accuracy of each system. The source code of the approaches is released1 under the GNU GPL license.

AB - In this paper we address the task of automatically assigning primary stress to out-of-vocabulary words in English. This work forms a necessary component in a scansion system for English poetry. We propose three different approaches based on (1) word similarity, (2) handwritten linguistic rules and (3) machine learning. The first and last approach require stress-annotated corpora to train a model for stress assignment, while the linguistic approach relies on grapheme to phoneme conversion, a syllabification procedure and hand-written stress assignment rules. An implementation of each approach is provided and the precision of the systems is compared. The linguistic approach proves to be the most effective, but the machine learning approach is not far behind. We anticipate that including part of speech information will improve the accuracy of each system. The source code of the approaches is released1 under the GNU GPL license.

UR - https://www.scopus.com/pages/publications/85054987387

M3 - Conference contribution

T3 - Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014

SP - 574

EP - 576

BT - Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014

A2 - Arabnia, Hamid R.

A2 - de la Fuente, David

A2 - Kozerenko, Elena B.

A2 - LaMonica, Peter M.

A2 - Liuzzi, Raymond A.

A2 - Olivas, Jose A.

A2 - Waskiewicz, Todd

A2 - Jandieri, George

A2 - Solo, Ashu M.G.

A2 - Tinetti, Fernando G.

PB - CSREA Press

Y2 - 21 July 2014 through 24 July 2014

ER -

Agirrezabal M, Heinz J, Hulden M, Arrieta B. Assigning stress to out-of-vocabulary words: Three approaches. In Arabnia HR, de la Fuente D, Kozerenko EB, LaMonica PM, Liuzzi RA, Olivas JA, Waskiewicz T, Jandieri G, Solo AMG, Tinetti FG, editors, Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014. CSREA Press. 2014. p. 574-576. (Proceedings of the 2014 International Conference on Artificial Intelligence, ICAI 2014 - WORLDCOMP 2014).

Assigning stress to out-of-vocabulary words: Three approaches

Abstract

Publication series

Conference

Other files and links

Fingerprint

Cite this