Elamur Research⌕
☰
home / papers / 2008 / ProPOSEL: a human-oriented prosody and PoS English lexicon for machine-learning and NLP

ProPOSEL: a human-oriented prosody and PoS English lexicon for machine-learning and NLP

Paper · 2008 · Claire Brierley and Eric Atwell

Summary

The paper presents ProPOSEL, an English lexicon that merges phonological, prosodic and part-of-speech information into one file. Its starting point is CUVPlus, enriched with data from CELEX-2 and CMU. Earlier pronunciation dictionaries were built for speech recognition and seldom record word class. The authors wanted a resource that can be joined directly to tagged corpora and used as prior knowledge in machine learning, especially for predicting prosodic phrase breaks in text-to-speech. Each of the 104,049 entries has 15 fields, including:

  • four PoS tagsets;
  • two phonetic transcriptions;
  • syllable count and lexical stress pattern;
  • a default content/function word label;
  • a consonant–vowel template.

CUVPlus packs several PoS variants into one field. The authors split these so that each (word form, C5 tag) pair becomes a unique key. Python/NLTK tools load the file as a dictionary and support filtered searches by sound, rhythm and word class. The paper reports no experiments, so the contribution is the resource itself. It shows how a plain-text dictionary can serve both NLP pipelines and people searching for words by partial form rather than meaning.

Key points

  • 104,049 entries, each with 15 pipe-separated fields. It is distributed as a text file and intended for open-source release with NLTK.
  • Four tagging schemes allow linking to different corpora:
    • C5 for the BNC;
    • Penn Treebank for Treebank-3;
    • LOB for MARSEC;
    • C7 for the BNC Sampler.
  • Pronunciation appears as SAM-PA and as syllabified, stressed DISC. Stress is coded with digits 0/1/2; for example, objectivity is 20,100.
  • Stress can depend on the tag. The noun/adjective present is 10, while the verb is 01.
  • Example searches:
    • four-syllable verbs stressed like obliterate returned 67 candidates;
    • other queries list all seven-syllable words or find rhymes for corpus using a regex over SAM-PA.

Interesting ideas

  • I1. Splitting entries that merged several PoS readings into one record per (form, tag) pair makes every value unambiguous and lets corpus tokens match by exact tuple lookup.
  • I2. Storing several tagsets side by side in one entry turns the lexicon into a word-level crosswalk between annotation schemes.
  • I3. A coarse default content/function flag is stored as a cheap, transparent feature, explicitly open to later revision by context.
  • I4. The authors frame the lexicon as a priori knowledge injected into data-driven models trained on corpora, rather than as a replacement for them.
  • I5. The dictionary can be searched from partial form — syllable count, stress shape, rhyme — not only by spelling or meaning.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (tokens): key the en lexicon on (form, UPOS) so lemma, morphology and other features attach to a specific reading and the tagger's choice directly selects the right entry.
  • I2 (tree): such parallel tag columns could check and explain Penn/C5-to-UD conversions when bringing older treebanks into our UD dialect.
  • I3 (tokens): a default content/function bit per lexeme gives the parser and chunker an explainable prior that context rules can override with a recorded reason.
  • I4 (learn): dictionary-derived features could act as a fixed prior alongside silver data, with negative controls showing whether the lexicon actually adds accuracy.
  • I5 (back): during generation or paraphrase, candidate words could be filtered by stress pattern or rhyme when retelling fables in verse-like or rhythmically similar wording.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of LeedsLeeds · United Kingdom40 works · 1,822 citations🎓University of Greater ManchesterBolton · United Kingdom2 works · 12 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.English · 2,263 worksSpeech & spoken language · 480 works

Cite

Cite as Claire Brierley and Eric Atwell. 2008. ProPOSEL: a human-oriented prosody and PoS English lexicon for machine-learning and NLP. Coling 2008: Proceedings of the Workshop on Cognitive Aspects of the Lexicon (COGALEX 2008).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{brierley2008proposel,
  title = {{ProPOSEL: a human-oriented prosody and PoS English lexicon for machine-learning and NLP}},
  author = {Claire Brierley and Eric Atwell},
  booktitle = {Coling 2008: Proceedings of the Workshop on Cognitive Aspects of the Lexicon (COGALEX 2008)},
  year = {2008},
  doi = {10.3115/1598848.1598854},
  url = {https://doi.org/10.3115/1598848.1598854}
}

References

23 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Atterer M., and E. Klein. 2002. Integrating Linguistic and Performance-Based Constraints for Assigning Phrase Breaks. In Proceedings of Coling 2002:29-35. find
  2. Atwell, E., G. Demetriou, J. Hughes, A. Schriffin, C. Souter, S. Wilcock. 2000. A comparative evaluation of modern English corpus grammatical annotation schemes. ICAME Journal, vol. 24, pp. 7-23. find
  3. Atwell, E. and A. Roberts. 2006. Combinatory hybrid elementary analysis of text. In Kurimo, M, Creutz, M & Lagus, K (editors) Proceedings of the PASCAL Challenge Workshop on Unsupervised Segmentation of Words into Morphemes. Venice. ↗
  4. Baayen, R. H., R. Piepenbrock, and L. Gulikers 1996. CELEX2 Linguistic Data Consortium, Philadelphia Bird, S., E. Loper, and E. Klein 2007a. NLTK-lite 0.8 beta [June 2007] Available online from : http://nltk.sourceforge.net/index.php/Main_Page (accessed: 21/06/07). ↗
  5. Bird, S., E. Klein, and E. Loper 2007b. Natural Language Processing Available online from: http://nltk.sourceforge.net/index.php/Book (accessed: 21/09/07). ↗
  6. Black A.W., P. Taylor, and R. Caley. 1999. The Festival Speech Synthesis System: System Documentation Festival version 1.4 Available online from: http://www.cstr.ed.ac.uk/projects/festival/manua l/festival_toc.html (Accessed: 07/03/08) ↗
  7. Brierley, C. and E. Atwell. 2007a. Corpus-based evaluation of prosodic phrase break prediction in: Proceedings of Corpus Linguistics 2007, Birmingham University. find
  8. Brierley, C. and E. Atwell. 2007b. An approach for detecting prosodic phrase boundaries in spoken English. ACM Crossroads journal, vol. 14.1. ↗
  9. Brierley, C. and E. Atwell. 2007c. Prosodic phrase break prediction: problems in the evaluation of models against a gold standard. Traitement Automatique des Langues, vol. 48.1. find
  10. Brierley, C. and E. Atwell. 2008 ProPOSEL: a Prosody and POS English Lexicon for Language Engineering. In Proceedings of LREC’08 Language Resources and Evaluation Conference, Marrakech, Morocco. May 2008.
  11. Burnard, L. (ed.) 2000. Reference Guide for the British National Corpus (World Edition) Available online from: http://www.natcorp.ox.ac.uk/docs/userManual/(accessed: 20/05/07). ↗
  12. Busser, B. W. Daelemans, and A. van den Bosch 2001. Predicting phrase breaks with memory-based learning. 4 th ISCA Tutorial and Research Workshop on Speech Synthesis. find
  13. Carnegie-Mellon University 1998. The CMU Pronouncing Dictionary (v. 0.6) Available online from: http://www.speech.cs.cmu.edu/cgi-bin/cmudict (accessed: 21/06/07). ↗
  14. Hartinkainen, E., G. Maltese, A. Moreno, S. Sham-mass, U. Ziegenhain 2003. Large Lexica for Speech-to-Speech Translation: frm specification to creation. EUROSPEECH-2003:1529-1532. find
  15. Hornby, A.S. 1974. Oxford Advanced Learner’s Dictionary of Current English (third edition) Oxford: Oxford University Press Johansson, S; Atwell, E S; Garside, R; Leech, G. 1986. The Tagged LOB Corpus - User Manual, 160pp, Bergen, Norwegian Computing Centre for the Humanities. find
  16. Kingsbury, P., S. Strassel, C. McLemore, and R. MacIntyre 1997. CALLHOME American English Lexicon (PRONLEX) Linguistic Data Consortium, Philadelphia Liberman, M.Y., and K.W. Church 1992. Text Analysis and Word Pronunciation in Text-to-Speech Synthesis. In Furui, S., and Sondhi, M.M., (eds.) Advances in Speech Signal Processing New York, Marcel Dekker, Inc. find
  17. Marcus, M.P., B. Santorini, M.A. Marcinkiewicz, and A. Taylor 1999. TREEBANK-3 Linguistic Data Consortium, Philadelphia Martelli, A., A. Martelli Ravenscroft, and D. Ascher 2005. Python Cookbook (second edition) Sebastopol: O’Reilly Media, Inc. find
  18. Mitton, R. 1992. A description of a computer-usable dictionary file based on the Oxford Advanced Learner’s Dictionary of Current English Available online and accessed (22/03/08) from: http://comp.lin.msu.edu/stabler-notes/1850/ascii_0710-2.txt ↗
  19. Mortimer, C. 1985. Elements of Pronunciation. Cambridge: Cambridge University Press Nancarrow, O. and E. Atwell. 2007.A comparative study of the tagging of adverbs in modern English corpora Proceedings of Corpus Linguistics 2007. Birmingham University. find
  20. Pedler, J. 2002. CUVPlus [Electronic Resource] Oxford Text Archive Available online from: http://ota.ahds.ac.uk/textinfo/2469.html (accessed: 21/06/07) ↗
  21. Roach P., G. Knowles, T. Varadi and S.C. Arnfield. 1993. Marsec: A machine-readable spoken English corpus Journal of the International Phonetic Association, vol. 23, no. 1, pp. 47—53. find
  22. Schryver, G. M. de. 2003. Lexicographers’ Dreams in the Electronic-Dictionary Age. International Journal of Lexicography 2003 16(2):143-199 find
  23. Teufel, S.. 1995. A support tool for tagset mapping. Proceedings of SIGDAT 1995. Workshop in cooperation with EACL 95, Dublin ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/brierley-2008-proposel-human-oriented-prosody-pos/