Elamur Research⌕
☰
home / papers / 2014 / Patterns in the English Language: Phonological Networks, Percolation and Assembly Models

Patterns in the English Language: Phonological Networks, Percolation and Assembly Models

Paper · 2014 · Massimo Stella and Markus Brede

Summary

The paper asks which properties of the English phonological network (PN) reflect how real vocabularies are organised, and which simply follow from how such networks are built. Each word is a node, and two words are linked if their phonological transcriptions differ by one edit. The authors build the PN from about 30,000 English words with IPA transcriptions. They then compare it with a series of null models of increasing strictness:

  • shuffled and percolation-style pseudolexica;
  • MCMC-like ensembles forced to match the real link count;
  • growth (attachment) models.

The degree distribution's power-law-like part turns out to be a side effect of the underlying space of phoneme strings. Random networks that match the PN's connectivity, however, are very hard to obtain. This suggests that the vocabulary grew by adding words that are small variations of existing ones. Even the matched models leave gaps, which point to extra constraints: real lexica avoid very high degrees, closed triangles and large non-giant clusters. The work matters because it gives a principled way to separate genuine linguistic structure from construction artifacts.

Key points

  • Data come from a Wolfram Research database built mainly on WordNet 3.0 and the British National Corpus (v3). Stress and accent marks were stripped, and homophones were removed before linking words at edit distance one.

  • The statistics closely match earlier networks built from the 20,000-word Hoosier Mental Lexicon:

    StatisticThis PNHoosier Mental Lexicon
    Share of nodes in giant component33%34%
    Clustering coefficient0.210.22
    Assortativity0.700.67
    Mean path length7.716.08

  • Small "linguistic islands" average 2.49 ± 0.04 words, close to an earlier estimate of about 2.52. The giant component is a small world with a cut-off power-law-like degree distribution.

  • The null models are of four kinds: site percolation, randomization, MCMC-style ensembles with fixed link counts, and attachment/growth models. The growth models reproduce high clustering, high assortativity and small-world structure qualitatively.

  • Against properly matched null models, the real English PN shows fewer closed triangles and a stronger tendency for similar-degree words to link than would be expected.

Interesting ideas

  • I1. Higher-order statistics such as clustering or assortativity are only judged after the null model already matches lower-order ones (link count, component sizes, degree sequence), so a difference cannot simply come from a mismatched baseline.
  • I2. The real lexicon is treated as the occupied subset of a space of all possible symbol strings, which separates effects of the space itself from effects of word choice.
  • I3. Growth models in which each new item is a small modification of an existing one explain connectivity far better than independent random generation does.
  • I4. Comparing a network with randomized ensembles exposes which structure is a true signal and which is an artifact of how the graph was constructed.

For Mova

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

  • I1 (learn): Our negative-control evaluations could compare a parser or repair rule against baselines matched on simple statistics (sentence length, label frequencies) before crediting it with structural gains.
  • I2 (tokens): Unknown-word handling could explicitly model "possible but unrealized" forms and use edit-distance-one neighbours of known lemmas to guess morphology or fix typos.
  • I3 (learn): Self-induced repair rules could be proposed as minimal edits of rules that already work, which keeps the rule set connected and easier to explain.
  • I4 (world): We could check our WordNet/FrameNet/Wikidata knowledge graph against degree-matched random graphs to find clusters or hubs that only exist because of the merge procedure.

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 SouthamptonSouthampton · United Kingdom4 works · 2,409 citations

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Social Sciences · 744 worksPhysics and Astronomy · 25 worksStatistical and Nonlinear Physics · 22 worksCultural Studies · 10 worksEconomics and Econometrics · 2 worksEconomics, Econometrics and Finance · 2 works

🧭 Topics 1Research subjects and languages the work is about.English · 1,373 works

Cite

Cite as Massimo Stella and Markus Brede. 2014. Patterns in the English Language: Phonological Networks, Percolation and Assembly Models. arXiv:1410.4445.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1088/1742-5468/2015/05/p05006 · arXiv 1410.4445

@misc{stella2014patterns,
  title = {{Patterns in the English Language: Phonological Networks, Percolation and Assembly Models}},
  author = {Massimo Stella and Markus Brede},
  year = {2014},
  doi = {10.1088/1742-5468/2015/05/p05006},
  eprint = {1410.4445},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.1088/1742-5468/2015/05/p05006}
}

References

47 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Newman M 2010 Networks: an introduction (Oxford University Press) find
  2. Barabási A L and Albert R 1999 Science 286 509–512 find
  3. Amaral L A N, Scala A, Barthelemy M and Stanley H E 2000 Proceedings of the National Academy of Sciences 97 11149–11152 find
  4. Ravasz E and Barabási A L 2003 Physical Review E 67 026112 find
  5. Peel L and Clauset A 2014 arXiv preprint arXiv:1403.0989 ↗
  6. Antonioni A and Tomassini M 2012 Advances in Complex Systems 15 find
  7. Hagen M, Kissling W D, Rasmussen C, Carstensen D, Dupont Y, Kaiser-Bunbury C, O’Gorman E, Olesen J, De Aguiar M, Brown L et al. 2012 Advances in Ecological Research 46 89–120 find
  8. Bullmore E and Sporns O 2009 Nature Reviews Neuroscience 10 186–198 find
  9. Sporns O, Chialvo D R, Kaiser M and Hilgetag C C 2004 Trends in cognitive sciences 8 418–425 find
  10. Dall’Asta L, Baronchelli A, Barrat A and Loreto V 2006 Physical Review E 74 036105 find
  11. Motter A E, de Moura A P, Lai Y C and Dasgupta P 2002 Physical Review E 65 065102 find
  12. Steyvers M and Tenenbaum J B 2005 Cognitive science 29 41–78 find
  13. Sigman M and Cecchi G A 2002 Proceedings of the National Academy of Sciences 99 1742–1747 find
  14. i Cancho R F and Solé R V 2001 Proceedings of the Royal Society of London. Series B: Biological Sciences 268 2261–2265 find
  15. Aitchison J 2012 Words in the mind: An introduction to the mental lexicon (John Wiley & Sons) find
  16. Griffiths T L, Steyvers M and Firl A 2007 Psychological Science 18 1069–1076 find
  17. Baronchelli A, Ferrer-i Cancho R, Pastor-Satorras R, Chater N and Christiansen M H 2013 Trends in cognitive sciences 17 348–360 find
  18. Solé R V and Seoane L F 2014 Santa Fe Institute Working Paper find
  19. De Deyne S and Storms G 2008 Behavior Research Methods 40 213–231 find
  20. i Cancho R F and Solé R V 2003 Proceedings of the National Academy of Sciences 100 788–791 find
  21. Vitevitch M S 2008 Journal of Speech, Language, and Hearing Research 51 408–422 find
  22. Fay D and Cutler A 1977 Linguistic Inquiry 505–520 find
  23. Luce P A and Pisoni D B 1998 Ear and hearing 19 1 find
  24. Sadat J, Martin C D, Costa A, Alario F et al. 2014 Cognitive psychology 68 33–58 find
  25. Vitevitch M S 1997 Language and Speech 40 211–228 find
  26. Vitevitch M S, Chan K Y and Goldstein R 2014 Cognitive psychology 68 1–32 find
  27. Vitevitch M S, Chan K Y and Roodenrys S 2012 Journal of memory and language 67 30–44 find
  28. Chan K Y and Vitevitch M S 2010 Cognitive science 34 685–697 find
  29. Arbesman S, Strogatz S H and Vitevitch M S 2010 International Journal of Bifurcation and Chaos 20 679–685 find
  30. Arbesman S, Strogatz S H and Vitevitch M S 2010 Entropy 12 327–337 find
  31. Siew C S 2013 Frontiers in psychology 4 find
  32. Hills T T, Maouene J, Riordan B and Smith L B 2010 Journal of memory and language 63 259–273 find
  33. Beckage N, Smith L and Hills T 2011 PloS one 6 e19348 find
  34. Stauffer D and Aharony A 1991 Introduction to percolation theory (Taylor and Francis) find
  35. Newman M E 2003 SIAM review 45 167–256 find
  36. Gruenenfelder T M and Pisoni D B 2009 Journal of Speech, Language, and Hearing Research 52 596–609 find
  37. Mandelbrot B 1953 Communication theory 84 find
  38. Miller G A 1995 Communications of the ACM 38 39–41 find
  39. Leech G 1992 Language Research 28 1–13 find
  40. Snijders T A 2002 Journal of Social Structure 3 1–40 find
  41. Robins G, Pattison P, Kalish Y and Lusher D 2007 Social networks 29 173–191 find
  42. Daraganova G, Pattison P, Koskinen J, Mitchell B, Bill A, Watts M and Baum S 2012 Social networks 34 6–17 find
  43. Ristad E S and Yianilos P N 1998 Pattern Analysis and Machine Intelligence, IEEE Transactions on 20 522–532 find
  44. Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D and Alon U 2002 Science 298 824–827 find
  45. Callaway D S, Hopcroft J E, Kleinberg J M, Newman M E and Strogatz S H 2001 Physical Review E 64 041902 find
  46. Storkel H L 2004 Applied Psycholinguistics 25 201–221 find
  47. Carlson M T, Bane M and Sonderegger M 2011 Global properties of the phonological networks in child and child-directed speech Proceedings of the 35th Boston University Conference on Language Development pp 97–109 find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/stella-2014-patterns-english-language-phonological-networks/