Publications

2022

1. Characterizing idioms: Conventionality and contingency Michaela Socolof, Jackie Chi Kit Cheung, Michael Wagner, and Timothy J. O’Donnell In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL 2022). Dublin, Ireland. 2022. [link]
2. Compositional generalization in dependency parsing Emily Goodwin, Timothy J. O’Donnell, Siva Reddy, and Dzmitry Bahdanau In Proceedings of the 60th annual Meeting of the Association for Computational Linguistics (ACL 2022). Dublin, Ireland. 2022. [link]
3. Evaluating distributional distortion in neural language modeling Benjamin LeBrun, Alessandro Sordoni, and Timothy J. O’Donnell In Proceedings of 10th International Conference on Learning Representations (ICLR 2022). 2022. [link]
4. Simplicity and learning to distinguish arguments from modifiers Leon Bergen, Edward Gibson, and Timothy J. O’Donnell Journal of Language Modeling. 2022. Accepted.
5. Synthesizing theories of human language with Bayesian program induction Kevin Ellis, Adam Albright, Armando Solar-Lezama, Joshua B. Tenenbaum, and Timothy J. O’Donnell Nature Communications. 1–13. 2022. [link]
6. Measuring morphological fusion using partial information decomposition Michaela Socolof, Jacob Louis Hoover, Alessandro Sordoni, Richard Futrell, and Timothy J. O’Donnell In Proceedings of the 29th International Conference on Computational Linguistics (COLING 2022). 2022.

2021

1. Linguistic dependencies and statistical dependence Jacob Louis Hoover, Wenyu Du, Alessandro Sordoni, and Timothy J. O’Donnell In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP 2021). Online and Punta Cana, Dominican Republic. Nov, 2021. [abstract] [arXiv] [link] [code]

   Are pairs of words that tend to occur together also likely to stand in a linguistic dependency? This empirical question is motivated by a long history of literature in cognitive science, psycholinguistics, and NLP. In this work we contribute an extensive analysis of the relationship between linguistic dependencies and statistical dependence between words. Improving on previous work, we introduce the use of large pretrained language models to compute contextualized estimates of the pointwise mutual information between words (CPMI). For multiple models and languages, we extract dependency trees which maximize CPMI, and compare to gold standard linguistic dependencies. Overall, we find that CPMI dependencies achieve an unlabelled undirected attachment score of at most ≈0.5. While far above chance, and consistently above a non-contextualized PMI baseline, this score is generally comparable to a simple baseline formed by connecting adjacent words. We analyze which kinds of linguistic dependencies are best captured in CPMI dependencies, and also find marked differences between the estimates of the large pretrained language models, illustrating how their different training schemes affect the type of dependencies they capture.

2. Systematic generalization with Edge Transformers Leon Bergen, Timothy J. O’Donnell, and Dzmitry Bahdanau In Advances in Neural Information Processing Systems (NeurIPS 2021). 2021. [link]
3. Jointly learning truth-conditional denotations and groundings using parallel attention Leon Bergen, Dzmitry Bahdanau, and Timothy J. O’Donnell arXiv. 2021. [link]

2020

1. Statistical evidence for learnable lexical subclasses in Japanese Takashi Morita, and Timothy J. O’Donnell Linguistic Inquiry. 87–120. 2020. [link]
2. Exploiting syntactic structure for better language modeling: A syntactic distance approach Wenyu Du, Zhouhan Lin, Yikang Shen, Timothy J. O’Donnell, Yoshua Bengio, and Yue Zhang In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL 2020). Online. 2020. [link]
3. Probing linguistic systematicity Emily Goodwin, Koustuv Sinha, and Timothy J. O’Donnell In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Online. 2020. [link]
4. The Jazz Harmony Treebank Daniel Harasim, C. Finkensiep, P. Ericson, Timothy J. O’Donnell, and Martin Rohrmeier In Proceedings of the 21st Annual Conference of the International Society for Music Information Retrieval (ISMIR 2020). Montreal, Canada. 2020. [link]
5. CLOSURE: Assessing systematic generalization of CLEVR models Dzmitry Bahdanau, Harm Vries, Timothy J. O’Donnell, Shikhar Murty, Philippe Beaudoin, Yoshua Bengio, and Aaron Courville arXiv. 2020. [link]
6. Towards ‘English’ phonetics: variability in the pre-consonantal voicing effect across English dialects and speakers James Tanner, Morgan Sonderegger, Jane Stuart-Smith, and Josef Fruehwald Frontiers in Artificial Intelligence. 1-15. 2020. [abstract] [DOI]

   Recent advances in access to spoken-language corpora and development of speech processing tools have made possible the performance of “large-scale” phonetic and sociolinguistic research. This study illustrates the usefulness of such a large-scale approach—using data from multiple corpora across a range of English dialects, collected, and analyzed with the SPADE project—to examine how the pre-consonantal Voicing Effect (longer vowels before voiced than voiceless obstruents, in e.g., bead vs. beat) is realized in spontaneous speech, and varies across dialects and individual speakers. Compared with previous reports of controlled laboratory speech, the Voicing Effect was found to be substantially smaller in spontaneous speech, but still influenced by the expected range of phonetic factors. Dialects of English differed substantially from each other in the size of the Voicing Effect, whilst individual speakers varied little relative to their particular dialect. This study demonstrates the value of large-scale phonetic research as a means of developing our understanding of the structure of speech variability, and illustrates how large-scale studies, such as those carried out within SPADE, can be applied to other questions in phonetic and sociolinguistic research.

7. Structured speaker variability in Japanese stops: Relationships within versus across cues to stop voicing James Tanner, Morgan Sonderegger, and Jane Stuart-Smith Journal of the Acoustical Society of America. 793-804. 2020. [abstract] [DOI]

   A number of recent studies have observed that phonetic variability is constrained across speakers, where speakers exhibit limited variation in the signalling of phonological contrasts in spite of overall differences between speakers. This previous work focused predominantly on controlled laboratory speech and on contrasts in English and German, leaving unclear how such speaker variability is structured in spontaneous speech and in phonological contrasts that make substantial use of more than one acoustic cue. This study attempts to both address these empirical gaps and expand the empirical scope of research investigating structured variability by examining how speakers vary in the use of positive voice onset time and voicing during closure in marking the stop voicing contrast in Japanese spontaneous speech. Strong covarying relationships within each cue across speakers are observed, while between-cue relationships across speakers are much weaker, suggesting that structured variability is constrained by the language-specific phonetic implementation of linguistic contrasts.

8. Recursive top-down production for sentence generation with latent trees Shawn Tan, Timothy J. O’Donnell, Alessandro Sordoni, and Aaron Courville In Findings of the Association for Computational Linguistics: EMNLP 2020. 2020. [link]

2019

1. Universality and diversity in human song Samuel A. Mehr, Manvir Singh, Dean Knox, Daniel M. Ketter, Daniel Pickens-Jones, Stephanie Atwood, Christopher Lucas, Nori Jacoby, Alena A. Egner, Erin J. Hopkins, Rhea M. Howard, Timothy J. O’Donnell, Steven Pinker, Max M. Krasnow, and Luke Glowacki Science. 1–17. 2019. [link]
2. Structured speaker variability in spontaneous Japanese stop contrast production James Tanner, Morgan Sonderegger, and Jane Stuart-Smith 666-670. Melbourne, Australia. 2019. [abstract]

   Studies of speaker variability in the realisation of stop voicing contrasts have demonstrated that differences across speakers are highly structured both within and across phonetic categories. These studies have focused on languages with similar voicing systems in scripted speech; it remains unclear how stop realisation varies in spontaneous speech more generally. This study examines speaker variability in two acoustic cues to stop voicing–Voice Onset Time and Voicing During Closure–in a corpus of spontaneous Japanese, a language undergoing change in its voicing contrast. Whilst speakers vary in both measures, this variability is highly structured: speakers with less aspirated stops are more likely to initiate voicing during the closure. However, no corresponding relationship is observed between how the two cues are used to mark the contrast. These findings extend previous work to demonstrate the structure of speaker variability in spontaneous speech.

3. Vowel duration and the voicing effect across English dialects James Tanner, Morgan Sonderegger, Jane Stuart-Smith, and The SPADE Data Consortium Toronto Working Papers in Linguistics. 1-13. 2019. [abstract] [DOI]

   The ‘voicing effect’ – the durational difference in vowels preceding voiced and voiceless consonants – is a well-documented phenomenon in English, where it plays a key role in the production and perception of the English final voicing contrast. Despite this supposed importance, little is known as to how robust this effect is in spontaneous connected speech, which is itself subject to a range of linguistic factors. Similarly, little attention has focused on variability in the voicing effect across dialects of English, bar analysis of specific varieties. Our findings show that the voicing of the following consonant exhibits a weaker-than-expected effect in spontaneous speech, interacting with manner, vowel height, speech rate, and word frequency. English dialects appear to demonstrate a continuum of potential voicing effect sizes, where varieties with dialect-specific phonological rules exhibit the most extreme values. The results suggest that the voicing effect in English is both substantially weaker than previously assumed in spontaneous connected speech, and subject to a wide range of dialectal variability.

4. Five ways in which computational models can help advancing Artificial Grammar Learning research Willem H. Zuidema, Robert M. French, Raquel G. Alhama, Kevin Ellis, Timothy J. O’Donnell, Tim Sainburg, and Timothy Q. Gentner Topics in Cognitive Science. 1–17. 2019. [link]
5. Harmonic syntax in time: Rhythm improves grammatical models of harmony Daniel Harasim, Timothy J. O’Donnell, and Martin Rohrmeier 2019. [link]
6. Morphological irregularity correlates with frequency Shijie Wu, Ryan Cotterell, and Timothy J. O’Donnell In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019). 2019. [link]
7. A thousand studies for the price of one: Accelerating psychological science with Pushkin. Joshua K. Hartshorne, Josh Leeuw, Mariela Jennings, Noah D. Goodman, and Timothy J. O’Donnell Behavioral Research Methods. 2019. [link]
8. Durational evidence that Tokyo Japanese vowel devoicing is not gradient reduction James Tanner, Morgan Sonderegger, and Francisco Torreira Frontiers in Psychology. 1-5. 2019. [abstract] [DOI]

   A central question in the Japanese high vowel devoicing literature concerns whether vowels are devoiced through a categorical process or via gradient reduction. Examining how vowel height and consonantal voicing condition phrase-internal CV duration in a corpus of spontaneous Tokyo Japanese, it was found that CVs containing high vowels are substantially shorter before voiceless consonants, whilst non-high vowels do not exhibit comparable shortening. This quantitative difference between CV durations suggests a controlled temporal compression of the CV, consistent with views that Japanese vowel devoicing is produced through a categorical process targeting high vowels preceding voiceless consonants, and supports previous observations made of elicited productions.

2018

1. A generalized parsing framework for generative models of harmonic syntax Daniel Harasim, Martin Rohrmeier, and Timothy J. O’Donnell In Proceedings of the 19th Annual Conference of the International Society for Music Information Retrieval (ISMIR 2018). Paris, France. 2018. [link]
2. A maximum likelihood model for the harmonic analysis of symbolic music Colin C. Aitken, Timothy J. O’Donnell, and Martin Rohrmeier In Proceedings of the 15th Sound and Music Computing Conference (SMC 2018). Limassol, Cyprus. 2018.
3. Boundary phenomena and variability in Japanese high vowel devoicing Oriana Kilbourn-Ceron, and Morgan Sonderegger Natural Language and Linguistic Theory. 175–217. 2018. Manuscript published online in 2017..
4. The emergence, progress, and impact of sound change in progress in Seoul Korean: implications for mechanisms of tonogenesis Hye-Young Bang, Morgan Sonderegger, Yoonjung Kang, Meghan Clayards, and Tae-Jin Yoon Journal of Phonetics. 120–144. 2018. Manuscript published online in 2017..
5. Mixed-effects design analysis for experimental phonetics James Kirby, and Morgan Sonderegger Journal of Phonetics. 70–85. 2018. [preprint]
6. Examining factors influencing the viability of automatic acoustic analysis of child speech Knowles, Thea and Clayards, Meghan and Sonderegger, Morgan Journal of Speech, Language, and Hearing Research. 2487–2501. 2018.
7. Model selection and phonological argumentation James Kirby, and Morgan Sonderegger 234–252. University of Chicago Press. Chicago. 2018.

2017

1. Evaluating hierarchies of verb argument structure with hierarchical clustering Jesse Mu, Joshua K. Hartshorne, and Timothy J. O’Donnell In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing (EMNLP 2017). Copenhagen, Denmark. 2017. [link]
2. Production planning and coronal stop deletion in spontaneous speech James Tanner, Morgan Sonderegger, and Michael Wagner Laboratory Phonology. 1–39. 2017.
3. Montreal Forced Aligner: trainable text-speech alignment using Kaldi Michael McAuliffe, Michaela Socolof, Sarah Mihuc, Michael Wagner, and Morgan Sonderegger In Proceedings of Interspeech 2017. 498–502. 2017.
4. Polyglot and Speech Corpus Tools: a system for representing, integrating, and querying speech corpora Michael McAuliffe, Elias Stengel-Eskin, Michaela Socolof, and Morgan Sonderegger In Proceedings of Interspeech 2017. 3887–3891. 2017.
5. A generative model of phonotactics Richard Futrell, Adam Albright, Peter Graff, and Timothy J. O’Donnell Transaction of the Association for Computational Linguistics. 73–86. 2017. [link]

2016

1. The medium-term dynamics of accents on reality television Morgan Sonderegger, Max Bane, and Peter Graff Language. 2016. Manuscript accepted for publication 10/2016; to be published in 2017. 53 pages..
2. Psych verbs, the linking problem, and the acquisition of language Joshua K. Hartshorne, Timothy J. O’Donnell, Yasutada Sudo, Miki Uruwashi, Miseon Lee, and Jesse Snedeker Cognition. 268–2888. 2016.

2015

1. Trajectories of voice onset time in spontaneous speech on reality TV Morgan Sonderegger In Proceedings of the 18th International Congress of Phonetic Sciences. Ed. for ICPhS 2015, The Scottish Consortium. Paper 903.1–5. The University of Glasgow. Glasgow, UK. 2015.
2. The private life of stops: VOT in a real-time corpus of spontaneous Glaswegian Jane Stuart-Smith, Morgan Sonderegger, Tamara Rathcke, and Rachel MacDonald Laboratory Phonology. 505–549. 2015.
3. Automatic Forced Alignment on Child Speech: Directions for Improvement Thea Knowles, Meghan Clayards, Morgan Sonderegger, Michael Wagner, Aparna Nadig, and Kris Onishi In Proceedings of Meetings on Acoustics. 25:060001. 2015.
4. The effect of word frequency on the timecourse of tonogenesis in Seoul Korean Hye-Young Bang, Morgan Sonderegger, Yoonjung Kang, Meghan Clayards, and Tae-Jin Yoon In Proceedings of the 18th International Congress of Phonetic Sciences. Ed. for ICPhS 2015, The Scottish Consortium. Paper 843.1–5. The University of Glasgow. Glasgow, UK. 2015.
5. Unsupervised Lexicon Discovery from Acoustic Input Chia-Ying Lee, Timothy J. O’Donnell, and James R. Glass Transaction of the Association for Computational Linguistics. 389–403. 2015.
6. A model of rapid phonotactic generalization. Tal Linzen, and Timothy J. O’Donnell In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP 2015). 2015.
7. Evaluating models of computation and storage in human sentence processing Luong Min-Thang, Timothy J. O’Donnell, and Noah D. Goodman In Proceedings of the Workshop on Cognitive Aspects of Computational Language Learning. 2015. [link]
8. Productivity and Reuse in Language: A Theory of Linguistic Computation and Storage Timothy J. O’Donnell The MIT Press. Cambridge, Massachusetts. 2015.
9. Production planning and coronal stop deletion in spontaneous speech James Tanner, Morgan Sonderegger, and Michael Wagner In Proceedings of the 18th International Congress of Phonetic Sciences. Ed. for ICPhS 2015, The Scottish Consortium. Paper 933.1–5. The University of Glasgow. Glasgow, UK. 2015.

2014

1. How children explore the phonological network in child-directed speech: A survival analysis of children’s first word productions Matthew Carlson, Morgan Sonderegger, and Max Bane Journal of Memory and Language. 159–180. 2014. [link] [preprint]

2013

1. Learning Non-concatenative morphology Michelle Fullwood, and Timothy J. O’Donnell Sofia, Bulgaria. 2013.
2. Arguments and Modifiers from the Learner’s Perspective Leon Bergen, Edward Gibson, and Timothy J. O’Donnell In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics. 115-119. Sofia, Bulgaria. 2013.

2011

1. Productivity and Reuse in Language Timothy J. O’Donnell, Jesse Snedeker, Joshua B. Tenenbaum, and Noah D. Goodman In Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Boston, MA. 2011.
2. Storage and computation in syntax: Evidence from relative clause priming. Melissa Troyer, Timothy J. O’Donnell, Evelina Fedorinko, and Edward Gibson In roceedings of the Thirty-Third Annual Conference of the Cognitive Science Society. 336–341. Boston, Massachusetts. 2011.

2009

1. Fragment Grammars: Exploring Computation and Reuse in Language Timothy J. O’Donnell, Noah D. Goodman, and Joshua B. Tenenbaum Cambridge, MA. 2009.

2007

1. Evolutionary Linguistics: a new look at an old landscape Marc D. Hauser, David Barner, and Timothy J. O’Donnell Language Learning and Development. 2007.

2005

1. Using Mathematical Models of Language Experimentally Timothy J. O’Donnell, Marc D. Hauser, and W. Tecumseh Fitch Trends in Cognitive Sciences. 284-289. Jun, 2005.