This week at the Montreal Computational and Qualitative Linguistics Lab meeting, Timothy O’Donnell will be presenting his Meditations on Compositional Structure, to makeup for last week’s postponement. This presentation attempts to synthesize several threads of work in a broader framework. We meet at 2:30 via zoom (if you are not on the MCQLL emailing list, please contact Emily Goodwin firstname.lastname@example.org for the meeting link).
McLing presents the second edition of VirtualLing, documenting our current virtual lives. Please send your photos and screenshots!
PhD student Vanna Willteron sends a photo of her tidied up office space, noting: “I am proud of how cozy and clean I managed to get it so I wanted to share. I do miss my office mates and the fast school wifi though!” Thanks Vanna, looks great!
Students, staff, and faculty gathered for a departmental Town Hall last week to discuss questions about the coming semester.
A small summer Malagasy reading group has formed for the summer, growing out of the Field Methods class last semester, and bringing in the expertise of Lisa Travis and Ileana Paul. If you’re interested in joining, please contact Jessica for details.
A new paper by James Tanner, Morgan Sonderegger, Jane Stuart-Smith (Glasgow), and Josef Fruehwald (Kentucky), entitled “Toward ‘English’ phonetics: variability in pre-consonantal voicing effect across English dialects and speakers” has just been published in the journal Frontiers in Artificial Intelligence: Language and Computation. Congrats all!
Welcome to the first of potentially many screenshot updates on our virtual linguistics lives. Since we don’t currently get to run into each other in the hall of 1085 Peel, please send screenshots of virtual meetings you have––classes, reading groups, small advising meetings––McLing wants them all.
Course lecturer Natália Brambatti Guzzo is teaching LING 355 ‘Language Acquisition’ in our first full semester online this summer. She delivers lecturers to the 50+ students via Zoom (pictured) and then posts the recordings online for the students who aren’t able to attend synchronously.
Faculty meetings continue by Zoom. Are they starting to miss room 117?
This week at the Montreal Computational and Qualitative Linguistics Lab meeting, Timothy O’Donnell will be presenting his Meditations on Compositional structure. This presentation attempts to synthesize several threads of work in a broader framework. We meet at 2:30 via zoom (if you are not on the MCQLL emailing list, please contact Emily Goodwin email@example.com for the meeting link).
McGill Linguistics graduate students had a successful round of funding applications! Students awarded grants include:
- Emily Goodwin, supervised by Tim O’Donnell and Siva Reddy of McGill and Dzmitry Bahdanau of ElementAI received a MITACs grant for a 4-month internship this summer. This project explores systematic generalization in neural models for semantic parsing.
- Jacob Hoover received a one-year Microsoft Research-Mila Collaboration grant supervised by Alessandro Sordoni and Tim O’Donnell.
- Bing’er Jiang, Clint Parker, and Michaela Socolof each received FRQSC bourse au doctoral en recherche grants. Michaela’s project is titled “Modeling the relationship between syntax and semantics in non-compositional structures”, Clint’s is “Grammar of the Shughni Language: A Community-based approach”, and Bing’er’s is “Computational Cognitive Modelling of Phonological Assimilation”.
Other student grants are not yet officially announceable, but stay tuned. Congratulations to all!
The next meeting of the Montreal Computational and Qualitative Linguistics Lab will take place on Wednesday May 6th at 2:30, via Zoom. Jacob Hoover will present an ongoing project on compositionality and predictability.
For abstract and more information see the MCQLL lab page. If you would like to participate but are not on the MCQLL or computational linguistics emailing list, contact firstname.lastname@example.org for the Zoom link.
The next meeting of the Montreal Computational and Qualitative Linguistics Lab will take place on Wednesday April 29th, at 2:30, via Zoom. Koustuv Sinha will present “Learning an Unreferenced Metric for Online Dialogue Evaluation” (ACL, 2020). For abstract and more information see the MCQLL lab page. If you would like to participate but are not on the MCQLL or computational linguistics emailing list, contact email@example.com for the Zoom link.
Congratulations to coauthors Oriana Kilbourn-Ceroun (PhD ’17), Meghan Clayards, and Michael Wagner, who have just had their paper “Predictability modulates pronunciation variants through speech planning effects: A case study on coronal stop realizations” accepted for publication in the journal Laboratory Phonology.
“Structured heterogeneity in Scottish stops over the twentieth century”, a paper by Morgan Sonderegger, Jane Stuart-Smith, Thea Knowles (McGill BA 2012, Asst Prof at U. Buffalo), Rachel Macdonald, and Tamara Rathcke, was published in the March issue Language. Congrats all!Abstract: How and why speakers differ in the phonetic implementation of phonological contrasts, and the relationship of this ‘structured heterogeneity’ to language change, has been a key focus over fifty years of variationist sociolinguistics. In phonetics, interest has recently grown in uncovering ‘structured variability’—how speakers can differ greatly in phonetic realization in nonrandom ways—as part of the long-standing goal of understanding variability in speech. The English stop voicing contrast, which combines extensive phonetic variability with phonological stability, provides an ideal setting for an approach to understanding structured variation in the sounds of a community’s language that illuminates both synchrony and diachrony. This article examines the voicing contrast in a vernacular dialect (Glasgow Scots) in spontaneous speech, focusing on individual speaker variability within and across cues, including over time. Speakers differ greatly in the use of each of three phonetic cues to the contrast, while reliably using each one to differentiate voiced and voiceless stops. Interspeaker variability is highly structured: speakers lie along a continuum of use of each cue, as well as correlated use of two cues—voice onset time and closure voicing—along a single axis. Diachronic change occurs along this axis, toward a more aspiration-based and less voicing-based phonetic realization of the contrast, suggesting an important connection between synchronic and diachronic speaker variation.
In recent years, we have seen major improvements to various Natural Language Processing tasks. Despite their human-level performance on benchmarking datasets, recent studies have shown that these models are vulnerable to adversarial examples. It is shown that these models are relying on spurious correlations that hold for the majority of examples and suffer from distribution shifts and fail on atypical or challenging test sets. Recent work has shown that large pre-trained models improve model robustness to spurious associations in the training data. We observe that superior performance of large pre-trained language models comes from their better generalization from a minority of training examples that resemble the challenging sets. Our study shows that multi-task learning with the right auxiliary tasks improves accuracy on adversarial examples without hurting in distribution performance. We show that this holds true for multi-modal task of Referring Expression Recognition and text-only tasks of Natural language inference and Paraphrase identification.
McLing is happy to report that PhD student Jeff LaMontagne has accepted a tenure-track position in the Department of French and Italian at Indiana University Bloomington, to begin this August. Congratulations Jeff!
McLing is happy to report that Jessica Coon and Stefan Keine’s paper “Feature Gluttony” has been accepted for publication in the journal Linguistic Inquiry.
This paper develops a new approach to a family of hierarchy effect-inducing configurations, with a focus on Person Case Constraint (PCC) effects, dative-nominative configurations, and copula constructions. The main line of approach in the recent literature is to attribute these effects to failures of phi-Agree or, more specifically, failures of nominal licensing or case checking. We propose that the problem in these configurations is unrelated to nominal licensing, but is instead the result of a probe participating in more than one Agree dependency, a configuration we refer to as feature gluttony. Feature gluttony does not in and of itself lead to ungrammaticality, but rather can create irresolvably conflicting requirements for subsequent operations. We argue that in the case of clitic configurations, a probe which agrees with more than one DP creates an intervention problem for clitic-doubling. In violations involving morphological agreement, gluttony in features may result in a configuration with no available morphological output.
Jacob Hoover and Mathieu Paillé traveled to Vancouver last weekend for the 2020 meeting of the West Coast Conference on Formal Linguistics (WCCFL), held at UBC. Jacob’s talk was titled “Accounting for variation in number agreement in Icelandic DAT-NOM constructions”, and Mathieu presented “Exhaustivity and the meaning of colour terms”.
This Tuesday, Francisco will lead a discussion of Baese-Berk et al.‘s (2019) “Not just a function of function words: Distal speech rate influences perception of prosodically weak syllables.” The reading is available at bit.ly/PReadingGroup. P* Reading Group takes place at 1pm on Tuesdays, in Room 117 of the Linguistics Building. All are welcome to attend!
We are pleased to announce that the next talk in our 2019-20 McGill Linguistics Colloquium Series will be by Laura Dilley (Michigan State University) on Friday, March 13 at 3:30 pm in the Education Building room 433.
The title of the talk is TBA. We will make an updated announcement shortly with the updated title and abstract. All are welcome to attend.