Maryland at SCiL and ACL

On speech perception, parsing, learning and language models.

July 3 and 4, the 9th meeting of Society for Computation in Linguistics (SCiL) takes place in sunny San Diego, California, co-located with the Annual Meeting of the Association for Computational Linguistics (ACL). This year's SCiL is co-organized by our own Naomi Feldman, who won an NSF grant to help SCiL authors attend the ACL. There are three presentations from our PhD students, Annika and Utku:

• Annika Shankwitz / Small neural networks as models of cross-linguistic speech perception
• Utku Türk, Eva Neu /  Quantifying the cross-linguistic effects of syncretism on agreement attraction 
• Utku Türk, Eva Neu, Özge Bakay, Brian Dillon *11 and Gaja Jarosz / Frequency modulates structural choice in Turkish suspended affixation: A latent-process account

Meanwhile over at the ACL main session, Philip has the honor of giving a keynote address on a topical issue:

• Philip Resnik / A New Balancing Act: Reflections on the Relationship between Computational Linguistics and AI

The main sessions have two papers involving Sathvik Nair, and one from Hillary Owusu, an advisee of Naomi's in the computer science department.

• Atrey Desai and Sathvik Nair / Filling in the Mechanisms: How do LMs learn Filler-Gap Dependencies under Developmental Constraints?
• Sathvik Nair and Byung-Doh Oh / Clozing the Gap: Exploring Why Language Model Surprisal Explains Cloze Surprisal
• Hillary Owusu and Naomi H. Feldman / Anchoring depends on confidence and post-training in Language Models

And finally there is action from our alums. The Class of 2018's Allyson Ettinger is featured in a symposium on "Linguistics and NLP in the LLM Era, while 2011 alum Brian Dillon has two papers with colleagues at UMass.

• Allyson Ettinger, Richard Futrell, Zoey Liu and Tom McCoy / Linguistics and NLP in the LLM Era
• Weiji Xu, Brian Dillon and Richard Futrell / Memory efficiency and resource-rational encoding in sentence processing
• Farah Adeeba, Brian Dillon, Hassan Sajjad and Rajesh Bhatt / UrBLiMP: A Benchmark for Evaluating the Linguistic Competence of Large Language Models in Urdu