RuCCS - Drs. Adam Jardine & Lesia Semenova - Double Feature Dialogue - Tuesday, October 13, 2026

Value of the Information & Strength of the Argument

The presentations offer strong value by linking theoretical computational constraints to empirical phenomena. Jardine’s argument is compelling: restricting the hypothesis space to finite-state transducers makes the learning problem tractable, and he provides a concrete case study with the English plural. The use of automata to model morphological and phonological mappings is elegant and supported by prior work. Semenova’s talk, though only described in the abstract, raises a crucial question about model interpretability, and her approach of analyzing data mechanics and analyst choices is a refreshing perspective. Both arguments are well-founded, though the transcript only partially captures the depth of the second talk.

Scientific Rigor, Source Quality, Title Accuracy

Jardine explicitly references the Chomsky hierarchy and cites specific researchers (Heinz, Chandlee) whose work underpins his approach. The logical flow from complexity classes to finite-state machines to learning algorithms is rigorous. The title of the video is purely administrative, not reflecting the content, but this does not detract from the scientific merit. The talk appears to adhere to academic standards, with clear terminology and formal definitions.

The title is purely descriptive, giving the event and speakers, but does not hint at the content. It remains adequate for an academic seminar announcement.

8/10

The talk presents rigorous computational and linguistic arguments, grounded in established frameworks (Chomsky hierarchy, finite-state automata) and recent research (Heinz, Chandlee). The reasoning is clear and well-structured, even though the transcript is partial.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and welcome; setup of the double-feature dialogue.
• 00:40Adam Jardine begins his talk on computational simplicity in language acquisition.
• 02:00Jardine poses the learning problem: how children infer underlying representations and phonological grammar from surface data.
• 05:00Introduction of the Chomsky hierarchy and argument that phonology resides in the regular class.
• 08:00Explanation of input strictly local processes, exemplified by the English plural alternation (cats vs. dogs).
• 12:00Proposal to model lexicon and phonological functions with finite-state machines, enabling simultaneous learning.
• 15:00Conclusion of Jardine's talk; transition to Lesia Semenova.

• Chomsky hierarchy — Mentioned as the framework for classifying complexity of patterns, referencing Choosenberger (i.e., Chomsky and Schützenberger).
• Heinz's thesis (2007) — Credited with proposing that phonology fits within the regular class and using that for learning.
• Jane Chandlee's dissertation — Her work on input strictly local processes is used to define the restricted hypothesis space.

• Heinz, J. (2007). The Inductive Learning of Phonotactic Patterns. — Jardine's approach builds on Heinz's thesis that phonological patterns are regular and learnable.
• Chandlee, J. (2014). Input-Strictly Local Functions. — Her characterization of phonology as input strictly local underpins the finite-state machine formalization.