I will discuss some recent findings on the impact of corpus size on the distribution of items in train-test splits for morphological inflection tasks. This can be tested empirically by taking many samples or calculated analytically. I will present and discuss some empirical results and an analytic solution.
What defines an author’s writing style such that we can differentiate them from other authors? As humans, stylometric choices (i.e. the linguistic choice we make when writing) are innate, meaning we subconsciously make them without much thought. How can we build a system to automatically detect these stylistic …
Thomas will present a concrete case in syntax where subregular complexity considerations favor one of two competing analyses even though both have the same complexity. Concretely, both the Ban on Improper Movement and Stefan Keine's probe horizons are TSL over dependency trees, but they differ in succinctness. With probe horizons …
I will talk about a project me and Thomas contributed to using a probabilistic version of TSL* for trees, combining Thomas's tree TSL with Connor Mayer's string pTSL. We fit a model to experimental data on island effects from Sprouse et al 2016.
*TSL: Tier-based Strictly Local; pTSL: probabilistic TSL …
Interacting phonological processes are interesting for phonological theory because they often lead to opaque surface forms: forms where a process applied in an environment where it shouldn't have (overapplication) and forms where a process does not apply even though it looks like it should (underapplication). In my presentation I'll review …
Gary will give a talk titled "A Brief History of the Logic of Time: Some Paradoxes from A to Z" this Friday. His talk will begin with a puzzle inspired by Erich Friedman [1998]. You can click here to read the puzzle with a picture of triplets or check the …
Topic: Learning sequential functions with sub-symbolic properties
Details: It has been shown that morphophonological processes represented with subsequential functions can be learnt from finite samples of data (Jardine et al. 2014). We show that the size of such a sample can be significantly reduced if we change the representation of …
Topic: Identification in the limit from positive data with noise
Details: We consider the problem of identifying formal languages in the limit when the presentations of positive data are corrupted by noise. Two kinds of noise are considered separately: when the data systematically omits positive examples and when negative examples …
We often say that syntactic dependencies are in the computational classes SL and TSL over trees (Graf 2018 et seq). But what does that mean? Actually, there several ways to generalize the string classes to trees. One of these involves command strings (c-strings), which encode the c-commanders of some node …
Sequence-to-sequence (seq2seq) models are a class of neural network models that learn the mapping between given input and target sequences. These models take an encoder-decoder structure where the encoder reads the inputs and passes the encoded information to the decoder, and the decoder decodes the encoded information and writes outputs …
We will not be meeting this Friday, because of a Blitztalk that takes place 1:00-2:00 pm on the same day. The Blitztalk is part of the virtual open house for the next cohort of PhD students, which, as Thomas emphasizes, "is perhaps the most important event of the …
Piecewise Local Expressions (Lambert 2022) are a kind of regular expression language for describing certain kinds of formal languages easily using the notion of substring and subsequence containment. These will be reviewed in the beginning. Then we will spend time thinking how to develop a similar expression language (i.e …
Thomas will take a subregular look at syntactic grafts, a mechanism first proposed by Henk van Riemsdijk to explain constructions like "John went to I believe it was Chicago". Grafts have not seen wide adoption by syntacticians because they are considered too powerful. But Thomas will argue that they are …
In our first organizational meeting, Thomas discussed the schedule for the spring semester, 2023. Jeff will take over as the faculty organizer and Jack remains the student organizer for this semester.
Please note that, in the future meetings, we will be meeting at 1:30 pm on Fridays, instead of …