Searcharxiv⌕ Search

arXiv subjects

Emily Upton

Publications and source records attributed to Emily Upton.

2 recordsLinked to original sources

Random Fibonacci Tilings

The Fibonacci spiral is formed by a square-joining rule: begin with a unit square and attach squares whose side lengths follow the Fibonacci sequence; in the standard construction, choices of direction produce a spiral and a tiling of the plane. We study a randomization of this construction, in which the attachment axis alternates deterministically, while fair coin flips determine direction. For a lattice point we ask when the growing rectangle first covers that point and how the expected covering time depends on its location. Tracking the rectangle's boundary reduces this question to random subset sums of even- and odd-indexed Fibonacci numbers. We extend the construction to d dimensions, where coordinate subsequences are super-increasing, allowing us to derive exact covering probabilities and formulas for the expectation and variance. For $d\geq2$, the expected covering time grows logarithmically with distance from the initial block, while the variance remains uniformly bounded. We obtain exponential tail bounds for the delay beyond the earliest feasible covering stage. Finally, we extend the Fibonacci spiral to the randomized tiling through a square-by-square construction using circular arcs and quintic connectors. This construction uses one-step lookahead, admits a continuously differentiable parameterization, and recovers the Fibonacci spiral for the standard tiling.

math.PR↗

NUVA: A Naming Utterance Verifier for Aphasia Treatment

Anomia (word-finding difficulties) is the hallmark of aphasia, an acquired language disorder most commonly caused by stroke. Assessment of speech performance using picture naming tasks is a key method for both diagnosis and monitoring of responses to treatment interventions by people with aphasia (PWA). Currently, this assessment is conducted manually by speech and language therapists (SLT). Surprisingly, despite advancements in automatic speech recognition (ASR) and artificial intelligence with technologies like deep learning, research on developing automated systems for this task has been scarce. Here we present NUVA, an utterance verification system incorporating a deep learning element that classifies 'correct' versus' incorrect' naming attempts from aphasic stroke patients. When tested on eight native British-English speaking PWA the system's performance accuracy ranged between 83.6% to 93.6%, with a 10-fold cross-validation mean of 89.5%. This performance was not only significantly better than a baseline created for this study using one of the leading commercially available ASRs (Google speech-to-text service) but also comparable in some instances with two independent SLT ratings for the same dataset.

cs.CL↗