Browsing by Author "Villa, Andres"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Augmenting BERT-style Models with Predictive Coding to Improve Discourse-level Representations
    (ASSOC COMPUTATIONAL LINGUISTICS-ACL, 2021) Araujo Vasquez, Vladimir Giovanny; Villa, Andres; Mendoza Rocha, Marcelo Gabriel; Moens, Marie-Francine; Soto, Alvaro
    Current language models are usually trained using a self-supervised scheme, where the main focus is learning representations at the word or sentence level. However, there has been limited progress in generating useful discourse-level representations. In this work, we propose to use ideas from predictive coding theory to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations. As a result, our proposed approach is able to predict future sentences using explicit top-down connections that operate at the intermediate layers of the network. By experimenting with benchmarks designed to evaluate discourse-related knowledge using pre-trained sentence representations, we demonstrate that our approach improves performance in 6 out of 11 tasks by excelling in discourse relationship detection.
  • No Thumbnail Available
    Item
    Training Program for Orthopedic Residents in Forefoot Osteotomy Skills Transference From a Simulator to a Cadaveric Surgical Scenario
    (2023) Ledermann, Gerardo; Kuroiwa, Aron; Gonzalez, Nicolas; Silva, Isadora; Villa, Andres
    IntroductionAn effective simulation program allows both the acquisition of surgical skills on the simulated model and the transfer of these skills to a surgical scenario. We designed a forefoot osteotomy training program and sought to determine the transferability to a cadaveric surgical scenario.MethodsEleven orthopedic residents and 2 foot and ankle surgeons were included. A foot simulator was used. All residents were instructed on the surgical techniques of Chevron, Akin, and triple Weil osteotomies. Eight junior residents (trainees) were enrolled in a supervised simulation program. Baseline assessment was performed on the simulator with the Objective Structured Assessment of Technical Skills (OSATS) and the Imperial College Surgical Assessment Device (ICSAD). After baseline, trainees completed a training program and had a final evaluation of proficiency on the simulator and on cadaveric specimens. Three senior residents with no simulated training (controls) and experts were assessed for comparison.ResultsAll trainees improved from a baseline OSATS score of 11 points (9-20) to a final score of 35 points (33-35) in the simulator and 34 points (32-34) in the cadaveric specimen (P < 0.01). Compared with baseline, the ICSAD results improved in path length (391 [205-544] to 131 [73-278] meters, P < 0.01) and number of movements (2756 [1258-3338] to 992 [478-1908], P < 0.01). The final OSATS and ICSAD scores did not differ from experts (P = 0.1) and were significantly different from untrained residents (P = 0.02).ConclusionsSimulated training of Chevron, Akin, and triple Weil osteotomies in orthopedic residents improved procedural proficiency, enabling successful skill transfer to a surgical scenario in cadavers.