Some Like It Small: Czech Semantic Embedding Models for Industry Applications

This article focuses on the development and evaluation of Small-sized Czech sentence embedding models. Small models are important components for real-time industry applications in resource-constrained environments. Given the limited availability of labeled Czech data, alternative approaches, including pre-training, knowledge distillation, and unsupervised contrastive fine-tuning, are investigated. Comprehensive intrinsic and extrinsic analyses are conducted, showcasing the competitive performance of our models compared to significantly larger counterparts, with approximately 8 times smaller size and 5 times faster speed than conventional Base-sized models. To promote cooperation and reproducibility, both the models and the evaluation pipeline are made publicly accessible. Ultimately, this article presents practical applications of the developed sentence embedding models in Seznam.cz, the Czech search engine. These models have effectively replaced previous counterparts, enhancing the overall search experience for instance, in organic search, featured snippets, and image search. This transition has yielded improved performance.

Evaluation of image registration for measuring deformation fields in soft tissue mechanics

High-fidelity biomechanical models usually involve the mechanical characterization of biological tissues using experimental methods based on optical measurements. In most experiments, strains are evaluated based on displacements of a few markers and represents an average within the region of interest (ROI). Full-field measurements may improve description of non-homogeneous materials such as soft tissues. The approach based on non-rigid Image Registration is proposed and compared with standard Digital Image Correlation (DIC) on a set of samples, including (i) complex heterogeneous deformations with sub-pixel displacement, (ii) a typical uniaxial tension test of aorta, and (iii) an indentation test on skin. The possibility to extend the ROI to the whole sample and the exploitation of a natural tissue pattern represents the main assets of the proposed method whereas the results show similar accuracy as standard DIC when analysing sub-pixel deformations. Therefore, displacement and strain fields measurement based on Image Registration is very promising to characterize heterogeneous specimens with irregular shapes and/or small dimensions, which are typical features of soft biological tissues.