4 March 2021 in IEEE Access
We have implemented a machine translation system, the PolyMath Translator, for LaTeX documents containing mathematical text. The current implementation translates English LaTeX to French LaTeX, attaining a BLEU score of 53.6 on a held-out test corpus of mathematical sentences. It produces LaTeX documents that can be compiled to PDF without further editing. The system first converts the body of an input LaTeX document into English sentences containing math tokens, using the pandoc universal document converter to parse LaTeX input. We have trained a Transformer-based translator model, using OpenNMT, on a combined corpus containing a small proportion of domain-specific sentences. Our full system uses this Transformer model and also Google Translate with a custom glossary, the latter being used as a backup to better handle linguistic features that do not appear in our training dataset. Google Translate is used when the Transformer model does not have confidence in its translation, as determined by a high perplexity score. Ablation testing demonstrates that the tokenization of symbolic expressions is essential to the high quality of translations produced by our system. We have published our test corpus of mathematical text. The PolyMath Translator is available as a web service at http://www.polymathtrans.ai.
16 January 2020 in bioRxiv
TIThe ONDRI investigators
+15
Large scale research studies combining magnetic resonance imaging data generated at multiple sites on multiple vendor platforms are becoming more commonplace. The Ontario Neurodegenerative Disease Research Initiative (ONDRI -http://ondri.ca/), a project funded by the Ontario Brain Institute (OBI), is a recently established province-wide natural history study, which has recruited more than 500 participants from neurodegenerative disease groups including amyotrophic lateral sclerosis, fronto-temporal dementia, Parkinson’s disease, Alzheimer’s disease, mild cognitive impairment, and cerebrovascular disease (previously referred to as the vascular cognitive impairment cohort). Because of its multi-site nature, all captured data must be standardized and meet minimum quality standards to reduce variability. The goal of the ONDRI imaging platform is to maximize data quality by implementing vendor-specific harmonized MR imaging protocols (consistent with the Canadi-an Dementia Imaging Protocol -http://www.cdip-pcid.ca/), monitoring protocol adherence, qualitatively assessing image quality, measuring signal-to-noise and contrast-to-noise, monitoring system stability, and applying corrections based on the analysis of images from two different phantoms regularly acquired at each site. To maximize image quality, this work describes the use of various automatic pipelines and manual assessment steps, integrated within an established informatics and databasing platform, the Stroke Patient Recovery Research Database (SPReD) built on the Extensible Neuroimaging Archive Toolkit (XNAT), and contained within the Brain-CODE (Centre for Ontario Data Exploration) framework. The purpose of the current paper is to describe the steps undertaken by ONDRI to achieve this high standard of data integrity. Data have been successfully collected for the past 4 years with the pipelines and assessments identifying deviations, allowing for timely interventions and assessment of image quality.