arXiv subjects

Deyan Ginev

Publications and source records attributed to Deyan Ginev.

6 recordsLinked to original sources

arXiv2026

Scaling Accessible Mathematics on arXiv: HTML Conversion and MathML 4

We report on the ongoing development of arXiv's HTML Papers offering, available on every new TeX/LaTeX submission since its initial release in 2023. The main highlights from 2025 and early 2026 are: (i) community-driven improvements to HTML fidelity and service health, with roughly half of 6,000 user reports resolved; (ii) corpus-scale conversion work aimed at 90% error-free HTML (currently 75%); (iii) initial MathML 4 Intent annotations for accessible speech output; (iv) an in-progress Rust port of LaTeXML, reducing compute costs and enabling faster previews on submission. The arXiv HTML Papers project remains experimental, but is gradually maturing as we better understand the needs of arXiv's readers and the technical opportunities presented by new standards and by advances in programming languages and AI.

cs.CL↗

arXiv2019

Scientific Statement Classification over arXiv.org

We introduce a new classification task for scientific statements and release a large-scale dataset for supervised learning. Our resource is derived from a machine-readable representation of the arXiv.org collection of preprint articles. We explore fifty author-annotated categories and empirically motivate a task design of grouping 10.5 million annotated paragraphs into thirteen classes. We demonstrate that the task setup aligns with known success rates from the state of the art, peaking at a 0.91 F1-score via a BiLSTM encoder-decoder model. Additionally, we introduce a lexeme serialization for mathematical formulas, and observe that context-aware models could improve when also trained on the symbolic modality. Finally, we discuss the limitations of both data and task design, and outline potential directions towards increasingly complex models of scientific discourse, beyond isolated statements.

cs.CL↗

arXiv2014

E-books and Graphics with LaTeXML

Marked by the highlights of native generation of EPUB E-books and TikZ support for creating SVG images, we present an annual report of LaTeXML development in 2013. LaTeXML provides a reimplementation of the $\TeX$ parser, geared towards preserving macro semantics; it supports an array of output formats, notably HTML5, EPUB, XHTML and its own $\LaTeX$-near XML. Other highlights include enhancing performance when used inside high-throughput build-systems, via incorporating a native ZIP archive workflow, as well as a simplified installation procedure that now allows to deploy LaTeXML as a cloud service. To this end, we also introduce an official plugin-based scheme for publishing new features that go beyond the core scope of LaTeXML, such as web services or unconventional post-processors. The software suite has now migrated to GitHub and we welcome forks and patches from the wider FLOSS community.

cs.DL↗

arXiv2014

NNexus Reloaded

Interlinking knowledge is one of the cornerstones of online collaboration. While wiki systems typically rely on links supplied by authors, in the early 2000s the mathematics encyclopedia at PlanetMath.org introduced a feature that provides automatic linking for previously defined concepts. The NNexus software suite was developed to support the necessary subtasks of concept indexing, concept discovery and link-annotation. In this paper, we describe our recent reimplementation and revisioning of the NNexus system.

cs.DL↗

arXiv2014

LaTeXML 2012 - A Year of LaTeXML

LaTeXML, a $\TeX$ to XML converter, is being used in a wide range of MKM applications. In this paper, we present a progress report for the 2012 calendar year. Noteworthy enhancements include: increased coverage such as Wikipedia syntax; enhanced capabilities such as embeddable JavaScript and CSS resources and RDFa support; a web service for remote processing via web-sockets; along with general accuracy and reliability improvements.

cs.DL↗

arXiv2011

The Planetary System: Executable Science, Technology, Engineering and Math Papers

Executable scientific papers contain not just layouted text for reading. They contain, or link to, machine-comprehensible representations of the scientific findings or experiments they describe. Client-side players can thus enable readers to "check, manipulate and explore the result space". We have realized executable papers in the STEM domain with the Planetary system. Semantic annotations associate the papers with a content commons holding the background ontology, the annotations are exposed as Linked Data, and a frontend player application hooks modular interactive services into the semantic annotations.

cs.DL↗