Software & Data

Research software, datasets, and trained models, all openly licensed. Everything below is on GitHub.

Platforms

DALME Online Database: The platform behind dalme.org. A digital environment for extracting, transcribing, analyzing, and publishing material culture information from textual primary sources, serving an international research consortium since 2020. Django, Django REST Framework, and Wagtail, with Elasticsearch full-text search and asynchronous task processing, deployed on AWS. Recipient of the Medieval Academy of America's Digital Humanities and Multimedia Prize, 2022.

Packages

CoNLL-U Tools: Python package, available on PyPI. A toolkit for CoNLL-U files, Universal Dependencies treebanks, and annotated corpora: bidirectional format conversion, validation against annotation guidelines, parser scoring against gold standards, linguistic pattern matching, and morphological feature normalization across four Latin treebank tagsets. Documentation.

Tree-sitter Chunk Grammar: A Tree-sitter parser for NLTK chunking-grammar syntax, with an accompanying editor package for syntax highlighting.

Research code and data

correction-as-annotation: Corpus, code, and trained models accompanying Correction as Annotation: Bootstrapping a Dependency Parser for Documentary Medieval Latin. Includes the DALME-Marseille corpus (1,804 hand-corrected sentences plus a 200-sentence gold standard, in CoNLL-U), the full analysis package, and nine trained Stanza pipelines. Every table and figure in the paper regenerates from the released data. Models and data: 10.5281/zenodo.22048602

rules-based-entity-extraction: A pipeline for extracting unnamed entities from Medieval Latin, combining rule-based lexical resources with a machine learning chunker trained on custom linguistic features.

course-scheduling-optimization: Course scheduling under institutional constraints, modelled as a generalized assignment problem and solved as a mixed integer program, with consensus ranking of faculty preferences.

baudot-murray-CCIR476-demo: Decoding Baudot-Murray/CCIR 476 teleprinter tape from microfilm images using computer vision.

marseille-luxury-objects: Data-driven analysis of material culture, wealth distribution, and social structure in medieval Marseille, on a DALME-derived dataset.

Vocabularies

Tropy-Vocabularies: CIDOC-CRM, Erlangen CRM, BIBO, schema.org, DINAA, and Open Context vocabularies converted and adapted for use with Tropy.

Teaching materials

Openly available workshop materials: Introduction to Python | Working with Data | Web Scraping | Understanding AI