🔬 Impresso Datalab Notebooks

A Toponym Resolution Pipeline for Digitised Historical Newspapers

This repository provides underlying code and materials for the paper 'Station to Station: Linking and Enriching Historical British Railway Data'.

An experimental infrastructure for computational critique of historical writing using RAG architectures, argumentation analysis, rhetorical mapping, and large language models.

Teaching materials for the workshop "Network Analysis in the Humanities" organised by DigiTS (Center for Digital Text Scholarship) at the University of Tartu.

Official repository of WikiTextGraph

Training classifier models on acoustic metadata to predict genres and subgenres.

Training classifier models to predict genres and subgenres on album cover data.

Word–color association from large-scale online image data (CIELch). Code for the PLOS ONE manuscript.

Replication package for PHTS Theory v3.0 — a two-tier structural framework for the Voynich Manuscript. Submitted to Cryptologia, 2026. Contains Python scripts, data files, and pre-computed results.

Impresso Python Library to interact with the Impresso Public API

A machine-checked formalization of the Fuxi 64-hexagram system. 40 claims verified; 7 corrected, including a clustering coefficient that is 0 not 5/12, and a divination kernel whose stationary distribution is not uniform.

Computational vocabulary analysis across Buddhist Vajrayana, Shakta Tantra and Baul Bengali texts · TF-IDF char n-grams · cosine similarity · Tara to Kali lexical migration · Ramprasad Sen OCR · Charyapada · 75 texts across 8 tradition layers

Chunk-level FAISS + Solar-10.7B RAG pipeline over 2,900+ Korean flash fiction texts.

5D semantic-dynamics framework separating Korean prose poetry from flash fiction. Code for the Scientific Reports manuscript.

Rights-aware archive index and research framework for modern graphic design history.

Information dynamics in Korean flash fiction — surprisal, coherence, and semantic-shift trajectories. Code for the Physica A manuscript.

A source-grounded, falsifiable atlas of humanity's shared story-patterns: 338 public-domain sacred texts, 106 traditions, 52k quote-anchored motif records, preregistered cross-cultural universality tests.