TextileBase

A Semantic Knowledge Graph for Textile and Dress History

TextileBase is a knowledge graph that interconnects databases and secondary sources—such as artefact photographs, contemporary images, drawings, and written descriptions—related to textile and dress history. It supports diverse areas of textile research, from historical dress studies to sustainable fashion.

Technically speaking, building an interactive website for textile history is like setting up a webshop—except all the product information is either missing or fragmented. There are no barcodes, producers, or standard descriptions. Instead, the relevant information is buried in academic knowledge, secondary sources, or scattered fragments of primary data.

This is where AI can significantly assist: by gathering, linking, and inferring data at a speed and scale beyond human capacity.

Connecting Surviving Artefacts

Unlike metal objects or buildings, textiles are fragile and rarely survive intact. They are worn, used, and discarded. Even from the 19th century, very few textiles remain.

Our goal is to create a tool that connects fragmented museum inventories and catalogues across institutions—linking rare textile artefacts in a coherent, searchable network.

The Challenges of Linked Open Data in Textile History

While digital humanities frequently refer to “linked open data” or the “Internet of Things,” textile history is still at the early stages of adopting these frameworks. A key goal of TextileBase is to give each artefact—like the Seto shirt—a unique URI. But generating URIs for garments in small rural museums or private collections is often time-consuming and expensive.

TextileBase seeks to streamline this process—making it faster, cheaper, and more reliable.

The Importance of Secondary Sources

Historical collecting practices were often biased. Textiles worn by everyday people—peasants, laborers, minority groups—were rarely preserved. Instead, most surviving collections reflect the lives of elites or were gathered during ethnographic expeditions with colonial or exoticizing perspectives.

That’s why secondary sources—such as period photographs, illustrations, and books—are so valuable. They often unintentionally document everyday garments, offering insights missed by early collectors.

Case Study: Finno-Ugric Traditional Clothing

We focus on the clothing traditions of small Finno-Ugric communities—such as the Setos and Livonians—due to the technical and linguistic challenges involved.

For example, Livonian communities in present-day Latvia were largely assimilated by the late 19th century. Setos are now found in southeastern Estonia and Russia’s Pskov oblast. Their artefacts are scattered across Finnish, Estonian, and Hungarian archives, usually documented in those languages—rarely in Livonian, Seto, or even English.

We must build a historical namespace of place names, languages, even spelling history, and garment classifications to search and understand this data. This includes terms like “shirt” in Finnish (paita), Hungarian (ing), or Estonian, and accounts for changing styles, materials, and local terminologies. This is one of the uses of TextileBase: it contains knowledge about words, placenames, their morphology, so that researchers not familiar with the collection’s language (and its historical changes, abbreviations) can work globally.

Building a Wikimuseum for Dispersed Collections

Inspired by Wikimedia Estonia’s multi-language, open-access model, we propose a virtual museum—a Wikimuseum—that brings together:

• Artefacts from rural museums (e.g., Mõniste, Saatse, Värska)

• Items in national museums (Estonia, Finland, Hungary)

• Private collections that would never be physically exhibited together

This approach connects dispersed knowledge—across time, languages, and borders—into a shared digital space.

How AI Supports Textile Research

TextileBase incorporates AI-driven tools—language models, translation engines, inference systems—that can:

• Search for historical placenames across multilingual archives

• Identify garments in photos or scanned books

• Detect and match synonyms or spelling variants

• Flag potentially relevant images and documents for expert review

By operating within the TextileBase framework, our AI tools remain human-controlled and explainable, reducing the risk of hallucinated or misleading results.