Explore - The Laßberg Letters Project

Quick search across all letters, persons, places and literature (keyword-based)

Explore the whole corpus, built from the register metadata of all 3,268 letters.

Correspondent:

Letters per year

View as table

Correspondents

Places

Blue: where letters were sent from. Purple (toggle in the layer control): places the letters talk about.

Historical persons discussed

Historical figures that the letters talk about: the subjects of Laßberg's antiquarian research, counted across the full-text letters.

Contemporary persons mentioned

Living contemporaries the letters mention: the social fabric of the scholarly network beyond the correspondents themselves.

Literature discussed

Works including medieval texts and contemporary scholarship that the letters discuss.

Searches by meaning, using BGE-M3 embeddings of the transcribed full-text letters. Letters that are only recorded in the register (no transcription yet) are not covered. The first search loads the ONNX model (~560 MB, cached afterwards).

You can also use the Hugging Face Inference API instead of downloading the model

HF token (stored locally only, needed only for the API variant):

The knowledge graph of the edition: every letter, person, place, work, and manuscript, connected by who wrote to whom and what is mentioned where. Search for a name or letter and explore its network.

SPARQL console (expert)

Queries run against an RDF file.

Experimental: LLM-orchestrated GraphRAG Research chat

Here, research is driven by an LLM: it formulates semantic queries, writes SPARQL against the edition RDF, traverses the knowledge graph, reads letters, and synthesizes a cited answer. Every step is shown in the output documenting how the answer was produced. Your API key is stored in your browser only and sent only to the provider you select. (SAIA API currently not working from Browser.) For local use, see MCP server on GitHub.

Provider API key Model Endpoint (base URL)

Semantic search inside the chat uses the same BGE-M3 model as the Search tab (one-time ~560 MB download on first use, or the HF-API option set there).

Combines embedding search with graph expansion (shared mentions, correspondence context) into a transparent, citable research context. Retrieval-only: no LLM is called and no API key is needed.