The Session Papers can be searched four ways. Each answers a different kind of question; together they make a research corpus out of a scattered archive.
Full-text search
Search queries the machine transcription of every processed page in the collection. Results return the document, its case context, and the matching passage, so you can judge relevance before opening anything. Because the index reaches the sentence level, the search rewards specificity: the name of a ship, a commodity, a farm, a legal term of art. Quoted phrases match exactly; unquoted terms match broadly.
Two things are worth knowing about what you are searching. First, the text is machine transcription — produced by vision-language models reading the printed pages, accurate enough for discovery but not a substitute for the page image, which is always one click away. We label it as uncorrected wherever it appears. Second, coverage follows processing: the printed papers are transcribed at scale, while handwritten materials follow a separate, slower path. A null result means the term is absent from the processed corpus — not necessarily from the archive.
Spelling in the papers is eighteenth-century spelling. The court's clerks wrote Falkirke, tobaccoe, and persewer; if a modern spelling finds nothing, try period variants — or move to semantic search below, which is largely indifferent to orthography.
The map
The map explorer plots the places the papers mention. Behind it stands a gazetteer of Scotland's historic geography: place references extracted from the documents are resolved to coordinates and, where possible, to the historic counties, parishes, and registration districts they belonged to — boundary data drawn from the UK Data Service's historic boundary collections — and displayed over National Library of Scotland historic map tiles, so the geography you see is the geography the litigants knew.
The court's world was not confined to Scotland, and neither is the map. Session Papers litigated Atlantic commerce, colonial inheritance, and imperial careers; the map plots Jamaica, Virginia, Bengal, and Boston where they belong. Zoom out and the collection's British-Atlantic dimension is immediately visible; zoom in and parish-level disputes take shape. Filters distinguish the roles places play in a case — where the dispute arose, where parties resided, where courts sat — because a place mentioned is not always a place at issue.
Roughly 70% of place references are currently mappable. The remainder are ambiguous (Scotland has many Newtons and more than one Aberdour), historical names still being identified, or places awaiting curation — the number grows steadily. Every mapped place links to its own page, with the cases and documents connected to it.
The network
The network explorer shows which people, places, subjects, and cases occur together in the documents. Choose any entity — a Glasgow merchant, the subject of sequestration, the parish of Lintrathen — and the explorer draws its neighbourhood: everything that co-occurs with it, with connections weighted by how many documents they share. Click a neighbour to re-centre and follow the chain: merchant to estate, estate to dispute, dispute to the advocate who argued it.
Co-occurrence is evidence of association, not proof of relationship — two people may share a document as opposing parties, joint petitioners, or mere neighbours in a rental roll. The network is a discovery instrument: it surfaces connections worth investigating and takes you to the documents where the real story is told. The legend doubles as a filter — toggle entity types to focus the graph.
Semantic search — the machine-readable archive (MCP)
Keyword search finds what a document says; semantic search finds what a document is about. The archive computes multilingual open-weight text embeddings (BGE-M3) over the full corpus, which lets software ask for documents by meaning: papers about coastal land disputes, about the wine trade, about contested guardianship — regardless of the words or spellings a clerk chose in 1774.
This capability is offered through a Model Context Protocol (MCP) endpoint. MCP is an open standard that lets research tools and AI assistants connect to external collections; once connected, your software can query the archive semantically, retrieve matching documents with their stable identifiers, and cite what it finds. Corpus-scale questions — trends across decades, clusters of similar litigation, documents that resist keyword description — become tractable without downloading or scraping anything.
Access is by application. Credentials are issued to researchers on request: contact the project team with a short description of your research and the tools you plan to connect. There is no charge. We ask that published work cite the archive, and that automated use respect the endpoint's rate guidance — it shares infrastructure with the reading interface.
Citing what you find
Every document, case, person, and place in the archive has a stable address, and document pages carry a ready-made citation. Beneath the visible URLs sits a system of permanent identifiers that resolve through the archive's own resolver, so a reference made today will survive future changes to the site's structure. Cite the document page; the archive will keep the address good.
