ATLASDiscovery Engine
Discoveries 0
Mitmachen Ask the Corpus Login
Corpus-based discovery across 14 years of research

A discovery layer over the combined cluster corpus.

ATLAS harvests every member's bibliography, retrieves the connected texts, and builds a searchable, analysable corpus. On top sits an exploratory layer — not a catalogue, but an instrument for examining the structural relations between hundreds of researchers across two clusters and their disciplines.

2
Clusters · BWG + MoA
14
Years of output
~600
Members & affiliates
12k+
Publications indexed*

From bibliography to discovery — the pipeline

01

Harvest

Pull every cited reference from all member bibliographies across both clusters.

● Ingested
02

Scrape

Your existing scraper fetches the connected full texts, abstracts & metadata.

● Connected
03

Index → RAG

Embed & chunk into a retrieval system; make it searchable and analysable.

◔ Building
04

Explore

Topics, bridges, discourses, dialogues — the exploratory discovery layer.

◔ This prototype

The Constellation

Every person, topic and discipline as a single network. Select a thread to trace its relations across the cluster.

Open the network →

Topics & Overlaps

Topic modelling over the corpus identifies the cluster's themes — and where two of them share literature and terminology.

See overlaps →

Bridge Actors

What connects research questions across disciplines? In the ID+ sense: not people, but themes, spaces, methods and tools — the non-human actors holding the network together.

Find bridges →

Discourses

Identify the discourse within each discipline, and trace where vocabularies and citations converge on a shared problem.

Trace discourses →

Disciplinary Dialogues

Personas grounded in different disciplines discuss a question, drawing on their field's stance. Observe, or contribute.

Open the dialogue →

Ask the Corpus

A semantic retrieval interface. Ask in natural language; receive an answer grounded in the corpus with citations.

Ask a question →

*Figures and contents in this prototype are illustrative mock data, shown to communicate the interface concepts.

Layers:
Mode:
Drag to pan · hover a node to trace its connections · click to inspect

⌖ Throw in a question or a topic

ATLAS finds matching literature and shows each source in context on the map →

Try something like “active matter in biology and design” or just a keyword. I’ll surface the most relevant sources from the corpus and light them up on the map.
Each dot is a publication, grouped by topic. Matches pulse in purple — click any dot to read it in context.

Topics & their overlaps

Topic modelling over the full corpus. Pick two topics to surface the literature and terminology they secretly share.

The ribbon diagram maps shared literature between topic fields — thicker ribbons mean more co-cited sources. Click a topic below to highlight its ties, then click a second to open the overlap.

Select two topics to reveal their shared sources & vocabulary…

Bridge actors

In the ID+ sense a bridge is not a person but something connecting — a theme, a space, a method, a tool, a material, a shared problem. ATLAS surfaces the non-human actors that hold otherwise-separate research questions together, ranked by how much of the network's flow passes through them.

Discourses & their intersections

Each discipline runs its own conversations. ATLAS names them, assigns publications, and — crucially — finds where two discourses converge on the same problem under different words.

Cast the panel

Each persona is grounded in one of the cluster’s research strands — image science, biomaterials physics, design research, history of science, and computational humanities (the fields of Bredekamp, Fratzl, Mareis, Schäffner and Stein). They voice their discipline’s stance, not any individual’s words. Pick 2–3 and let them argue.

Question:
Pick your panel and a question, then start the debate. You can jump in any time.
Retrieval over 14 years of citations

Ask the corpus anything.