How to explore: drag to rotate · scroll to zoom · hover for a quick label · click for full context · toggle the detail panel with the button ↗ · sky-blue dots pulse along thematic relay arcs (motion off if your system prefers reduced motion)
The dissertation tests whether Soviet-era talking points on famine commemoration transform predictably when they cross register boundaries - secret memo → embassy/Novosti → domestic news → diaspora/scholarly print - and whether that transformation is dated and measurable (Montreal 1983 as anchor, post-1983 explosion as qualitative rewire). These 3D models make Lotman’s spatial vocabulary falsifiable: they read from the same open JSON as Figures 1-12.
Imagine each ingested document as a coloured dot inside nested gold spheres. The spheres are not decoration - they are Lotman’s semiosphere made spatial: inner shell = secret KGB/CC voice; middle shell = autocommunication (inward reporting); third shell = Soviet outward channels (embassy, Novosti, UN, domestic news register); outer shell = diaspora, mainstream, and scholarly rival print. Coloured tubes between dots are traced relays - pair-read connections we can defend with a proof tier in JSON.
Distance from the centre encodes institutional register, not geography. Montreal is the anchor event: spr. 1206 (24 Mar 1983) sits on the inner shell; embassy/Novosti denial sit outward; diaspora and AP commemoration sit farther out. The faint red plane marks the 1983 boundary where internal scare quotes and «сборище» do not copy outward - boundary non-equivalence made visible. A dot with no outward arc (e.g. Pravda 1983) can still be evidence: domestic silence while denial runs abroad.
In this visualization, a connection is not a vague “link.” It is a pair-read adjudication stored in JSON: two text surfaces share a topic (famine, symposium, gathering) and we document how wording travels - or refuses to travel - between them. An arc therefore encodes an evidentiary claim with a proof tier (proven thematic relay, proven verbatim, memo-claim only, and so on).
What connections tell us collectively: (1) Internal hostility (сборище, scare quotes) stays on the inner shell and does not relay to embassy, Novosti, or UN - 0 scare-quote operator hits on those Soviet-outward surfaces in the register-relay matrix (48 internal hits across three memos). (2) Outward Soviet lines share denial frames (drought, kulak sabotage, no artificial famine) across embassy and agency. (3) Rival and diaspora surfaces carry commemorative language the outward line rejects. (4) Domestic Soviet newsprint in 1983 is largely silent on Holodomor denial - shown as absence of arcs on the newspaper shell; tier-B tracks may show incidental quote-mark matches (1933 baseline idiom), not denial relay. Together, arcs sketch a governed split narrative, not one story told louder or softer.
What a missing arc also means: No line between spr. 1206 and Pravda is itself evidence - denial routed abroad, not through verified domestic clips in the anniversary window. The model is honest about silence and pending tiers.
Try it: click spr. 1206, then the sky-blue arc to the embassy - read the panel sections “What this connection tells us” and “What it does not prove.” Compare to the green embassy↔Novosti arc (outward coordination). Notice Pravda has no denial arc - silence as finding. Ideal full semiosphere (fabricated demo) →
Each panel opens one step on the way from the archive's JSON to the picture, with the equation it uses.
Each of the 13 documents gets a shell from its register, the kind of reader it was written for. The shell radius rs is 1.4 for the secret memos, 2.1 for inward reporting (empty in this pilot), 2.9 for Soviet texts aimed outward and the Soviet domestic newspapers, and 3.7 for rival, diaspora and mainstream print. The order is what matters; the gaps between shells are a display choice.
Within a shell, documents are spread out with golden-angle spacing so they don't collide. N = 13 and i is the document's place in the list. Only the distance from the centre carries meaning. The direction around the sphere is a spacing device. The 0.85 squashes the sphere vertically, so documents near the top and bottom sit a little inside their shell. Read their shell from the colour and the label.
Each arc joins two documents A and B along a smooth curve that bows outward, because its control point M is the midpoint pushed 12% away from the centre. An arc is drawn only when the register-relay JSON holds a pair-read for that pair, and its colour is the relay type. This pilot has five: two proven thematic relays (spr. 1206 to the embassy release and to Novosti), one verbatim relay (embassy and Novosti carrying the same text), and two memo-claim links from spr. 1249 to the Globe text it cites.
O(o, s) is how many times the operator detectors matched wording habit o on document s. The same sum over the embassy release, Novosti and the UN statement is 0. Pravda, Izvestiia and Radianska Ukraina show 1, 2 and 3 matches, which are stray quotation marks in tier-B material from the 1933 idiom rather than relay. Figure L2 draws every count.
Play the relay story walks the view through 1983. The camera keeps its angle, moves its focus to spr. 1206, the embassy release, Novosti and Pravda in turn, and zooms in on each, while an amber halo marks the document and the caption under the figure says what happened there. It takes about half a minute and ends back where it started. The white dots with a blue glow travel along the two thematic relay arcs. Each dot's position along its arc is (0.00012 × time in milliseconds + an offset) mod 1, so a pass takes a little over eight seconds. They run on their own unless your system is set to reduce motion; Play relays and Pause relays switch them either way. Turn the model spins the spheres slowly. Dragging still works throughout.
The page is generated in the thesis lab by build_lotman_3d_visualizations.py from the register-relay JSON and copied here by scripts/sync_from_thesis_lab.py. The data object at the bottom of the page holds every document, operator count and arc with its tier. This section is kept in explain/fig-lotman-semiosphere-3d.html in the site repo and added back on every sync.
The secret memo of 24 March 1983 uses hostile wording: scare quotes, and a dismissive word for the symposium (a "gathering"). That wording never appears in the embassy or Novosti releases about five weeks later, but the topic does, now framed as drought and kulak sabotage. The embassy and Novosti carry the same text. Pravda and the other domestic papers have no denial arc at all: silence at home while the denial ran abroad.
This is the dissertation's central picture. Talking points change in predictable ways when they cross from one audience to another, and the change can be dated and counted. In Lotman's terms the boundary works as a filter: it translates some things, recodes others and refuses the rest. The semiosphere is the whole space those documents live in, and the spheres here are one way of drawing it so that its claims can turn out wrong.
The same question runs through the AI experiments on this site. When one model hands text to another, is there a boundary that filters like this? Mostly there isn't, unless someone builds one into the channel.
This is the core of the dissertation and the strongest kind of evidence on the site, because every arc rests on reading two documents side by side. Its limit is size: thirteen surfaces and five arcs, an empty inward shell, and a Globe 1987 link that rests on a memo's own claim. Here more data is the whole answer, and the open threads say what kind.