samhan.ai

Relinkings ALPHA

V0.2 安東權氏成化譜 1476 — PEOPLE — RELATIONS — GENERATIONS ENGINE — LOADING A research environment for discovering marital relinking chains and loops in medieval Korean genealogical data

Start with a person. Type a name, a hangul reading, or a biographical number, and the man is drawn with the kin the 1476 print records for him: his father above, his sons below, his sons-in-law beside them. Daughters are not written in this genealogy, so a marriage appears as a line from the bride's father to her husband — the woman herself is that line. Click any man in the drawing to open his record; double-click him to draw his own kin beside him, next to what is already drawn, and hold Alt while clicking to take him off the drawing again. None of this alters the record: the drawing is a view of the genealogy, never an edit of it. The depth boxes say how many generations up and down to take at once. Below, the lineage index lists every choronym with the number of men it holds, and the folio browser goes the other way: name a page of the print and see who stands on it.

Who do you want to look at?

Leave these people out

Generations up ↑

Generations down ↓

How much to keep

Every choronym in the corpus and how many men carry it. Click one to draw its patriline — the father-to-son tree within that lineage alone, largest connected part first.

Filter

At most this many men

A folio citation in this corpus is a volume, a leaf and a side of the printed page. Give one and see who the print puts there; leave the leaf empty to see a whole volume.

Volume

Leaf

Side

A relinking is what happens when two families already joined by a marriage are joined by another. Marry your daughter into a house your grandfather already married into, and the two lines close a circle; that circle is what this part looks for. There are two ways of finding one, and they answer slightly different questions. A chain starts from a man with more than one line of issue and follows two of them down until they meet again in a single descendant — the classic query of Douglas White's work on marital relinking. A loop forgets the direction of descent and looks for any closed circuit in the web of fathers and fathers-in-law, whichever way round it runs; it is the exhaustive census, and it finds circles no single ancestor commands. Pick a tab, say how long a path you will allow, and run it. Each result is written out as a sentence. Click one to draw it — alone, or with as much of its surroundings as you ask for.

Solid lines are father to son; the dashed line is a marriage, drawn from the bride's father to her husband. On the left, two lines of descent leave one man and rejoin in another: a chain. On the right, the same closure with no single man at its head: a loop. Both are relinkings.

Two separate lines of descent leaving one man and arriving at one man. Where the two lines share nobody in between they close a circuit.

Shortest path allowed

Longest path allowed

Lineage at the top

Lineage at the foot

At least one marriage in the path ? Only pairs joined twice over (circuits)

Stop after

Draw each result on its own

focuscontext

Beyond nought, the circle keeps its ink and everything within that many steps of it is drawn faded around it.

Results

Not yet run.

The tools above look for particular circles. These look at the corpus as one object and ask what shape it has. The endogamy core is the largest group of men within which every pair is joined by at least two independent paths of kinship: take any one man out and the rest stay joined. White and Jorion called that condition structural endogamy — the part of a population that keeps marrying back into itself. The marriage matrix reduces the same data to a table of who gives daughters to whom, lineage by lineage, so a repeated alliance shows as a heavy cell. The block model then asks whether lineages group better by whom they marry than by what they are called. And the null model asks the hardest question of the four: whether a pattern is a rule or an accident. It shuffles the recorded marriages at random — keeping every man's number of wives, every father's number of daughters given, and never crossing a generation — and counts the circles again. If the real genealogy holds far more of a kind of circle than the shuffles do, the preference was real.

Treating descent and marriage alike as ties without direction, the largest bicomponent is the structurally endogamous core (White & Jorion 1992).

Draw at most

Inside the core

Is the core one dense mass or a chain of overlapping neighbourhoods? The split tree divides it in two again and again by whom its members are tied to; the k-cores peel it from the outside in.

Split depth

Cap at

One row and one column per lineage. The cell where a row meets a column counts the marriages in which the row's lineage gave a daughter to the column's. A dot marks two or more — a repeated alliance. Click a cell for the marriages behind it.

How many lineages

CONCOR splits the lineages by how alike their marriage profiles are, until k groups remain. The image matrix then marks which pairs of groups are joined densely enough to count as a tie at all.

What to group

How many groups

Ignore lineages under

First the count: every circuit in the corpus tallied by side, by kin type and by lineage pair. Then the test: the same count over marriages shuffled at random, so the two can be compared.

Longest circle counted

Against chance

Shuffles

Seed

Everything above treats the genealogy as if it were true. This part does not. The apparatus lists the places where the 1476 print — or our reading of it — does not hold together: a father and son with different surnames, a line of descent that leaves the Andong Kwŏn, a man with two fathers (usually an adoption, 繼後), a person who exists only because some other man's entry named him, a lineage attributed by inference rather than stated, a generation number we had to raise to keep the arithmetic in order, a note of a second or a former husband. Each item carries its folio so it can be checked against the page. Nothing here alters the graph: it is a reader's apparatus, not a correction, and what any of it means is your decision. This part also holds the export tools and the glossary of every technical word the application uses.

Show

Nothing drawn yet.

The query travels in the link itself and is restored when the link is opened. Nothing is sent anywhere.

Save this query as

Every word in this application that means something particular, with what it means here. Hovering a dotted word anywhere in the sidebar shows the same definition.

Look up a word

Layout

Colours

Colour descent by patriline

With patrilines on, each lineage takes its own ink and dash, and the legend becomes editable — rename a line and the name sticks. A marriage is never a patriline.

Names on the figures Hangul on a second line Office on a further line Birth order on the lines Generation grid

Figure size 15

Line width 2

Line opacity 78

Name size 15

Names are never drawn below 11 px on screen. The A4 export enlarges them again so they survive the page.

Legend

HAS_SON 子father → son

HAS_SIL 室father-in-law → son-in-law

安東權 · otherfilled / hollow; dotted = lineage inferred

Each row of the grid is one generation from the apex 權幸. A dashed line is a marriage; the arrow runs from the bride's father to her husband, because the 1476 print records daughters only through their men.

P-graph · after White & Jorion

sonsolid — a man leaving his natal couple

daughterdashed — a woman the print does not name

A node is a couple, an edge is a person. In this representation a relinking is simply a cycle, and the eye needs no legend to find one.

Relinkings

Starting

安東權氏成化譜 · 1476