Academic knowledge is written almost entirely in natural language, which makes it readable by humans but hard to query across disciplines that use different words for related things. We describe a tag layer that assigns every paper the same small set of field-neutral facets — claim type, phenomenon (a normalised join key), claim relations, replication status, and limitations — under one non-negotiable rule: every tag is grounded in an exact, verbatim sentence from the paper's own text. We report a proof-of-format on a curated corpus of 126 papers spanning 39 disciplines (1866–2023), of which 116 were fully faceted. The format is feasible and verifiable: 127 claim-relations and 87 limitations, each substring-checked against the source. It is also honest about what it did not do: a novelty scan surfaced no cross-disciplinary link that the literature had not already drawn, and the phenomenon facets, written as specific verbatim-anchored strings, rarely collide across fields — indicating that a canonical join vocabulary, not free text, is the missing ingredient for surfacing new links. We argue the value of the layer is as a compass for "far reading" that points where slow close reading may be worthwhile, not as an answer engine.
Shir Sivroni (Sat,) studied this question.