📄 Polysemy in specialized lexicon from Old English to Present-Day English

Summary

The article traces how English specialized vocabulary has gained new meanings over time instead of always coining new forms. Pérez Iglesias treats polysemy as a tool of linguistic economy and flexibility. Her material is a set of terms taken from the OED, each recorded with the date, author and work where a given sense first appears. Most of the terms belong to seven fields: Chemistry, Physics and Mathematics, along with Law, Commerce, Computing and Technology. The study is qualitative and organised by period, from Old English through Middle, Early Modern and Modern English to Present-Day English. It follows paths such as general word → specialized sense (Old English ceap, craeft), technical term → spread into other fields (Middle English alchemy and astronomy), and older nautical, military, craft or legal words → Computing senses (port, cartridge, spooler). The main finding is that new senses mostly pile up beside old ones rather than replacing them, usually through analogy or metaphor, and occasionally through ellipsis. In Present-Day English, technological change is the main driver. The work matters because it documents how sense inventories grow over time, and that growth drives lexical ambiguity in technical text.

Key points

• Data source: the OED and its supplements. Each sense carries a first-attestation date and source, for example craeft in the sense 'strength' (893, Ælfred's Orosius) and the 'trade/profession' sense around 900.
• Domains covered: Mathematics, Chemistry and Physics, together with Law, Commerce, Computing and Technology. Present-Day Computing senses are traced back to Law, Mathematics, sports and games, Fine, Practical and Minor Arts, and Nautical and Military Arts.
• Middle English cross-domain senses: Jupiter, Mars and Mercury each name a deity, a planet and a metal in alchemy (tin, iron and quicksilver; Chaucer, 1386).
• Early Modern English mathematical terms with other field senses:
  • calculus: medicine vs. mathematics
  • helix: geometry, architecture and anatomy
  • hexagon: geometry vs. fortification
  • parabola: rhetoric vs. geometry
  • prism: geometry vs. optics
• Conclusions of the paper:
  • Most OED addenda are new senses of existing words, not new words.
  • Older words tend to have more senses.
  • The author states that a technical term should keep one sense within each science.

Interesting ideas

• I1. The paper builds each sense inventory with a first-attestation date and a source, so meaning becomes a dated, layered history rather than a flat list; this makes it possible to explain why a sense exists and when it is plausible.
• I2. The paper treats a technical term as unambiguous within one field, with cross-field senses behaving like homonyms; the field context therefore picks the sense.
• I3. The paper describes ellipsis as a sense source: eye-glass / drinking-glass shrink to glass, and the full compound comes back whenever ambiguity arises.
• I4. In the paper's material, old senses survive beside new ones, and older words collect more senses; word age works as a rough prior for how ambiguous a word is.
• I5. The paper finds recurring metaphor paths in Computing vocabulary, such as nautical → data (navigation, anchor, port).

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (world): Domain and era tags on each WordNet/ontology sense would let the world model rule out anachronistic readings in older fables and tales, and say why.
• I2 (frames): Frame building could take the document's domain (computing, law, geometry) as an explicit, logged constraint when choosing between senses like port = harbour vs. access route.
• I3 (back): During generation, Mova could check whether a bare noun is ambiguous in the target meaning graph and fall back to the fuller compound to keep the text → graph → text round trip lossless.
• I4 (learn): Evaluation sets for sense selection could deliberately include old, sense-heavy nouns (craft, cell, port) as hard cases, with domain-swapped contexts serving as negative controls.
• I5 (world): Storing such systematic sense mappings as rules in the knowledge graph could help the world model handle new metaphorical uses without memorizing each one separately.