R0021/2026-03-25/Q008/SRC02/E01¶
Academic research on the prevalence and nature of polysemy in natural language
URL: https://direct.mit.edu/coli/article/50/1/351/118497/
Extract¶
Key findings from polysemy research:
- Polysemy is pervasive — most content words in natural language are polysemous to some degree
- Frequency correlates with polysemy — more frequently used words tend to be more polysemous
- Context typically resolves ambiguity for human speakers, but polysemy has proved "notoriously difficult to treat both theoretically and empirically"
- Polysemy is distinct from homonymy — polysemous senses are related (unlike random coincidence of homonyms)
- Formal treatment challenges — formal systems like Type Composition Logic (TCL) and Modern Type Theory (MTT) attempt to handle polysemy through type-theoretic approaches, but this remains an active research problem
Relevance to Hypotheses¶
| Hypothesis | Relationship | Strength |
|---|---|---|
| H1 | Supports | Polysemy is pervasive and difficult to formalize — confirms the gap |
| H2 | Contradicts | Research confirms polysemy is a genuine challenge |
| H3 | Supports | Context resolves ambiguity for humans, qualifying the gap |
Context¶
The finding that polysemy is "notoriously difficult to treat both theoretically and empirically" is directly relevant to prompt engineering. If formal linguistics cannot reliably handle natural language ambiguity, expecting natural language prompts to serve as engineering specifications faces a fundamental limitation.