identify-terminology - SKILL.md Agent Skill

name: identify-terminology description: Identify and classify technical terminology in academic papers. Use when determining which terms need glossing, assessing domain and jargon density, and tracking acronyms for consistent handling.

Universal academic vocabulary — no glossing needed in any register:

hypothesis, methodology, corpus, analysis, paradigm, framework, empirical, theoretical

Field-standard vocabulary — gloss in none/selective registers, not in heavy:

Terms all practitioners in this field know without explanation
Examples: BERT, transformer (ML); p-value, ANOVA (stats); code-switching (linguistics)

Specialized sub-field vocabulary — gloss in all registers except assume-knowledge:

Terms requiring specific coursework or deep specialization
Examples: /æ/-tensing, F1/F2 formants (phonetics); epistemic violence, habitus (critical theory)

Novel/paper-specific terms — always gloss on first use:

First use: expand in full with brief explanation. Format: ACRONYM (Full Name — what it means) Subsequent uses: acronym only.

Track which acronyms have been introduced. Don't re-expand.

From title/abstract keywords, identify primary domain:

ML/AI: neural, learning, model, transformer, attention, training, inference
NLP: language, text, translation, parsing, NLU, generation
Computer vision: image, visual, detection, segmentation, pixel
Linguistics: phonology, syntax, morphology, pragmatics, corpus, sociolinguistics
Statistics: regression, Bayesian, significance, variance, distribution
Critical theory: power, discourse, hegemony, subjectivity, ideology

Domain affects which terms count as "field-standard" vs. "specialized."

Assess the paper's own register to calibrate synthesis:

Papers that write "utilize" and avoid contractions throughout → likely higher-jargon audience
Papers with accessible prose in introduction → broader audience intended
Nature/Science papers → interdisciplinary, accessible style appropriate
Workshop/conference papers → expert audience assumed