You are curating scientific keywords for the research domain "{domain_title}".{domain_description}

A VALID keyword is ONE well-formed noun phrase in its source language
that names exactly one of:
  C  CONCEPT / phenomenon / theory / model
     e.g. "phase transition", "active matter", "morphogenesis"
  M  METHOD / technique / instrument / numerical scheme
     e.g. "atomic force microscopy", "molecular dynamics", "femtosecond laser"
  O  OBJECT of study / material / system / organism / device class
     e.g. "polymer brush", "budding yeast", "granular medium"
  P  PLACE or SETTING studied as such: a kind of environment, habitat
     or site, not a particular named place
     e.g. "tidal flat", "urban heat island", "deep sea"
  D  FIELD: the name of a discipline or research field
     e.g. "biophysics", "fluid mechanics", "géomorphologie"

SCOPE RULE (read carefully):
Researchers of the domain legitimately work at interfaces with
neighbouring disciplines (astrophysics, chemistry, biology, computer
science, mathematics, earth sciences, engineering...). Such
interdisciplinary terms are SIGNAL, not noise. ACCEPT them as long as
they name a real scientific concept / method / object — even if they sit
outside the core of the domain. Only treat a term as out-of-scope if it is clearly NOT a
natural-science / engineering / quantitative-method term at all (pure
humanities, pure law, pure literature, etc.); in that case use code H.

PERSON NAMES (read carefully):
In some fields (history, philology, archaeology, art history...) persons
are legitimate OBJECTS of study. A person-name keyword is ACCEPTED
(code O) ONLY when BOTH conditions hold:
  1. The person is famous in the research field AS a research object —
     someone studied, edited, excavated or collected (ancient authors,
     historical figures, notable collectors: "Galen", "Titus Livius",
     "Vivant Denon").
  2. The keyword is the complete, unambiguously recognizable name of
     that person — NEVER a fragment of a name, a bare common-word name
     part ("vivant"), a lone surname ("garcia"), or a citation artefact
     ("vellmoor (ed.)").
Contemporary researchers, co-authors, collaborators, supervisors and
cited scholars NEVER qualify, however often they appear in the corpus —
reject them with N. When in doubt about a person name, REJECT with N.{person_whitelist_block}

ACCEPT only if ALL of these hold:
  1. The phrase is fluent in its source language. No reordering, no
     missing preposition or conjunction would be required to translate.
  2. The phrase is informative on its own — a scientist could place it
     under some scientific subfield from the term alone.
  3. It names a specific concept / method / object — not a bare ultra-
     generic noun.

REJECT with the most specific code:
  N  person (unless accepted under the PERSON NAMES rule above), name
     fragment, citation artefact ("x (ed.)") / lab / institution /
     journal / publisher / city / venue
  K  admin, CV, project-management or generic action-verb phrase
     ("team leader", "funded project", "PhD supervision")
  G  bare ultra-generic noun usable in any field with no specific
     scientific content (measurement, simulation, function, structure,
     property, method, system, approach, result...); a single everyday
     word ("shape", "weight", "poids") unless it is a term of art of
     the field ("entropy" in physics)
  F  broken piece, not a term: a phrase cut out of a longer one, words
     split across a phrase boundary, debris of a sentence, header
     artefact, garbled or OCR-broken token, words in an order no
     speaker would use; bare acronym you cannot confidently expand.
     F is ONLY for broken pieces (see PROCESS OF AN OBJECT below).
  H  a real term, but not informative in THIS field: too common among
     its researchers to tell them apart, or outside its scope.

K, G and F say the phrase is never a keyword, in ANY field: use them
only when you are sure of it. When a phrase could be a keyword
elsewhere, or you are unsure, use H.

PROCESS OF AN OBJECT (read carefully):
A phrase that joins a process, a property or a measure to an object of
the field ("protein folding", "enzyme turnover rate", "repliement des
protéines", "croissance des levures") is a VALID keyword, usually C,
when both parts belong to the field. French, Spanish, Portuguese and
Italian often write it "X des Y", "X de Y", "X della Y"; English and
German as a compound ("Proteinfaltung"): judge the whole phrase, and do
not reject it in favour of its object alone. German writes every noun
with a capital: a capital alone does not make a person or a place.

WHEN IN DOUBT, prefer ACCEPT for any term that looks like a real
scientific concept / method / object, even if interdisciplinary or
unfamiliar. Silently dropping a valid term loses information; the only
things that *must* go are admin / proper-noun / generic / fragment
cases.

CANONICAL {reference_language_name} FORM (only for C / M / O / P / D):
  - Give the canonical form of the term in {reference_language_name}.
  - Singularize plurals ("polymers" -> "polymer").
  - Prefer the substantive: when a natural noun names the same notion,
    the canonical form is that noun, not the adjective ("hagiographic"
    -> "hagiography"). Keep an adjective-derived form when it IS the
    discipline's name ("numismatics", "ceramics" stay as they are).
    When unsure, keep the corpus's dominant surface form.
  - Translate terms from other languages into {reference_language_name}
    (e.g. when the reference language is English, "matière molle" ->
    "soft matter", "polymère" -> "polymer", "fluide complexe" ->
    "complex fluid", "weiche Materie" -> "soft matter", "materia
    blanda" -> "soft matter"). Keep established technical terms that
    specialists in this field use untranslated.
  - When a {reference_language_name} term of the list names the same
    thing as a term in another language, give the other term exactly
    the canonical form you give the {reference_language_name} term:
    terms merge across languages only when their canonical forms are
    identical.
  - Expand standard abbreviations ("NMR" -> "nuclear magnetic
    resonance", "AFM" -> "atomic force microscopy").
  - Reorder to a fluent {reference_language_name} noun phrase (e.g. the
    French adjective-after-noun -> English adjective-before-noun flip,
    "matière molle" -> "soft matter"). Never insert "of", "and", "the",
    "interface", "between", or any new word that does not directly
    translate one of the input words.

OUTPUT FORMAT — exactly one line per input term, in input order:
  C <lang> <term>=<canonical>
  M <lang> <term>=<canonical>
  O <lang> <term>=<canonical>
  P <lang> <term>=<canonical>
  D <lang> <term>=<canonical>
  N <term>
  K <term>
  G <term>
  F <term>
  H <term>
Where <lang> is the two-letter ISO code of the term's source language
and <canonical> is its canonical {reference_language_name} form. No
numbering, no commentary, no blank lines. If the input has N terms,
output exactly N lines.
