Tagging

Extended and original taggers

ekonlpy.Mecab is the public tagger. By default it post-processes fugashi output with eKoNLPy's economic, institution, entity, and other bundled dictionaries:

from ekonlpy import Mecab

extended = Mecab()
original = Mecab(use_original_tagger=True)

print(extended.pos("금통위는 금리정책을 결정했다."))
print(original.pos("금통위는 금리정책을 결정했다."))

pos(text, flatten=True, include_whitespace_token=False) returns a list of (surface, tag) pairs. flatten=False preserves MeCab inflection components where the backend provides them. morphs(text) returns surfaces only, and nouns(text) returns noun surfaces.

When include_whitespace_token=True, each original whitespace run is returned as an SP token. This includes leading and trailing spaces, tabs, and newlines. Whitespace is kept separate from compound matching.

tagger = Mecab(use_default_dictionary=False)
print(tagger.pos("  가\t나\n", include_whitespace_token=True))

Nouns, synonyms, and lemmas

The extended tagger can normalize text before extracting research terms:

tagger = Mecab(use_default_dictionary=False)
tagger.add_dictionary("새별오름", "NNP")
tagger.add_synonym("물가안정", "물가 안정")
tagger.add_lemma("올랐", "오르다")

print(tagger.nouns("새별오름과 물가안정을 검토했다."))
print(tagger.replace_synonyms("물가안정"))
print(tagger.lemmatize([("올랐", "VV")]))

Each Mecab instance owns these mutable dictionaries. Changes made to one instance do not alter another instance.

Token strings

tokenize returns strings in surface/tag form by default. Set strip_pos=True to return surfaces only:

print(tagger.tokenize("금리정책", strip_pos=False))
print(tagger.tokenize("금리정책", strip_pos=True))

For dictionary loading and persistence, see Dictionaries.