Dictionaries

Add words for one tagger

Use add_dictionary for words that should receive a MeCab tag. The tag must be in the installed tag set unless force=True is intentional:

from ekonlpy import Mecab

tagger = Mecab(use_default_dictionary=False)
tagger.add_dictionary(["금융위", "새별오름"], "NNP")
print(tagger.pos("금융위와 새별오름"))

load_dictionary(path, tag) reads one word per line. add_terms and load_terms add words to the internal term categories used by nouns and sent_words; their accepted categories include COUNTRY, INDUSTRY, NAME, SECTOR, ENTITY, GENERIC, CURRENCY, and UNIT.

use_default_dictionary=False disables the bundled word loading, but the tagger still initializes its synonym, lemma, and term-category resources. Use the instance methods above to add or replace the entries relevant to a study.

Synonyms and lemmas

Synonyms map a word to a replacement used by replace_synonyms and downstream sentiment word extraction. Lemmas map an inflected form to its base form:

tagger.add_synonym("한은", "한국은행")
tagger.add_lemma("올랐", "오르다")

print(tagger.replace_synonyms("한은"))
print(tagger.lemmatize([("올랐", "VV")]))

Use load_synonyms(path, tag="NNG") or load_lemmas(path) for files. Each vocabulary row has two values separated by a comma, for example:

한은,한국은행
인상했다,인상하다

Blank lines and lines beginning with # are skipped. A nonempty malformed row raises ValueError with the file name and line number. Loading is read-only; the input file is never rewritten. persist_synonyms() and persist_lemmas() write the current maps to the package data directory, so use them only when you intend to modify that installation.

Instance isolation

Custom dictionaries, synonyms, lemmas, terms, and mutable tagger settings are owned by each Mecab instance. Create a fresh instance for an independent experiment or test.