NEWS

---

RmecabKo 0.3.0

RmecabKo grows from a thin wrapper into a Korean text-analysis layer over RcppMeCab, filling the gap left by the archived KoNLP and elbird packages.

Tokenizers and tidytext

• Tokenizers now follow the tokenizers package contract, so token_morph(), token_words(), token_nouns(), and token_ngrams() drop directly into tidytext::unnest_tokens(token = ...). Each gains a simplify argument.
• Add a drop_pos argument to token_morph() and token_ngrams() to remove tokens by POS tag, complementing keep_pos.
• Breaking: unnamed input now yields an unnamed list. Previously each element was named after its (potentially long) source document. Named input is unchanged.

Korean data

• Add stopwords_ko, a curated table of Korean function morphemes, with the accessors stopwords_ko_words() (surfaces) and stopwords_ko_tags() (POS tags for drop_pos).
• Add lexicon_knu(), which downloads and caches the KNU Korean sentiment lexicon for tidy sentiment joins.
• Add the demo_ko demonstration corpus.

User dictionaries

• Add dict_add_words(), dict_words(), dict_remove_words(), dict_compile(), dict_use(), and dict_path() to manage a MeCab user dictionary from a data frame of words, filling the mecab-ko-dic context IDs and final-consonant flag automatically.

Analysis helpers

• Add token_lemma() and lemmatize_morphemes() to recover dictionary forms of Korean predicates. Fused tokens whose stem RcppMeCab does not expose return NA.
• Add keywords_tfidf() and keywords_textrank() for morpheme keyword extraction, and kwic() for keyword-in-context concordances.
• Add text_normalize() for NFC composition, width folding, and repeated- character squashing.

Documentation

• Add the "Korean text analysis with RmecabKo" vignette walking through a full tidy workflow.

RmecabKo 0.2.1

• Require the companion RcppMeCab 0.0.1.7 release used by the 0.2.0 migration to centralized MeCab and dictionary management.
• Skip the end-to-end integration test when a check platform has no compatible Korean backend while retaining deterministic dictionary-validation tests.
• Use a valid external URL for the Korean README in built source packages.
• Encode Korean documentation examples with R Unicode escapes so the PDF reference manual builds with CRAN's LaTeX toolchain.

RmecabKo 0.2.0

• Use RcppMeCab as the sole native MeCab engine and verify that the active dictionary produces Korean POS tags.
• Make noun, content-word, and morpheme tokenizers consistent for scalar, vector, named, list, empty, and missing inputs.
• Filter punctuation, numbers, and POS categories after morphological analysis.
• Add multiple n-gram sizes, exact skip-grams, POS filters, and stopword boundaries. Empty and too-short documents now safely return character(0).
• Deprecate the obsolete standalone MeCab installer.
• Preserve UTF-8 text on Windows and normalize all analyzer input at the R boundary (fixes #1).
• Support compiled cross-platform user dictionaries through user_dic, with compilation delegated to RcppMeCab::dict_index() (fixes #2).

RmecabKo 0.1.7.0

• Replace package-local POS tagging with RcppMeCab.
• Fix encoding issues.

RmecabKo 0.1.6.2 (2018-02-13)

• Fix install_mecab() subdirectory handling.
• Fix local encoding checks in POS and noun extraction.