How to Use Korean Corpora Without Mistaking Frequency for Importance

Core examples: 말뭉치; 빈도; 용례; 공기어; 장르; 문어체; 구어체; 기사체; 자막; 검색식; 용례 수; 표본; 맥락.

The problem: frequency can look like a curriculum

A corpus can show that a Korean word, ending, particle, or phrase appears very often. That feels objective. Learners like objective lists because they promise efficiency: study the most frequent forms first and Korean will organize itself.

Frequency matters. But frequency is not the same as importance for a specific learner. A form may be frequent because the corpus contains news. A phrase may appear often in legal documents but be useless in conversation. A slang expression may appear in subtitle or comment data but be unsafe in formal writing. A bound form may be frequent but not independently learnable without a larger pattern.

The serious rule is: frequency is evidence, not a command.

Basic corpus terms

A token is an occurrence. If 그런데 appears 100 times, those are 100 tokens. A type is a distinct form. 빈도 is frequency. 용례 is an example of actual use. 공기어 or collocation refers to words that appear together more than chance. 장르 is genre: news, conversation, fiction, subtitles, academic writing, forms. 사용역 is register: formal, casual, spoken, written, institutional, intimate.

Metadata matters. A corpus without source type, date, speaker information, or genre labels is still useful, but its evidence is weaker. A learner should always ask: Frequent where? Frequent for whom? Frequent in what kind of text?

Why Korean makes corpus use especially valuable

Korean forms are context-sensitive. Particles, endings, honorifics, spacing, and pronoun omission all depend on discourse. A dictionary tells you what -더라고요 means. Corpus examples show where speakers actually use it. A grammar book explains 은/는. Corpus examples show how topic continuity and contrast appear across paragraphs. A list says 부탁드립니다 is polite. Corpus lines show its email, notice, and customer-service environments.

Corpus evidence also protects learners from invented examples. Natural Korean often omits subjects, compresses nouns, uses formulaic collocations, and prefers certain verb-noun pairings. A corpus can show whether a phrase is common, rare, genre-bound, or translationese.

Common mistakes

The first mistake is treating frequency as usefulness. A beginner does not need every frequent formal ending from news before learning everyday request forms. An advanced translator may need those formal endings urgently.

The second mistake is ignoring genre. If you search a news-heavy corpus, you will overlearn 밝혔다, 전했다, 관련, 전망, 정부, and according-to structures. If you search subtitles, you may overlearn casual fragments, interjections, and drama conflict. If you search official notices, you may overlearn nominalized bureaucratic Korean.

The third mistake is copying fragments. Concordance lines often show partial context. Korean meaning may depend on omitted subject, previous sentence, or genre. Do not turn a truncated line into a model sentence unless the full context is clear.

The fourth mistake is overgeneralizing from surface form. A form such as 다 can be a declarative ending, part of another construction, a dictionary form, or a fragment depending on context. Search results need interpretation.

A practical corpus workflow

Search a form. Inspect at least 20 lines, not just the first two. Group examples by pattern. Check genre. Compare with dictionary and grammar explanations. Decide what learner action follows: memorize, notice, defer, or avoid.

For example, search 덕분에. You may find thank-you expressions, causal explanations, public messages, and ironic uses. The learner action is not simply “덕분에 = thanks to.” It is “덕분에 marks positive attribution in many contexts, but can be ironic; it often pairs with 감사, 도움, 가능, 성장, or result clauses.”

Technical-review guardrail: corpus evidence is descriptive, not automatically normative

A corpus shows what appears in collected texts. It does not automatically tell you what is standard, polite, advisable, current, or safe. For spelling, pronunciation, and official norms, check authoritative references separately. For usage, compare corpus evidence across genres.

Mini practice: evaluate the corpus result

Learner workflow: responsible corpus note

1. Record the searched form.
2. Record corpus name, date accessed, genre, and filter settings.
3. Group examples by pattern.
4. Copy full sentences only when context is clear.
5. Add register and domain labels.
6. Decide learner action: active card, recognition note, domain glossary, or ignore.

Suggested functions:

1. Search log fields: query, corpus, genre, date, filter.
2. Concordance grouping: collocation, ending pattern, phrase frame.
3. Frequency warning: high frequency but low learner priority.
4. Register tagger: spoken, written, news, official, academic, online.
5. Card export: only after context and register are confirmed.

Final rule

Do not let frequency pretend to be importance. Korean corpus evidence is powerful when you ask where the form appears, what genre produced it, and what you should do with it as a learner.