Building Korean Vocabulary by Register Instead of Frequency Alone

Article body

Frequency lists are useful, but they can mislead serious learners. A high-frequency word may be too broad to use well. A lower-frequency word may be essential in a domain you read every week. A word can be common in news but strange in conversation, common in school notices but rare in novels, common in product pages but useless in academic writing. Korean vocabulary grows faster when you organize by register and family, not by frequency alone.

Take the family 학-. A flat list may teach 학교, 학생, 학습, 학문, 학점, and 학기 as separate words. A serious deck sees a network: school institution, learner/person, learning process, scholarship/discipline, credit, semester. The shared Hanja root helps, but the register differs. 학교 is everyday. 학습 is formal/education-register. 학문 is academic/intellectual. 학점 belongs to university administration. 학기 belongs to school calendars.

Now take 정- as in 정치, 정책, 정보, 정리, 정확. Not every 정 is the same Hanja or same semantic family, so the learner must avoid fake grouping. Word-family learning works only when you verify roots, meanings, and collocations. It is not “all syllables that sound the same go together.”

Register tags matter. 말, 언어, and 스피치 do not behave the same way. 말 is native and everyday. 언어 is formal, academic, and general. 스피치 is a loanword tied to presentation, training, and performance contexts. 일, 업무, and 워크 similarly belong to different source worlds. A deck that translates all as “work” is actively harmful.

A better Korean vocabulary deck has fields: word, source sentence, word family, register, domain, collocations, near-synonyms, active/passive status, and warning. Learners do not need to produce every word they can recognize. For many formal terms, recognition and accurate reading are enough at first.

Register-first deck fields

Guided example

Instead of one card that says:

지원 = support

Build a family:

• 도움: everyday help
• 지원: support/assistance, often institutional
• 후원: sponsorship/support, often charity/culture
• 보조금: subsidy/grant-like money
• 지원하다: apply for / support, depending on object and context

Now the learner can read 지원하다 in both “회사에 지원하다” and “취약계층을 지원하다” without collapsing the meanings.

Learner traps

Do not trust frequency without genre. Do not group homophonous Sino-Korean syllables without checking Hanja or meaning. Do not make giant cards with ten related words and no usage. Do not treat formal words as better words. Often the right Korean word is the less fancy one.

Reusable workflow

1. Choose one anchor word from a real source.
2. Add two to five related words, not twenty.
3. Tag each by register and domain.
4. Add one collocation for each word.
5. Decide active vs recognition-only.
6. Review by family occasionally, but review individual words in sentences.

Additional practice and repair

The central fix is to stop treating frequency as a synonym for usefulness. High-frequency vocabulary is important, but advanced reading grows from families, registers, collocations, and source tasks. This article should make the deck architecture concrete enough that a learner can build it the same day.

Remediation diagnostic

Before/after deck repair

Weak deck entry:

개선하다 — improve

Remediated entry:

개선하다 — to improve a system/process/condition. Common objects: 제도, 환경, 문제, 상황, 서비스, 품질. Register: formal/workplace/policy. Related family: 개선, 개선책, 개선 방안, 개선되다. Example: 서비스 품질을 개선하기 위해 새로운 절차를 도입했다.

Added workflow: the 3-tag minimum

Every serious vocabulary card should have at least three tags:

1. Layer: native, Sino-Korean, loanword, hybrid, idiom, slang.
2. Register/source: casual, formal, official, academic, workplace, media, product, regional.
3. Collocation frame: common verbs, objects, modifiers, or endings.

A learner can still use frequency, but frequency should decide review priority, not card structure.

The word-family deck builder should prevent isolated cards by default. If the user enters 학습, the tool should prompt for related family members, common collocations, and a source sentence before exporting to Anki.

Suggested interactive/tool module

Build a word-family deck builder that imports a sentence, suggests related words, asks the learner to tag register/domain, and warns against homophone-based fake families. Export cards by recognition, production, collocation, and reading-only status.