Repository navigation
feat(writing): add metaphor verbs and pseudo-concrete words to the AI-smell taxonomy - #251
Merged
Merged
Conversation
…-smell taxonomy AI-specific metaphor verbs (「地味に効く」「静かに壊れる」), pseudo-concrete words (「解像度を上げる」「腹落ち」) and framing fillers were neither caught by the textlint fallback nor described in norms. They depend on context, so they go into norms for LLM judgment rather than prh.yml. Pass 3 of writing-proofread also listed a stale category count and omitted mannered prose; it now points at the taxonomy without a count. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概要
AI が書いた日本語に多い比喩動詞・具体的に見せかける語・前置きと結びを、
norms.mdの AI-smell taxonomy に 1 項目として追加します。nanaism/yomiyasu のreferences/slop-catalog.md（MIT）から、技術文書で誤検出にならない語だけを選んで取り込みました。変更内容
home/dot_claude/references/japanese-writing/norms.mdに「Metaphor verbs and pseudo-concrete words（AI 偏愛語）」を追加するhome/dot_claude/skills/writing-proofread/SKILL.mdの Pass 3 を直すなぜ
prh.ymlではなくnorms.mdなのか候補の語を並べた検証用の文書を今の textlint 設定（ai-writing プリセットと
prh.ymlを含む）にかけたところ、指摘は助詞の重複 1 件と**ラベル**:形式の箇条書き 2 件だけでした。「いかがでしたでしょうか」「静かに壊れます」「解像度が上がりました」「革命的です」はどれも検出されませんでした。一方で、これらの語は文脈しだいで正しくも誤りにもなります（「壊れる」と「お腹を壊す」、「潰す」と「バグを潰す」）。文字列で照合する prh では区別できないため、LLM が判断する
norms.mdに入れました。取り込まなかったもの
norms.mdの方針と衝突する。yomiyasu のコーパスにlint.pyをかけると、書き直し後の 48 本の指摘数は 1 本あたり 1.16 件で、AI の素の出力（1.12）と変わらず、人間の文章（0.68）に近づかなかった検証
just test→✓ All checks passed!just fmt-check→✓ All files formatted correctly!tests/writing-bench/で変更前後を比べる必要があります🤖 Generated with Claude Code