| id | academic-english-discourse |
| name | Academic English Discourse |
| version | 0.2.0 |
| kind | style |
| description | Academic English discourse database and writing guidance for literature review drafting and EN/CN translation in Scholar AI. |
| entry_mode | assistant |
| ui_visibility | skill_assisted |
| display_group | writing |
| experimental | false |
| supported_scopes | ["selection","paragraph","section","full_draft"] |
| tags | ["academic-writing","translation","literature-review","phrasebank"] |
| permissions | {"model.llm":true,"retrieval.read":true,"draft.read":true,"draft.write":true,"files.read":true,"files.write":true,"network":true,"script.execute":true,"storage":true} |
| script_policy | {"has_scripts":true,"safe_to_execute":false} |
| model_policy | {"allow_llm":true,"allow_embedding":false} |
| root_policy | {"allowed_roots":["skill_root","project_root"]} |
| privacy_notes | The generator writes source-derived chunks only to the local workspace_artifacts tree. Do not paste bulk source text into prompts or public artifacts. |
| rollback_hint | Restore the package from .rollback_snapshots or delete workspace_artifacts/generated/output/english_discourse and rebuild from the original local sources. |
Academic English Discourse
Use this skill when Scholar AI needs to improve English academic writing, revise literature reviews, normalize scholarly terminology, or translate between Chinese and English while preserving disciplinary stance.
Core Intent
This is a database-backed writing skill, not an agent workflow. It supplies a local discourse database and a compact writing protocol so Scholar AI can:
- choose the right rhetorical move before choosing wording;
- rewrite Chinese thought into English academic argument structure;
- make literature review prose more cautious, connective, and evidence-bound;
- keep translated terms stable across a draft;
- avoid generic "AI-polished" prose that sounds fluent but academically vague.
Required Database
Build or refresh the local database with:
& .\.venv-1\Scripts\python.exe .\extension_packages\skills\academic-english-discourse\scripts\build_discourse_db.py --downloads-preset --include-phrasebank --ocr-engine auto
The generated artifacts live under:
workspace_artifacts/generated/output/english_discourse/
Expected outputs:
academic_english_discourse.sqlite3
chunks.jsonl
phrases.jsonl
discourse_frames.json
academic_english_habits.json
manifest.json
build_report.md
PDFs with no selectable text are rendered page-by-page and passed through Windows OCR when --ocr-engine auto is used on Windows. The resulting records use source_type="ocr_pdf" so they remain auditable.
After editing references/english_discourse_habits.md, rebuild the database so academic_english_habits.json.policy_markdown contains the latest official policy text.
Runtime Use
Load the knowledge in this order:
academic_english_habits.json.policy_markdown as the authoritative strategy text.
academic_english_habits.json structured fields for machine-readable checks.
discourse_frames.json for move-specific writing frames.
- SQLite/JSONL records for source-derived examples and phrase patterns.
When revising or translating, first map the user request to a discourse function:
territory: establish the research area.
gap: identify an unresolved problem or limitation.
aim: state what the paper or section does.
method: describe design, materials, data, or procedure.
result: report findings without overclaiming.
interpretation: explain what a result means.
comparison: compare studies, theories, datasets, or outcomes.
causality: express mechanisms and causal relations cautiously.
limitation: mark scope, uncertainty, or boundary conditions.
implication: state contribution, usefulness, or future work.
transition: move between claims or sections.
citation: attribute claims and synthesize sources.
Then retrieve matching chunks or phrases from the local database and produce adapted language. Do not copy a phrase mechanically if it does not match the claim, tense, certainty, evidence type, or discipline.
English Academic Habits
Use references/english_discourse_habits.md as the human-readable policy for this skill. The core habits are:
- choose the discourse move before choosing wording;
- order information from known context to new claim;
- bind every claim to evidence and scope;
- calibrate certainty through verbs and modality;
- keep technical terms stable across the draft;
- synthesize relations among studies instead of listing sources.
For Chinese-to-English work, rewrite the argument before polishing the surface. Typical conversions include:
- "่ฏดๆ/่กจๆ/่ฏๆ" ->
suggests, indicates, shows, or demonstrates by evidence strength;
- "็ ็ฉถไธ่ถณ" -> a specific gap in population, method, mechanism, dataset, theory, or context;
- "ๅ
ทๆ้่ฆๆไน" -> a concrete implication rather than generic importance;
- "ๆ็ ็ฉถ่กจๆ" -> a reporting verb plus a source relation.
Translation Protocol
For Chinese to English academic translation:
- Identify the discourse move before translating sentence by sentence.
- Convert Chinese topic-comment order into English claim-support order when needed.
- Preserve hedging and evidence boundaries; add caution only when the Chinese claim is underspecified.
- Stabilize key terms across the paragraph and avoid synonym drift for technical nouns.
- Prefer active scholarly verbs for argumentation and passive/nominalized forms only when the method or object should be foregrounded.
For English to Chinese translation:
- Preserve the rhetorical relation, not only the literal connector.
- Keep terms stable and record alternative translations only when the field genuinely uses both.
- Do not flatten hedging; distinguish "may", "might", "suggest", "indicate", "demonstrate", and "show".
Output Discipline
Good output should be:
- move-aware: the sentence has an identifiable academic function;
- evidence-bound: claims are scoped to the cited study, dataset, or review;
- cautious but not weak: hedge uncertainty, not established facts;
- cohesive: old information leads to new information;
- terminologically stable: one concept keeps one English form unless a distinction is needed.
See prompts/main.txt for the full prompt, references/english_discourse_habits.md for the distilled habits, and references/schema.md for the database schema.