review-cet-writing

CET-4/CET-6 Writing and Translation Correction

Perform strict correction of CET-4/CET-6 writing and translation, and output actionable suggestions and a visual report.

Workflow

1. Identify task type and level

• Determine task_type：writing or translation。
• Determine level:CET4 or CET6。

1. Read scoring criteria

• Writing: read references/writing.md。
• Translation: read references/translation.md。
• Chart/Expository writing: additionally read references/expository-writing.md。

1. Produce correction results

• First give the total score (0-15) and dimension scores (usually each item 0-4）。
• Mark three types of issues:critical、warning、tip。
• Give the original sentence, replacement suggestion, and reason for each issue.
• Give 3-5 score-improvement suggestions, avoiding vague template phrases.

1. Generate JSON and HTML report

• Organize the JSON according to the following fields:
  • task_type、level、overall_score、dimensions、issues_summary
  • original_text、text_with_highlights
  • suggestions（id/type/title/original/replacement/reason）
  • improvement_tips（title/content）
• Run:

python scripts/generate_report.py cet_review.json --output output/report.html

• The script will automatically convert the report score (106.5 scale).

1. When only doing score conversion

python scripts/convert_score.py 12
python scripts/convert_score.py cet_review.json --output cet_review_converted.json

Resources

• scripts/generate_report.py: Render the correction JSON as an HTML report.
• scripts/convert_score.py: Conversion from the 15-point scale to the 106.5 report score scale.
• assets/report-template.html: Report template.
• references/*.md: Scoring criteria and examples.

Constraints

• Always score according to strict standards; do not give inflated encouragement scores.
• Do not fabricate "official detailed rules"; base conclusions only on references/ and the user's text.
• The suggestions.id in the report must correspond one-to-one with the text_with_highlights in data-id .