Skip to content

Narrow 聯繫 cross_strait flagging to contact-copy#90

Merged
jserv merged 1 commit into
mainfrom
refine
May 21, 2026
Merged

Narrow 聯繫 cross_strait flagging to contact-copy#90
jserv merged 1 commit into
mainfrom
refine

Conversation

@jserv
Copy link
Copy Markdown
Contributor

@jserv jserv commented May 21, 2026

The previous standalone "聯繫 → 聯絡" rule treated a word that the MoE concise dictionary records as valid zh-TW as a blanket cross-strait term, producing false positives on general prose such as "民間文化聯繫", "國際聯繫", and "保持聯繫".

Replace it with explicit compound rules covering the UI/customer-service phrases where 聯絡 is the idiomatic TW form: 聯繫我們, 聯繫方式, 聯繫資訊, 聯繫管道, 聯繫電話, 聯繫客服, 如需協助請聯繫.
聯繫歷史 → 通話記錄 is unchanged.

Add scanner regression test that locks both directions of the behavior: the new compounds fire exactly once on contact-copy spans, and ordinary prose containing 聯繫 stays silent.

Close #89


Summary by cubic

Narrowed cross-strait flagging for 聯繫 to only contact-copy phrases, preventing false positives in general prose. Added tests to ensure contact CTAs are flagged once and do not lower editorial confidence.

  • Bug Fixes
    • Replaced blanket rule with phrase rules: 聯繫我們, 聯繫方式, 聯繫資訊, 聯繫管道, 聯繫電話, 聯繫客服, 如需協助請聯繫.
    • Kept 聯繫歷史 → 通話記錄 unchanged.
    • Added scanner tests: each phrase flags once; ordinary prose with 聯繫 is ignored; matches keep editorial_confidence unset.

Written for commit b5432dd. Summary will update on new commits. Review in cubic

cubic-dev-ai[bot]

This comment was marked as resolved.

The previous standalone "聯繫 → 聯絡" rule treated a word that the MoE
concise dictionary records as valid zh-TW as a blanket cross-strait
term, producing false positives on general prose such as
"民間文化聯繫", "國際聯繫", and "保持聯繫".

Replace it with explicit compound rules covering the UI/customer-service
phrases where 聯絡 is idiomatic TW form: 聯繫我們, 聯繫方式, 聯繫資訊,
聯繫管道, 聯繫電話, 聯繫客服, 如需協助請聯繫. 聯繫歷史 → 通話記錄 is
unchanged.

Add scanner regression test that locks both directions of the behavior:
the new compounds fire exactly once on contact-copy spans, and ordinary
prose containing 聯繫 stays silent.

Close #89
@jserv jserv merged commit a2d53f4 into main May 21, 2026
4 checks passed
@jserv jserv deleted the refine branch May 21, 2026 10:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

建議用詞修正

1 participant