RagLeap
LIN ZHANGSummary
2 pull requests (2 merged), 0 issues opened, 8 comments, 0 reviews, 0 inline review comments, 8 commits.
Pull requests (2)
- #1195 feat(voice): normalize Japanese kanji cardinal numbers — merged 2026-11-07 · opened 2026-11-05
- #1194 feat(voice): normalize Mandarin Chinese cardinal numbers — merged 2026-11-05 · opened 2026-11-05
Comments on issues and pull requests (8)
- 2026-11-07 · on #1195 feat(voice): normalize Japanese kanji cardinal numbers — comment: “Noticed [ac74702] on main already preserves zh tokens with out-of-order 十/百/千, so there is nothing left for me to open a follow-up zh PR for — verified the three tokens stay unchanged on main and the…”
- 2026-11-07 · on #1195 feat(voice): normalize Japanese kanji cardinal numbers — comment: “Synced with main: merged latest main into this branch and kept both language entries intact (zh and ja) in metrics.py, the tests, CHANGELOG and README. Local run of testbenchnumbers.py: 82 passed (bot…”
- 2026-11-06 · on #1195 feat(voice): normalize Japanese kanji cardinal numbers — comment: “Good catch. Fixed: each section now remembers the rank of the most recent 十/百/千 and returns None (preserves the whole token) when a multiplier equal or larger in rank follows, e.g. 二十三百, 三千二千, 十二百; th…”
- 2026-11-05 · on #1195 feat(voice): normalize Japanese kanji cardinal numbers — comment: “Thanks! Agreed — the intended behavior is now implemented: digit runs consist only of 〇 and private unit digits (一二三, 二〇二六, 一〇一 convert to digits), while 零 strictly follows positional-zero validation…”
- 2026-11-05 · on #1195 feat(voice): normalize Japanese kanji cardinal numbers — comment: “Thanks! Same class of fix as on the zh PR, applied here: a multi-character token starting with 零 (零一, 零零, 零百, 零一百) is now preserved unchanged and covered by regression tests; 零 alone -> 0, and 〇 digit…”
- 2026-11-05 · on #1194 feat(voice): normalize Mandarin Chinese cardinal numbers — comment: “Thanks for the full-suite pass and the quick review! Agreed on the leading-zero case. Fix pushed: a multi-character token starting with 零 (零一百, 零一, 零零, 零百) is now preserved unchanged, with regression…”
- 2026-11-05 · on #801 voice: number words for every language in the WER normalizer (claim a language) — comment: “I'd like to claim Japanese (kanji, ja). Native speaker of Chinese checked the shared CJK construction rules against CLDR ja RBNF and the Japanese numeral references; a fluent Japanese speaker has not…”
- 2026-11-05 · on #801 voice: number words for every language in the WER normalizer (claim a language) — comment: “I'd like to claim Mandarin Chinese (zh). Native speaker. The module, tests, changelog and README entries are ready and follow the numbersms.py pattern; the PR will be linked here in the next comment.…”
Commits (8)
- 2026-11-07 · Merge remote-tracking branch 'origin/main' into feat/voice-numbers-ja (
6ab475d) - 2026-11-06 · fix(voice): validate multiplier order within ja sections (
fd71e9e) - 2026-11-05 · fix(voice): ja digit runs use 〇 only; 零 stays a positional marker (
b509335) - 2026-11-05 · fix(voice): repair guard line break in numbers_zh (
5af85f9) - 2026-11-05 · fix(voice): preserve malformed Japanese tokens with a leading zero marker (
888dc0a) - 2026-11-05 · fix(voice): preserve malformed tokens with a leading zero marker (
a0fd694) - 2026-11-05 · feat(voice): normalize Japanese kanji cardinal numbers (
698344c) - 2026-11-05 · feat(voice): normalize Mandarin Chinese cardinal numbers (
7fea162)
How to verify this record
Everything above comes from the public GitHub API for ragleap/ragleap-core. Check it yourself: pull requests · issues opened · issues and PRs commented on · reviews · commits. GitHub is the source of truth.