Inconsistent POS tag conventions cause incorrect pronunciations for "read", "reread", "wound"

[heyaphra]

Problem

Present tense sentences are pronounced incorrectly:

• "I read books every day" → "red" instead of "reed"
• "I reread this chapter weekly" → "re-red" instead of "re-reed"

Root Cause

The dictionary mixes two tagging conventions inconsistently:

Misaki's own tags (work correctly):

Tag	Entries	Usage
DEFAULT	789	Fallback pronunciation
VERB	272	Generic verb form
NOUN	425	Noun form
ADJ	63	Adjective form
None	32	Stressed function words

Penn Treebank tags (used inconsistently):

Tag	Entries	Issue
VBD	4	Correct (past tense)
VBN	3	Correct (past participle)
VBP	3	All 3 are wrong

Since Misaki uses spaCy (en_core_web_sm), the tagger outputs Penn Treebank tags. When spaCy returns VBP (present tense, non-3rd person singular), the dictionary maps it to past-tense pronunciations:

"read":   {"VBP": "ɹˈɛd", ...}   // VBP="I read daily" should be "reed"
"reread": {"VBP": "ɹiɹˈɛd", ...} // VBP="I reread often" should be "re-reed"
"wound":  {"VBP": "wˈWnd", ...}  // VBP="I wound it" should be "woond" (injury)

It appears VBD/VBN/VBP were added specifically for these homographs, but VBP was mapped using past-tense semantics rather than Penn Treebank's definition (present tense).

Suggested Fix

Option A: Fix the 3 entries (minimal change)

"read":   {"DEFAULT": "ɹˈid", "VBP": "ɹˈid", "VBD": "ɹˈɛd", "VBN": "ɹˈɛd", "ADJ": "ɹˈɛd"}
"reread": {"DEFAULT": "ɹˌiɹˈid", "VBP": "ɹˌiɹˈid", "VBD": "ɹiɹˈɛd", "VBN": "ɹiɹˈɛd", "NOUN": "ɹˈiɹid"}
"wound":  {"DEFAULT": "wˈund", "VBP": "wˈund", "VBD": "wˈWnd", "VBN": "wˈWnd"}

Option B: Standardize tagging convention (architectural)

Consider whether to:

• Use Penn Treebank tags consistently (VB, VBD, VBG, VBN, VBP, VBZ)
• Use Misaki's own tags consistently (VERB, NOUN, ADJ, etc.)
• Document the hybrid approach and ensure mappings are correct

Affected Entries

These are the only 3 entries with VBP tags, and all three are incorrect:

read, reread, wound

---

Happy to submit a PR for Option A, or discuss Option B further.

[joshwhiton]

I'd appreciate a PR for this, even if simply Option A. Since PRs seem to be piling up here I've been accepting them on my fork, and Hexgrad may log in and accept them eventually too.

[joshwhiton]

Here's my misaki fork with a PR of mine and several others incorporated until this main repo starts accepting PRs again. You can still just submit the PR to this official branch if you prefer and I'll grab it.