Record empirical basis for repeated glossary passes

This commit is contained in:
2026-11-11 15:51:30 +00:00
parent 2b5b0d6051
commit 2261a38a6b

View File

@@ -3,8 +3,9 @@
**Updated:** 2026-11-11
**Status:** Topic 1 resolved in ADR 0003; topics 2–4 decided, with implementation
pending; topic 5 deferred; grammar naming and scope in topic 6 decided, with
description updates pending; the remaining topic 6 choices await discussion.
pending; topic 5 deferred; grammar naming/scope and glossary/homophone overlap
in topic 6 decided, with description updates pending; language/examples and
meaning-reversal policy await discussion.
Work through the remaining topics in order. Their recommendations are starting
points for discussion, not newly adopted restrictions or guarantees.
@@ -15,6 +16,12 @@ The review of revision `6d34602` covered all eleven
the context and response handling for proposals, validation, and repair.
It was a design review, not a measurement of live-model quality.
The maintainer reports that nearly all module behavior was selected through
practical experimentation. Treat that experience as evidence when reviewing the
design: overlapping responsibilities and repeated passes are not inherently
redundant. The deferred evaluation work does not imply an absence of empirical
support for the existing module design.
Retain the strengths: conservative transcript editing, explicit literal-span and
deletion semantics, distinct glossary context and protection, full before/after
validation, independent repair conversations with withdrawal, and native
@@ -167,8 +174,17 @@ without waiting for the deferred evaluation work:
editorial allowance. Update the relevant module descriptions and canonical
documentation during implementation; no rename or broader grammar-correction
remit is planned.
2. **Glossary/homophone overlap:** assess the benefit of their overlapping remit
and repeated glossary passes before changing the default sequence.
2. **Glossary/homophone overlap — Decided:** retain both modules, their useful
overlap, and the existing default sequence documented in
[configuration](../config.md#pipeline-and-glossary-selections). The maintainer's
practical experience establishes that two glossary passes help, and some
workloads benefit from three or more. Preserve operators' ability to configure
additional passes. Clarify the complementary requested roles: glossary stages
focus on known terms supported by selected glossaries; homophone stages look
more broadly for plausible spoken-word mistranscriptions and may also use
glossary evidence. Do not exclude an otherwise supported homophone correction
merely because it targets a glossary term. No module consolidation or default
sequence change is planned.
3. **Language and examples:** decide whether spoken English and campaign-specific
examples are the right bundled defaults, and how broadly those defaults
should apply.
@@ -176,5 +192,5 @@ without waiting for the deferred evaluation work:
of a mistranscription. Initially retain the conservative default while
explicitly evaluating that tradeoff.
The remaining items are product discussions, not selected changes to module
sequencing, language support, or validation policy.
The remaining items are product discussions, not selected changes to language
support or validation policy.