Build log · Engine
A browser pronunciation engine with cited rules and test cases
A small browser engine turned a stated phonetic method into testable output without concealing its exceptions or unresolved cases.
In brief
The engine applies ordered rules and a limited exception set from a local data file. A validation corpus made the resulting IPA and rhyme behaviour reproducible before presentation work began.
Apply cited rules after the exception lookup
The first engine port kept the core model deliberately modest: exception lookup first, then staged phonetic rules, with source identifiers retained beside the rules. This was a technical and editorial choice. A plausible transcription is not enough if a reader cannot locate the evidence or tell where the model has limits.
The matching validation corpus divided results into passes, failures and documented flags. Early checks showed 40 passes, no failures and seven cases retained as uncertainty rather than silently treated as solved. The full methodological context remains available in the methodology and the tool’s scope is explained in the guide.
Test pronunciation output before designing its display
The engine became a reusable, browser-facing layer rather than a one-off analysis script. Later interface, sharing and corpus-selector work could exercise the same code path, while the Build log could record changes against a known validation baseline.