Jev in Channel Finder: using typed relevance scoring for similar-channel search
A practical case study on where semantic judgment belongs, what ordinary code should still own, and how to evaluate the boundary.
A practical case study on where semantic judgment belongs, what ordinary code should still own, and how to evaluate the boundary.
Separate publication time, capture time, and evaluation time so historical creator activity filters can be reproduced without pretending stale observations are current.
A small JavaScript layer for pitch-ratio calculations, signed fine-tuning corrections, and the browser playback-speed trap.
A manual comparison procedure for checking vocal identity and artifacts alongside a speaker-similarity metric.
I make Learn Thai Alphabet & Tones, so Thai text is a useful place to be precise about what a JavaScript "position" means. For this article I wrote and ran an original diagnostic to inspect two string
Resolve aliases to confirmed channel IDs, group matching identities, and preserve source rows for review and correction.
มา and ขา are a useful pair for a very small Thai tone-rule exercise. Both use the long vowel า, both are open live syllables, and neither has a written tone mark. The relevant difference is the initi
How to define the observation window, compute a median correctly, and avoid reading more into a YouTube metric card than the data supports.
I built the Key Changer Engine because I wanted a desktop Chrome transposer that sounded acceptable to my own ears in both directions. Upward vocals could sound too cartoonish to me, while downward sh