A regex matched my LLM pipeline on clean text. Then synthetic speech broke both.
An operator on a plant radio says "Close valve two-oh-three, line Bravo." The field technician answers "Copy, closing two-oh-eight, line Bravo." Valve 208 is on a different line. The read-back exists so somebody catches that, and sometimes nobody on the channel does. I spent the AMD Developer Hackathon (ACT III, lablab.ai) building SayAgain , a checker for exactly that moment. This post is not…
The operator on a plant radio instructed, "Close valve two-oh-three, line Bravo." The field technician responded, "Copy, closing two-oh-eight, line Bravo." Valve 208 was on a different line, causing confusion. The read-back system was in place to catch discrepancies, but sometimes operators failed to catch them. At the AMD Developer Hackathon, the author developed SayAgain, a tool to detect these issues.
This post is not the launch of SayAgain; instead, it details a surprising discovery: a plain regex worked just as well as the language model when processing clean text, but synthetic speech caused both to underperform in a way that headlines often overlook.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
