Now this is thought-provoking. Google’s Marissa Mayer said explicitly Wednesday at the Searchnomics conference that Google’s voice-activated mobile search tool, 1-800-GOOG-411, is set to influence the company’s video search efforts by sharpening its speech-to-text knowledge, according to the Red Herring blog.

We had heard that GOOG-411 — which by the way is so much better than the other free 411 offerings I’ve tried, with no irrelevant ads, way better comprehension, and the option to connect you immediately — was perhaps just a front for harvesting voice data. And clearly a key application for speech recognition technology is video search. But it’s quite interesting to hear Mayer spell it out.

“What we have running is reasonably high quality in GOOG-411,” Ms. Mayer said. “The more accurate our voice-to-text becomes… Once you have accuracy you can build a transcript and time markers, and make video search more powerful… As you broaden (voice) recognition, pitches, genders, accents, ultimately that means you can apply it to the video corpus and text transcripts… A breakthrough is likely in the next one to two years. Visual search is promising but is probably further out than that.”