The Kmiec Ramblings

Overview
Our good friends at Google have added in a new feature to YouTube video search: speech recognition. Previously their search algorithm/method was based 100% on the video’s title and the tags associated with the video. YouTube video search will now integrate the actual spoken words in the video into the search.

Here’s a great example. Let’s say you have a video titled: NASCAR Accident and the tags used are: cool, random, car, NASCAR, crash. Well if someone is looking for Dale Earnhardt Jr. videos, the “NASCAR Accident” would never show up. However, with the new speech recognition feature, YouTube would include the video because the announcer’s voice over in the video mentioned Dale Earnhardt Jr.

Adding to the cool factor and increasing the usability is the ability to jump right to a specific point in a video based on what was being said. Check out this image to see an example.

Speech Recognition

Right now Google is only testing this with political videos. It’s only a matter of time until all of the videos can leverage this new functionality.

The Example

1. Go to this page
2. Look on the lower left for the module/section called “what did the candidates say”
3. Enter in some search terms and enjoy

Possible Implications
For a while now, we’ve know that a good title, with great tags, and an eye-catching first frame help get videos found, noticed, and watched. With the ability to search and find videos using the spoken words in them, things will change quickly. More attention to the word selection in the copy will be needed. Using music as the audio could actually hurt us; because the spoken words from the song will be part of the search. Sound Sweetening and audio mixing will become very important and will need to be budgeted for. It’s possible that we can run the spoken audio at a low sound level so that YouTube picks up on it, but the user never really hears it.

Trackback URI | Comments RSS