The topic of video search optimization was tackled by a group of panelists at the Search Engine Strategies conference held in New York City this week. Eric Papczun, Director of Natural Search for Performics, talked about some of the trends he’s observed in the video search space. In particular, he made references to user-generated tags that segment scenes within a video file, so that instead of waiting several minutes to get to the part of interest, viewers can jump to a particular portion of the video to begin watching the content by clicking on one of the tags.

In the absence of a transcript to help search spiders crawl and properly index video files, it’s of no surprise to see real humans and not machines taking the lead to organize and segment video content available online. Video search engine PodZinger is tackling the tagging problem by running all of its indexed audio and video through proprietary speech-to-text technology to create text files that put a time stamp of every word mentioned in a file. Through this automated process, PodZinger makes it possible to search the entire content inside a video, and take the user directly to the section of the video of interest.
Neither approach to tagging video content is flawless. In the manual way, human errors can be made and it lends itself for the spamming of content by rogue content creators. Manual tagging will depend on online-users for policing and fixing misleading tagging of content. The automated way to tagging is not free of errors either. Even though machines and algorithms are making the decisions how to tag content, the human factor is still present and is much needed to continue refining and tweaking the algorithms to avoid incorrect tagging caused by music playing in the background, poor sound quality, or a thick foreign accent.

Personally, I believe automated and human tagging of video content will co-exist and not become mutually exclusive of each other since there are way too many factors to take into consideration for tagging video content correctly so that it can be properly indexed by web search engines.