Searching For Sound
from the limitations dept
Many people have pointed out that search engines (yes, mainly Google) are now the “front end to the internet.” However, how does that work when the internet is increasingly not just about text? Especially as broadband catches on around the world, more and more content is audio and visual content. Both new and old search engines are now working on better ways to sort through that content – using metadata and speech recognition to understand what’s being said. The article uses NPR as the main example, describing how they use voice recognition technology to create immediate transcripts of their audio, which are completely searchable. They admit that these transcripts are later replaced by “more accurate” human written transcripts, but that the automated ones work well enough. The article also focuses on StreamSage, which seems to be one of the more advanced tools. It uses voice recognition to transcribe audio – but also tries to add in some contextual analysis to create an automated “table of contents” for the file, so searching through it is much easier.