Now consumers can use the same technology used by the CIA to find podcasts.
PodZinger officially launched its audio search service on Wednesday, based on speech recognition software that parent company BBN Technologies sells to government agencies.
PodZinger aims to make it easier for users to help determine relevance by displaying the text surround the search term in snippets. Alex Laats, president of BBN’s Delta Division, said that the combination of audio and visual cues let people quickly gauge relevance without having to listen to the audio segment.
“We expose as much as we can in search results, giving multiple snippets of the words around the search term to allow your brain to scan it for relevance,” Laats said. “Relevance is the objective. Speech-to-text is not perfect, but by providing as much result as possible, you can assess whether it’s a relevant or irrelevant result.”
Users can click to listen from the PC, download the podcast or subscribe to it via RSS feed
PodZinger crawls the Web daily, adding new podcasts to its index. It then uses speech-to-text technology to create a text index of the audio that’s searchable by keyword.
PodZinger, which was launched in beta in September, has a new interface that lets users preview a podcast by clicking on the keywords in the search results. It now supports multiple audio players and browsers.
It also has a new business model: pay-per-click advertising. “It’s a model that has been proven to be viable,” Laats said. PodZinger runs both keyword and contextual advertising from Google, which increases the revenue potential: Even if no advertiser has bid on the keywords in a user’s search, Google knows PodZinger’s context is television and multimedia.
PodZinger also is working on deals with large media organizations to use PodZinger search on their sites.
Laats counts on viral distribution by bloggers and podcasters for PodZinger. They can put PodZinger search on their own sites, limiting searches to their own content or opening them up to the Internet. People also can save searches and share them with others either via their blogs or e-mail.
“This is the way blogs have become so viral,” he said, “and PodZinger brings that viral capability to the world of multimedia as well.”
Other multimedia search engines employ text-to-speech to index audio and video content. Blinkx launched podcast search in July 2005, following its release of video search in December 2004. TVEyes provides subscription-based television search to corporate clients; in January 2005, it began providing text-to-speech services to Yahoo .
There are two other methods for indexing video and audio content: using the metadata provided by content creators and indexing the closed captioning text generated during broadcasts.
PodZinger’s parent company, BBN, operates the first metro quantum cryptography network, the first real-time foreign broadcast monitoring system, and has developed the world’s first stereoscopic digital mammography system.
PodZinger’s speech-to-text capability is based on BBN’s AVOKE, developed primarily for government customers in the defense department and intelligence agencies. A DARPA program known as EARS funded research that would help the agencies extract information from broadcast or recorded content, Laats said, and BBN has been a leading recipient of those grants.
“By 2005, BBN had established the state of the art for speech-to-text technology,” Laats said. “My job is to apply that technology to emerging market needs. So we productized AVOKE.”
The company saw its opportunity when Apple announced that iTunes would support podcast retrieval in June 2005. “We said, ‘The ability to convert speech-to-text across a broad range of content is what we do best.'”
