The launch of Podzinger last week (January 13, 2006) as a podcast search engine has improved the state of podcast search on the web. The previous norm in podcast search functionality was to allow visitors to search on categories of podcasts. Podzinger is different in that it converts a podcast into a plain text format using speech recognition research powered by BBN Technologies to transform audio into words. The transcribed words or text is then indexed and searchable. Similar to how Talkr converts text into audio, Podzinger does the reverse process.
In visiting Podzinger and searching on a term like "baseball." A set of results are immediately returned. You can order the results by relevance or by date. Each result gives you several options. You can:
- Download the podcast
- Play the podcast
- Subscribe to the podcast
- Or play portions of the podcast
The last item listed above, which is the ability to play portions of the podcast, is an awesome feature. Basically, you will see a list of sub results under your initial result that displays the occurrences of your search keyword in the actual podcast. In the search result for "baseball" below you can see this feature in action. I feel it allows you to get right to the point and find what you are looking for more quickly then having to listen to the entire file. Some users have found the feature to be slow to load but I assume that is based on the configurations and software of the PC is that is loading and playing the audio clip. An RSS feed of your search query is also presented which could come in extremely handy if you are looking to perform your search for a particular type of podcast on a regular basis.
One potential improvement that was recommended by Michael Arrington of TechCrunch here is to produce a public version of the podcasts in a text format since they are already being converted to text for search purposes. He felt podcasters and bloggers alike would then be able to link to these podcasts as post. I agree this would be a sweet feature but taking it a step further Podzinger could to link the podcast text version back to the original podcast audio feed. FeedBurner began offering FeedFlare in December of 2005. The release of a Podzinger text version of a podcast could interact with FeedBurner's FeedFlare enabled feeds. FeedBurner released the API so that some slick programmer could create additional interactivity like this functionality. The added feature would offer the ability to render the text of at least a of the podcast with the actual audio podcast link. I feel this feature would be extremely helpful in sorting through podcasts and allow the ability to read podcasts as well as listen to them from one feed.
One other feature that might make searching for podcasts easier would be the addition of a tag cloud (like the image to the right) of popular podcast search terms to Podzinger's homepage. I think it would just help spark search ideas to further use of the services of Podzinger.
For additional perspectives from the blogosphere on Podzinger check out these posts on Newsome.org, Audio Activism, Bloggers Blog and Harry Chen Thinks Aloud. Give Podzinger a try and you should be able to leverage keyword search to find more relevant podcasts.
