« Previous post
Next post »

January 14, 2006

Podzinger Launches, Moves Podcast Search Forward

Michael Arrington

28 comments »

Podzinger, which I wrote about briefly last month along with two other services, officially launched a couple of days ago. Podzinger is based in Cambridge, Massachusetts

In my post last month, admittedly when Podzinger was still deep in development, I could not make it work. I am still having problems with core funtionality.

Podzinger uses speech recognition technology that is supposedly also used by the CIA and other government organizations to turn podcast audio into searchable text. In a few quick searches, it seemed to be returning relevant results. Results can be listed by date or relevancy.

Users can listen to podcasts direclty from the search results, or download the file. Users can also subscribe to the feed or the search results via RSS.

This is where I ran into problems. A “killer” feature of Podzinger is the ability to listen just to the snippet that contains the search term. Links appear within the result that can be clicked on to hear the snippet. I see the links and click on them, but cannot get a single one to play.

This issue is addressed in the FAQs, and it suggest either using Real Player or being patient. However, I am getting actual errors, not just long delays.


When I click on a snippet, why is there sometimes a long delay from the time I hit play until the time I hear the audio playing?
Most media players including Quicktime require that the file be downloaded before play begins. You should not experience this delay with RealPlayer, since it supports something called the Range feature of the HTTP1.1 protocol that makes it possible to begin playing almost immediately at any point in an audio file.

Assuming they get this working properly, it’s a great way to search for content within podcasts and other audio files. A perfect new feature would be the ability for publishers and users to tag additional content within the file and allow others to jump right to that.

Also, Podzinger is obviously fully transcribing the podcasts…although I cannot seem to find the full transcriptions anywhere. If the transcription exists and has a permanent URL, I imagine podcasters would link to it like crazy.

Podzinger is supported by advertising and is also rolling out a cost-per-click product to allow podcasters to sell advertising into transcripts.

  • Sphere It

This entry was posted on Saturday, January 14th, 2006 at 2:38 pm and is filed under Company & Product Profiles. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

Trackbacks/Pings (Trackback URL)

  1. Rise of the PHX

    January 14th, 2006 at 3:26 pm

  2. swylie » Blog Archive » Podcast Chapter Points

    January 14th, 2006 at 4:09 pm

  3. It looks obvious

    January 14th, 2006 at 4:38 pm

  4. Somewhat Frank

    January 22nd, 2006 at 11:54 am

  5. bad credit history

    May 23rd, 2006 at 5:44 pm

  6. Techcrunch » Blog Archive » All The Cool Kids Are Deep Tagging

    October 1st, 2006 at 10:37 pm

  7. All The Cool Kids Are Deep Tagging « BIG Interactive

    October 2nd, 2006 at 2:30 pm

  8. All The Cool Kids Are Deep Tagging » JenIT

    October 11th, 2006 at 8:23 am

  9. Gotuit SceneMaker Lets Users Tag And Cut Online Video at Swiss Podcast Directory and Blog

    December 11th, 2006 at 9:41 pm

  10. Techcrunch » Blog Archive » Gotuit SceneMaker Lets Users Tag And Cut Online Video

    December 11th, 2006 at 10:32 pm

  11. TechCrunch Japanese アーカイブ » Gotuit SceneMakerでオンラインビデオのカットとタグづけ

    December 12th, 2006 at 9:24 am

  12. Start-up.co.nz » Blog Archive » Gotuit SceneMaker Lets Users Tag And Cut Online Video

    December 12th, 2006 at 2:10 pm

  13. Gotuit SceneMaker Lets Users Tag And Cut Online Video

    January 8th, 2007 at 5:35 pm

  14. A Search Engine for Spoken Words in YouTube Videos | Think Artificial

    May 27th, 2007 at 9:28 am

  15. Video Ads: Somebody Needs to Solve this Problem

    July 6th, 2007 at 4:09 pm

  16. techcrunch » Blog Archive » Gotuit SceneMaker Lets Users Tag And Cut Online Video

    August 14th, 2007 at 5:20 am

  17. techcrunch » Blog Archive » All The Cool Kids Are Deep Tagging

    August 14th, 2007 at 5:22 am

Comments

RSS feed for comments on this post.

  1. Ed Dunn

    January 14th, 2006 at 2:59 pm

    I mentioned this in the last posting, but there is a company in the market that can record speech to text and make the content intranet searchable. They are not public facing and generate their revenue from enterprise call centers and government agencies.

    I do not see the need for this in the public sector. It’s just as practical for a podcast to be processed through a desktop speech-to-text program and publish to the web for text-based search engines to index.

    I also realize they are using speech-to-text technology from BBN. What is stopping anyone else from using this technology? Isn’t that like bragging about using Map/GPS data from Navtech?

  2. PXLated

    January 14th, 2006 at 4:01 pm

    Speech to full text would be great. I can read/scan far faster than podcasters can talk.

  3. Dimitar Vesselinov

    January 14th, 2006 at 8:12 pm

    G’Day World #68 - Alex Laats, Podzinger
    http://gdayworld.thepodcastnet.....podzinger/

  4. Rod Boothby

    January 14th, 2006 at 9:10 pm

    This is going to have the most amazing impact on enterprise use of Web Office technology.

    You can already email your blog. Now you can leave a voicemail and have it posted on your blog. Or have the transcript of a conference call posted up.

    I am currently building an enterprise blogging system for my firm. We have 130,000 employees. Although blogging and Wikis have taken off on the public web, broad adoption within the enterprise is going to be a big challenge.

    It is important to find as many ways for people to easily transition from the tools they use today to Web Office tools such as enterprise blogs and enterprise wikis.

  5. Nathan

    January 15th, 2006 at 8:48 am

    I just don’t see how this would have a big impact, especially in enterprise and the much anticipated Web Office.

    Web Office itself is going to have a big climb uphill in terms of adoption. Enterprise culture is slow at adopting new technologies, and they are right to be so, because there are inherent risks in adopting new technologies.

    Where they are even slow to adopt Instant Message (although many employees use it, employers don’t see it that useful).

    Don’t forget about the allmighty SOX (Sarbanes-Oxley). Critical office documents created and stored on server based technology, located god knows where under god knows what security standards.

  6. Kevin Marks

    January 15th, 2006 at 10:42 am

    That response is almost totally false:

    Most media players including Quicktime require that the file be downloaded before play begins.

    False - QuickTime has supported ‘fast start’ playback while downloading since the first browser plugin.

    You should not experience this delay with RealPlayer, since it supports something called the Range feature of the HTTP1.1 protocol that makes it possible to begin playing almost immediately at any point in an audio file.

    QuickTime supports this too - if you make a reference movie pointing partway into an online file, it starts at that offset.

  7. David

    January 15th, 2006 at 12:06 pm

    Alex Laats…the leader of the team who developed the specific speech to text technology that powers Podzinger…has a very interesting background.

    He’s quite a different sort when compared to many Web 2.0 entrepreneurs who bootstrap with a mashup, partner with a technology provider based only on the strength of an idea for a “business development opportunity” or cobble together just enough capital to license an existing technology and then hope to get to either cash flow break even or a VC funding / buyout liquidity event before time runs out.

    I’m curious…does anyone here have an informed perspective on whose speech to text technology is stronger between Podzinger and Podscope…and why?

  8. David

    January 15th, 2006 at 12:08 pm

    Oops!! I forgot to post the link to Mr. Latt’s bio.

    http://www.bbn.com/For_Commerc....._Team.html

    And a little more here…

    http://www.mitforumcambridge.o.....sep02.html

  9. viirya

    January 15th, 2006 at 11:04 pm

    Podzinger’s killer feature - to listen just the snippet that contains terms you search is working and impressively cool. Thanks fo r this introduction.

  10. p

    January 18th, 2006 at 4:37 pm

    Just to answer David’s question about the difference between podzinger and podscope technology. I don’t speak for either company but here is my 2c. I’m also going to trivialize the work that goes into building these systems (i.e you’re not going to code either system over a weekend)

    podzinger is speech to text i.e transcription. so you take a speech signal and figure out the probability a certain set of sounds matches a known word in your dictionary. after you’ve done the conversion, search is regular old text search (so it’s super fast).

    podscope is in speech search which basicaly extracts the phoenemes and creates an index based on this info. whe n you do a search you’re search for a match of sounds. it’s not as fast the podzinger method but is better for stuff like names, places etc which may not exist in text dictionaries.

    my ignorant opinion is that you really want a hybrid model where the frontline search uses the podscope method but then trains the podzinger system with new vocabulary. so over time you can get much better transcription.

    It’s great to see people jumping into this game. It’ll only improve the pace of speech search research. i expect the big boys to make their first moves around the middle of this year.

  11. cengnlxady

    May 21st, 2006 at 9:14 am

    ellvarapxai uubysqye vizjccydkdi

« Previous post
Next post »