Thursday, March 31, 2005

The Hobson & Holtz Report - Podcast #20: March 31, 2005

Show notes for March 31, 2005

Content summary: Listeners’ comments: audio comments and a few written ones, too, on VNRs, multimedia search, and new blogs; Elizabeth Albrycht’s advice column; Nielsen-Norman’s top 10 intranets; Microsoft’s newly named stripped-down Windows XP for Europe; new bloggers at GM’s Fastlane blog; Yahoo 360; faux blogs; and monitoring employees’ onliine behavior.

Show notes for March 31, 2005

Download MP3 podcast

Welcome to For Immediate Release: The Hobson & Holtz Report, a 63-minute conversation recorded live from Los Angeles, California, USA, and the UK.

Download the file here (MP3, 28MB), or sign up for the RSS feed to get it and future shows automatically. (For automatic synchronization with your iPod or other digital player, you’ll also need software such as the FeedDemon RSS aggregator, or the free ipodder or DopplerRadio).

In this edition:

Intro:

  • 00:29 Shel on what’s in this week’s show; how to give your feedback; show notes
  • 02:01 Comments from the last show

Features:

  • 28:36 Elizabeth Albrycht’s new blog advice column
  • 29:53 Nielsen-Norman’s 2005 top 10 intranet listing
  • 35:29 Microsoft and the EU settle on a new name for Windows XP in Europe
  • 35:29 New executives are blogging to the General Motors Fastlane blog
  • 39:50 Yahoo launches Yahoo 360
  • 44:34 Faux blogs
  • 47:49 Employee monitoring

Outro:

  • 58:07 Comments and show notes reminder

Music:

If you have comments or questions about this show, or suggestions for our future shows, email us at .(JavaScript must be enabled to view this email address), or call the Comment Line at +1 206 984 0931. You can email your comments, questions and suggestions as MP3 file attachments, if you wish (max. 5Mb attachment, please!). We’ll be happy to see how we can include your audio contribution in a show.

So, until Thursday March 31…

Click “More” to read Carl Rogat’s e-mail about multimedia search.

In your March 28th podcast # 19 you mentioned that it would be quite useful if there was a way to search audio files for specific occupancies or words or subjects without having to depend on the rather meager data included in metatags.  Several companies are now marketing search products that use new approaches to accomplish this task.

E-Content Institute (http://www.econtentinstitute.org) states in their May/June 2004 resource section that:

“Nexidia’s ( http://www.nexidia.com ) NEXminer product processes unstructured multi-language audio and video content so that users can search actual audio or video files phonetically for any set of sounds. Rather than using speech recognition technology, which tries to match your words against a spoken lexicon, or transcription technology where you search a full text index, they have created a way for you to find what you’re looking for simply by spelling it out fo-net-ik-lee (or as it sounds, not necessarily how it’s spelled). Which, of course, comes in handy when a proper name or place or thing can be spelled several ways and when the content you’re searching comes in a variety of languages (for example, international news clips). Your search terms get resolved down to phonemes (the units of sound at the heart of any language) and the search engine returns all spoken references that match. Without having to transcribe or index, the process of getting new content into the system is fast and so is performance—over 30 hours of material can be searched in less than one second. What about homonyms (words that sound the same but are spelled differently)? That’s a problem. So are text mining strategies like pattern searching for related concepts (remember—there’s no text index). The system currently supports American English, Latin American Spanish, Modern Standard Arabic and Iraqi. Nexidia has partnered with others in the search market, like Convera http://www.convera.com  , to provide an audio search capability where users can search for spoken words without having to convert the audio to text.‰

Source: http://www.econtentinstitute.org/issues/ISarticle.asp?id=149459&story_id=8987132500&issue=05012004&PC=&RType= )

Nexidia (an outgrowth of a company formerly known as Fast-Talk)) has also partnered with: CACI International Inc. (http://www.caci.com), IBM (http://www.ibm.com Dicta,phone (http://www.dictaphone.com), and SAIC (http://www.saic.com)

Source: http://www.nexidia.com/partners2.asp

Blinkx (http://www.blinkx.com) a new search engine for audio video content with a focus on Television, also uses a phoneme search technology, possibly Convera‚s platform.  An Article by By Dinesh C. Sharma on Dec 16, 2004 notes that Blinx uses:

‰Ground breaking automatic transcription technology, which transcribes content straight from the cable box on the fly or from video already stored on the Web, together with advanced phonetic matching speech recognition technology, automate the process of searching TV clips for the first time,” Blinkx founder Suranga Chandratillake said in a statement.‰

Source: http://news.com.com/Blinkx+unveils+video+search+engine/2100-1032_3-5493660.html

In Blinkx‚s own words” The system technology is based around decomposing digitized speech into its phonetic constructs. The phonetic sequence is then analyzed in conjunction with acoustic model and statistical improbabilities to calculate which is the most probable sequence of hence words and utterances.”

In addition to Convera, another company that has a phoneme search engine is Aurix (http://www.aurix.com)  Their site states:

„Aurix miner enables fast searching of large audio or video archives for specific spoken words of phrases.  Aurix miner meets the need of trawling through large audio or video archives by a two-stage approach. The first stage is a one pass phonetic recognition of the audio stream using Aurix asr to determine a phonetic representation of the data. The phonetic representation accurately encodes not just the most likely sequence of phonemes but also less likely phoneme sequences. This enables the miner to smooth over phonetic recognition errors and to retrieve a higher percentage of relevant material in the audio archive. The phonetic representation is converted into an ‘index’ for efficient searching.

In the second stage the user-specified search keywords are converted to phonetic sequences by looking them up in a phonetic dictionary, or by using letter-to-sound rules. A fast search algorithm then searches through all the indices in the archive for the phonetic sequences and returns a list of ‘hits’ sorted by audio by confidence. „

Source: http://www.aurix.com/dynamic/technology/developers/Technology.aspx?story=196

An interesting summary of phoneme development by Leavitt Communications (http://www.leavcom.com)  can be read at http://www.leavcom.com/ieee_oct02.htm .

Another interesting articles Article appeared on Panda Central‚s website out of Oslo, Norway.  It discussed Blinkx and it’s phoneme search engine. Check out http://www.pandia.com/sw-2004/65-blinkx.html .

Finally, an excellent, article that appeared in the December 13,2002 issue of InfoWorld, though dated, it is worth checking out at http://www.infoworld.com/articles/ap/xml/02/12/16/021216apfastalk.html.

I enjoy your excellent podcasts and find your meticulous show notes of tremendous value.  I am sure that it takes a great deal of time and effort to produce your shows and hope that the two of you pace yourselves and do not burn out too quickly.  If you were to drop off the podcast scene it would be a real loss to what I am sure is a growing audience.

Carl S. Rogat, Pres.
Notes4Review.com

Posted by shel on 03/31 at 02:11 PM
Show Notes • (2) CommentsPermalink