| 7:19 pm on Jan 22, 2004 (gmt 0)|
I've been expecting that development, it will definitely have some appeal for the less computer literate.
Could be big.
| 7:41 pm on Jan 22, 2004 (gmt 0)|
Remember when google labs had a voice related search labs1.google.com/gvs.html? You still had to be in front of the computer in order to see the results.
The coolest feature of this would be if it was completely independent of the computer, like they are suggesting. So long 411
|Off-line search capabilities allowing you to conduct web searches while away from your PC |
I am looking forward to trying this out and seeing where it takes me.
| 9:34 pm on Jan 22, 2004 (gmt 0)|
This could be cool - reminds me, I need to get one of those WAP mobile phones.
Though...it'll be interesting to see if any of these newcomers actually get some search volume. Will it have it's own DB, or will it be using a 3rd party index?
| 10:37 pm on Jan 22, 2004 (gmt 0)|
|Remember when google labs had a voice related search labs1.google.com/gvs.html? |
They still have...
dunno if it works though, as i am not in the US and this makes me reluctant to give it a go (international call rates being what they are, etc)
| 10:56 pm on Jan 22, 2004 (gmt 0)|
Time to (as I was recommending for years now, some buddies got nice pocket money on that tip) scale up your software recognition stock portfolio...
It's coming, not violently, not tomorrow but it's coming.
For the mentioned site though: needs much cleaner design, but that's just me...
| 7:38 am on Jan 23, 2004 (gmt 0)|
|needs much cleaner design, but that's just me... |
No, it's not (just you). I think that the design is kind of a mess, with the layer the ads are on overlapping some of the listings on my browser too.
I'm wondering how the voice recognition might interpret some of the searches. Makes me think of the punch lines of many jokes. I could be wrong... and I hope I am... but this feels like a late-90s kind of idea to me. Can't imagine that the accuracy needed for such a wide vocabulary could be accomplished without training the software.
I have many tee-shirts from companies that started like this that aren't around any more.
| 8:51 am on Jan 23, 2004 (gmt 0)|
And: why would a speach recognition service need an entrance portal anyway, let alone any other content?
An IP should be enough and the rest is done via hardware that controls the software, algorithm at the target IP, then spitting out clean, pure target offers as SERs - again voice activated too - you get to pick with another voice command.
Late 90's idea - I totally agree but only in terms of the execution. The idea as such is not new and will be flying, but it needs to get as close as possible to real human behavior.
Read Nicholas Negroponte about the issue, he's been right with his digitalization theories all along. If not MIT the development will be driven by the real big players in the market on the elements that are there already. All it needs is fine-tuning and linking them up. Kindoflike:
IBMoogleSoftNuance type cooperation. Interesting who's gonna pull the trigger first on that direction.
| 10:49 am on Jan 24, 2004 (gmt 0)|
|needs much cleaner design, but that's just me... |
I concur - it's not just you. It is appalling.
I'm sure with the coverage they're getting at the moment they'll get in a designer to re-do it properly. But the face-lift will be easy, the important thing is whether or not the guts actually work.
| 4:45 pm on Jan 26, 2004 (gmt 0)|
I've been in speech reco./wireless speech reco. for many, many years. If this works, I'll be the 1st to take my hat off and yell "Hooray".
Forgive my "stumping" and parade raining...
There are a littany of variables that such portals will have to overcome before this begins to work well..
Google's experiment was a dismal failure, for the most part excepting calls from good 'ol analog-type phones and 900mHz
Most cellular networks have gone from CDMA/TDMA flavors to GSM; GSM requires pretty healthy compression to work well in massive volumes. Depending upon which wireless carrier one uses, this compression can range from fairly nice to "clippy". The caveat here is that these carriers are quickly dicovering acceptable bitrates can't be had w/o trimming down sampling rates ergo the frequency respone (Nyquist) is slimmed accordingly down to good 'ol "telephone" bandwidth - carriers are also discovering channel capacity is so stressed during peak hours that even this presents problems.
Add in to the mix that most carriers are using some form of IP to process Johhny Lunchbucket's demand for MSM, camera phones and WAP the QoS for "voice" becomes secondary.. "Can you hear me now?" becomes a real challenge.
Even so, there lies another pitfall: Voice "Gateways".
If a search portal uses voice in, the speech must travel through a gateway to begin interfacing with the recognizer, and recognizers only work with native digital; so now we have a variety of compression algo's (there's no standard in sight) the gateway must interpret/uncompress/convert/re-sample to hand the user's speech to the core recognizer. The server-side horsepower to do this requires either massive amounts of money or a QoS trade-off.. And I'm curious if this portal has the liquidity to do this right.
Sorry for the diatribe.. But I'm still the Doubting Thomas.
|The Subtle Knife|
| 4:08 pm on Jan 28, 2004 (gmt 0)|
This doesn't look like a search engine to me!
Does anyone know where the results are coming from?
| 2:34 am on Jan 29, 2004 (gmt 0)|
results are definitely coming from inktomi. But there seems to be a kind of addition of results from somewhere else. All my high first page listings in Ink are in limo but lower on the page. About a 5 to 10 position drop acrossed the board. DEF ink. tho
| 2:44 am on Jan 29, 2004 (gmt 0)|
Yeah, they will have to cut down on that page size! Took almost 10 seconds to load everything on a DSL Line.