Ajay Juneja

Dialogue makes communication work

  • Sphinx 3.4 coming out soon...

    So I just attended a talk by Arthur Chan, formerly of SpeechWorks (now ScanSoft), who is now staff at Carnegie Mellon, and he's working on a new version of the Sphinx 3 recognizer (Sphinx 3.4). It's a very sophisticated approach using a tree-based lexicon (which improves speed, and marginally reduces accuracy of words that are slurred or cut off, since the language model cannot boost confidence scores with this approach), and a continuous HMM (Hidden Markov Model).

  • Friend for hire...

    So I have a really good friend here at CMU, great guy, and needs an internship this summer. Please ping me via this blog if you are hiring for the summer and could use a CMU student who:

    - Has a 4.0 in our computer science dept. (Only like 4 students a year can pull that off here).
    - Is teaching a class on C# right now.
    - Helped write a C++.net book published by Deitel and Deitel last summer.
    - Has a great personality, super creative, awesome guy, and super enthusiastic.

    I'm not posting his resume so he doesn't get spammed, but ping me if you've got anything interesting :)
    I'll try and help him through my own contacts at IBM and elsewhere, too.
    Thanks!

  • The Passion of the Christ -- Great Movie... but what does the Baby with the Devil towards the end of the movie symbolize?

    I just came back from watching this much talked about movie -- it's a fabulous movie, I think. It's a very good portrayal of the "real story" (or at least as real as the story we have documentation on) and shows each of the characters well. The messages and motifs are left in tact, and it appears that Mel Gibson still had some artistic interpretation inserted in.

    One thing confused me -- I don't understand what the Devil carrying the sort of old, evilish baby at the end symbolizes? Anyone got any takes on that?

    Great movie, go see it, and especially see the movie if you have read the old and new testaments.

  • Ah what a joyous February

    Guten Morgen! This month and semester is off to a great start! Parallel programming is a really cool class, as is getting access to a large cluster of IBM Power5's.

    Hopefully I'll be able to incorporate some of the techniques I learn from this class into future programs I write.

    Also, this month, aside from being really frigid, has been very good to me! I get to go to two joyous weddings in March, and I may also get the chance to meet Ken Rehor, one of the original creators of VoiceXML.

    I also had a few job interviews lately, one with Amazon.com (which had quite a comedic end to it -- I think the recruiter was probably laughing when she got my email after the interview -- needless to say, I don't want to work there), one with BeVocal (which went very well and I liked a lot), and I may have one with Scansoft soon, too. Looks like the jobs are out there -- don't lose faith for all of you who are looking :) And if anyone out there needs help, I have contacts at lots of companies, especially on the west coast.

    I am also getting plans ready for a trip to Germany after school ends in March! Should be the perfect respite before starting work. Not to say that I don't have some business planned to do there soon.

    At school, we had a fantastic talk about RFID from one of the Vice Presidents of SAP Labs' research center in Palo Alto -- I'm going to try and schedule an in-person chat with them over spring break.

    I do believe a Mr. Bill Gates is coming to CMU in a couple weeks -- will be quite interesting to hear what's on his mind. Maybe I can find a plush penguin to give him as a thank-you-for-visiting.

    I'll be out in San Francisco from March 6th - 13th,. and would love to chat with any of the bloggers out in the bay area -- especially would love to chat with the now famous Mr. Robert Scoble. Robert, would you care to do lunch or dinner that week and have a nice chat about the future of lots of technologies, from speech to RFID to user interface design?

    There's also a lot of behind the scenes things going on with me in the dialogue systems world :)

    The one bit to share is a great way of explaining what I'm working on (Special thanks to my dear friend Lauren at Pitt for help on coming up with this):

    Think of the speech recognizer as your ears. Think of the text to speech engine as your brain. Think of the dialogue system as your brain. I work on the brain :)

  • The tale of the $40 DVD player -- or why I hate the CHEAP! CHEAP! CHEAP! mentality of Americans today

    So my brother decided to get one of those cheapo-dvd players for $40. I was given the task of setting this up for my mom, and my first thought was "what an utter piece of junk this thing is. I'll be surprised it it lasts 6 months."

    To no surprise, it failed within 2 hours of use. I told my brother to return it and PLEASE let me go with him to pick out a better model -- something that was at least $100 with a brand name, since he insisted on being so cheap! He declined and came back with another $40 player with the hopes that it wouldn't die so quickly. We'll see how long this one lasts...

    I looked at the specs -- audio signal to noise ratio of 60dB -- how pathetic -- analog cassettes have a better s/n ratio than that! I truely wonder exactly how cheap the D/A converters are in that thing. I see DVD-Rom drives going for like $16 these days.. the decoder card I can see being $10 -- I don't think they can meet the standard consumer electronics price point of selling at 4 times manufacturing costs. If they indeed do with these no-name chinese players, then how they manufacture them for $10 is beyond me.

    I do in fact personally own the player that started this cheap! cheap! cheap! craze -- the Apex AD600A -- which I bought solely for the novelty of playing MP3-CD's in 2000, when it was the only player that could do so. While it still works today, I regret the purchase in a way, because it was just a sign of things to come. I do fully intend to purchase a Sony 999ES or the earlier 9000ES one day in the next two years. Then I'd have a DVD player on par with the rest of my audio equipment, but since I am more of a music person than a video or movie buff. Nor do I have the space or money in college for a really nice TV.

    I suppose my point is, why do people WANT to buy this crap, solely on the fact that it's cheap? What happened to people willingly spending money on things that LAST, rather than things that are CHEAP! As I said, standard fare in the consumer electronics world is sell at 4 times manufacturing costs. If they somehow got the costs to $10 in china for that player, it's no wonder that the parts inside do indeed, suck. And the quality control does too. And under what labor conditions was this thing manufactured to meet that price point and still make a profit?

    I look at a lot of my stereo equipment, a lot of which is 10-year old Rotel gear that is solid as a brick (and yes, made in china), some 10-year old Sony ES components (Made in Japan), and I have a 20-year old Tape deck (Technics 3 head) which retailed for $400 when it was new. All of it still works, and works beautifully. I bet you it will all still be working in 10 more years. But I spent the time to earn the extra money and spend it on *quality.*

    I look at those $500 computers in the same way. Sure it's cheap, but what was the cost to get there? Memory shared with the Video card? CPU with high clockspeed but castrated L2 cache? Hard drive that will fail in one year? I still have some $2000-$3000 machines that are 6 years old and solid as a brick. I have a working Apple II+ still.

    C'mon people, demand quality again. And Quality costs MONEY. I had a discussion with a GM engineer a couple months ago who was shocked that Toyota (Lexus) switched steel suppliers on a whim for their RX330 manufacturered in Canada, because the hood panel gap was something like .05" too much, and the Japanese engineer decided it was the steel they were using that was the problem, so they are now shipping steel in from Japan rather than using a local supplier in Canada.

    My response to him was, "Well you do that 128 times, and you end up with a better car, not just a better part."

  • Time for an update!

    Whew! Finals are finally finished, so I finally get to blog again. Recently, I upgraded my soundcard to a Terratec DMX6 Fire 24/96 -- my old soundblaster live wasn't cutting it anymore. Amazing what a difference a sound card made for audio makes, as compared to one for gaming! One day I'll be able to afford one of those Lynx Two's

    Ok, before anyone else says I'm going overboard, I'm an audiophile, and well, my work has to do with speech recognition and dialogue systems. So, sound is my life so to speak. I also got an excellent new microphone, the Andrea ANC-750 which also is helping my demos.

    On the dialogue systems front, there are many interesting things happening, though I'm not saying what just yet. All I'll say is watch this space closely.... there should be a new build of ariadne by mid-January, I may not make a public release this time, though.

    Lots of trips coming up for me, too:
    Winter Break: New York City, Jan 1st - 10th (philadelphia the other part of winter break)
    Spring Break: San Francisco, March 6th - 13th
    post graduation: Germany (probably Freiburg), June 1st - July 29th