Slashdot Mirror

← Back to Stories (view on slashdot.org)

Hardware Suggestions for Linux IVR?

Posted by Cliff on from the as-long-as-0-still-gets-you-an-operator dept.

Lester Hightower writes "I am the CTO of a vertical market application service provider, and we have a couple of applications which could benefit from an interactive voice response (IVR) system. We are an almost-all Linux shop, and most of our production systems are CGI in Perl. I would like to get some feedback and/or recommendations from the Slashdot community on what hardware and software works well, is reliable, easy to maintain, and so forth." Recommendations against hardware, that do not work well for this type of application, are also welcome.

16 comments

  1. Hardware recomendation by Anonymous Coward · · Score: 0

    This phone may not work too well with your system....

  2. Microphones by Anonymous Coward · · Score: 0

    You'll probably need microphones.

    And speakers. Definitely speakers.

  3. "Vertical market"? by Anonymous Coward · · Score: 0

    I don't know much about marketingspeak - what the heck is a "vertical market"?

    1. Re:"Vertical market"? by Anonymous Coward · · Score: 0

      A big market, spread over many floors.

  4. More information required by DRACO- · · Score: 0, Troll

    You neglected to mention weather this IVR is going to be a palm sized device, a laptop, a desktop, a server tower, a beowolf cluster answering telephones. Or is this going to be set up like on startrek shows, "computer do such and such" with all the appropriate beeping tones?

    What the hell is the application other than "an interactive voice response (IVR) system"

    Will this need to sample audio figure out what the hell people are trying to say, and then form a response? Or is it supposed to understand a command such as "AZIZ, LIGHT!" (or more appropriate "Lights please").

    Please either reply with appropriate information or try submitting a fuller question next time. Otherwise you will only get halfassed answers, and by luck you might get a suggestion such as mine, to redefine your question.

    DRACO-

    --
    Consider yourself blessed if you are sneezed on by a dragon and only get wet, it could have been a fireball.
  5. IVR by sigwinch · · Score: 3, Insightful
    IVR is the standard lingo for telephone voice response systems. "Press one for an operator. Press two for an annoying screech. Press three to have F16s bomb your current location. Etc."
    Will this need to sample audio figure out what the hell people are trying to say, and then form a response?
    Hopefully *NOT*. Speech recognition is a pain in the ass. They almost certainly just want to use telephone tones.
    Or is it supposed to understand a command such as "AZIZ, LIGHT!"
    <nothing happens> "I've changed it to 'Illuminate'". ;-)
    Otherwise you will only get halfassed answers,...
    The people who can give useful answers will have understood it. <knocks on wood>
    --

    --
    Kuro5hin.org: where the good times never end. ;-)

    1. Re:IVR by hairyian · · Score: 2, Interesting

      Speech recognition is a pain in the ass. They almost certainly just want to use telephone tones.

      Very large dictionary VR is still a little unpredictable, even when they're trained to one specific person. For limitted dictionary (say, a few dozen words), speaker independent(*) VR is an awful lot more successful. One such project was on slashdot a long time ago and has been in development since - Sphinx.

      As for picking out spoken numbers, back in the days when I was figuring out how to do simple VR I trained myself to the point where I could recognise a number (between 0 and 30 or so) by only looking at a graph of it's waveform - creating a markov model for that purpose would get results as good if not better.

      (* This depends, in Sphinx's case, on the quality of the language model. A language model well suited for recognising mid-west American accents would work poorly on Icelandic... for obvious reasons.)

      Ian Woods

    2. Re:IVR by gmhowell · · Score: 2

      "I've changed it to 'Illuminate'". ;-)

      Damn. All I did was learn how to crochet.

      --
      Jesus was all right but his disciples were thick and ordinary. -John Lennon
  6. Not enterprise level, but... by Vrallis · · Score: 2, Interesting

    I did this once for a non-profit research institute. We used Lucent's online text-to-speech page to produce a library of vocabulary, with a patched copy of sox to convert them to the appropriate type (gmr?). We used USR voice modems, and about 300 lines of Perl code to handle everything from pickup, to producing the menu, building phrases out of the vocabulary, reading responses, and spitting out the resulting data. It was a simple menu system for reading off meteorological data, so at least the vocabulary was fixed and controlled.

    I'd say it worked pretty well, and making changes worked out okay. For our purposes, text-to-speech voices did the job and saved us from the issues of having a proper studio to produce useful sample.

  7. Definitions by Tychoma · · Score: 1

    Most people are going to have some clue as to what IVR is, but you might do well to define "vertical market application service provider".

    --
    Karma: Shitty (mostly due to American moderators)
  8. Solutions by dth · · Score: 4, Informative

    Check out the Bayonne project if you want to create the system yourself. It doesn't support a whole bunch of hardware, and to be honest this will be your biggest problem. Find hardware that is supported by linux and you'll probably find software/libs/apps for free.

    Dialogic cards are probably your best bet, but they're not cheap, and you'll have to be careful in regards to which models work (well) under Linux. Some can be a nightmare. You might want to check out Pika cards too. I haven't used them, but I've heard they do the job.

    If you're looking for a relatively cheap box that does all this for you, take a look at Ostel's IVR100B. Around $2k for a 4 port box.

    1. Re:Solutions by Bazzargh · · Score: 2

      Someone want to mod the parent to this up?? its sitting at 1 but its the only post that actually tries to answer the question posed!!!

    2. Re:Solutions by TokyoJimu · · Score: 3, Informative

      Dialogic cards are probably your best bet

      We have used Dialogic and NMS cards and have never been truly happy with those companies. They seem to promise the world and deliver buggy drivers and it can take months or years to get fixes out of them.

      So we've now started using a totally open-source hardware platform, the Tormenta Card. Linux drivers are provided.

      It costs us just a few hundred dollars for a card that handles a few T1 spans. These cards are kind of like Winmodems, in that they have very little intelligence on them, with all the work being done by the host. But we're running two Ts on a 1.3 GHz machine and we're testing a 4-span card too. One of the card's developers told us he expects to be able to run two or three of their 4-span PCI cards in a dial-processor 1.7 GHz machine.

      Compared to the NMS cards we were buying at $10k a pop, this is downright cheap, and with full open-source we don't have to wait for Dialogic's developers in New Zealand to get around to our problem.

  9. package deals by gCGBD · · Score: 2, Informative

    You can look at commercial packages like that offered by VoiceGenie, or Nuance.

    Bayonne, suggested earlier, has pretty strict licensing on it for commercial use.

    VoiceGenie may be a little young yet with regards to their Linux offering, but it does seem to work ok.

    You can check out LinuxTelephony.org, for more ideas.

    Caveat with the Dialogic hardware:

    • The drivers are closed source and only works on the 2.2.5-14.0 kernel.

    • It is dependent on the archaic LiS (Linux Streams) modules.
      Good Luck trying to install security patches or upgrades.

      Their hardware pricing is also very strange and counterintuitive.
      Often, More Features != Higher Cost.

      There is a new version of the Dialogic drivers coming out, but I've heard they are pretty unstable still, and may not be solid for many more months.
    --

    O=='=++
  10. bayone by h2odragon · · Score: 3, Informative
    Avoid OSTel's hardware; mine has been running for a month and locks up regularly. Cheap 1u system; i suspect it may be the phone card hardware driver.

    Avoid OSTel's application development; they took something like $5,000 to develop an IVR app on that box for their customer, and delivered a non-functional, almost demo after 2 extra months. It was a constant struggle to talk to them, too; they're either far too busy to be taking new customers or they were out to screw us from the start. I really think they're just busy, but i dunno...

    That said, bayonne is pretty cool. With it and a bit of perl, I managed to make the application the customer wanted in 92.5 hours of development time, without much prior experiance applicable to the job. The scripting language is simple, if minimal; it has an embedded Perl (and supposed to be an embedded python too in newer versions) to do anything not possible in the script lang (like math).

  11. Here's a few boards that I know have Linux support by MrJerryNormandinSir · · Score: 3, Informative

    Here are a few boards that I know work and can make a decent IVR.

    Natural Microsystems:
    Alliance Generation® DSP (digital signal processor) boards
    Linux Open Source drivers avaialable!
    http://www.opentelecom.org/news/linuxannouncemen t. html

    I used to use the OLD Watson boards for IVR applications.

    Dialogic has some decent board too!
    http://www.linuxdevices.com/news/NS2735991983.ht ml

    My first commercial IVR appliction was written for "Fall River Celibrates America", this is a yearly celegration that starts around August 8th. I wrote it 10 years ago.. before the days of Linux.
    I used OS/2 at the time. I guess a decent IVR
    that's got Multiline and Faxback capability would
    be a cool killer app for Linux.