Microsoft has a better switchboard than Google

Have you ever tried to call someone at Microsoft or Google? I have. I’ve even memorized Microsoft’s number but won’t give it here cause I’d hate for Jeff Sandquist to get a ton of crank calls. Anyway, both Microsoft and Google have switchboard numbers. Today I called someone (or tried to, anyway) at Google and found the experience to be far less useful than when you try to call someone at Microsoft. By the way, at both companies all you need is someone’s name. Yeah, calling Bill Gates or Eric Schmidt probably won’t prove fruitful, but try someone deep down in the bowels of the company and you’ll probably get right through.

Anyway, Microsoft has a really cool voice-controlled system. You dial the number, an electronic voice answers and asks you for the person’s name you’re trying to reach. You SAY it. “Jeff Sandquist” and then it comes back with a recording of Jeff Sandquist so you know you’re going to the right place. “Are you looking for ‘Jeff Sandquist, Channel 9′”? Why yes I am! The call goes right through. If it doesn’t, or you hacked Jeff’s name, it prompts you for a whole lot of other things. “Do you know the email address of the person you are trying to reach?” Etc. Etc.

But compare that to Google’s switchboard. When it answers it immediately prompts you for that person’s extension. That’s not nearly as human as Microsoft’s approach to “Welcome to Microsoft. To reach a specific person just say their name at any time.”

I didn’t know the extension of the person I was trying to reach. So, the voice on the phone says to push 8 to reach someone by their name. I entered in the first few digits of the person’s last name I was trying to reach. It immediately sent me to the voice mail box of someone completely different without prompting the way Microsoft’s system does to make sure I entered the person’s name right.

Why is Microsoft’s system so good? Because they’ve been working on speech stuff for a long time. It’s also why the acquisition of TellMe was so interesting. Imagine if Microsoft just focused its search efforts on voice and gave up the attempt to clone Google.com. Now THAT would be a winning strategy.

Imagine calling a number and saying “Starbucks, San Francisco” and getting routed to the right answer.

Maybe Microsoft’s search engineers should call the switchboard to see just how to beat Google and move the goalposts to a new game?

UPDATE: Richard Sprague, who runs the speech team at Microsoft, says Google can buy its solution if it wants to improve its switchboard.

93 thoughts on “Microsoft has a better switchboard than Google

  1. I have never called either of those companies. But I think computer answering systems awful!!! I would much rather speak to a person, even if the person is completely clueless. By the way thanks for the fast service on the other thing. Benny’s Tavern is going great.

    Like

  2. I have never called either of those companies. But I think computer answering systems awful!!! I would much rather speak to a person, even if the person is completely clueless. By the way thanks for the fast service on the other thing. Benny’s Tavern is going great.

    Like

  3. “Starbucks, San Francisco”

    TellMe Business Search already does this. 😉
    1-800-555-8355 and say “Business Search”

    Like

  4. “Starbucks, San Francisco”

    TellMe Business Search already does this. 😉
    1-800-555-8355 and say “Business Search”

    Like

  5. I think Microsoft is focusing much of its efforts on voice search, and Tellme gets them a good way to the solution.

    Caveat: Google already has a very effective 411 voice search service for businesses. It’s “experimental,” but very functional, and the speech recognition is a breakthrough. Try it out at 1-800-GOOG411 with your “Starbucks, San Francisco” example. It’ll ask you, “What city, and state?” I and other speech industry people have been testing this out vs. Tellme and other VS-DA services, and Google is extremely impressive, and their recognition performance is a genuine breakthrough.

    Not to pour cold water on anything, we’re rooting for Microsoft and love the TellMe acquisition. Some of us, in fact, were begging mgmt to buy TellMe back when we worked there. 😉 I just want to point out that there is a lot of work to be done yet and there may be a performance gap between TellMe and Google.

    Like

  6. I think Microsoft is focusing much of its efforts on voice search, and Tellme gets them a good way to the solution.

    Caveat: Google already has a very effective 411 voice search service for businesses. It’s “experimental,” but very functional, and the speech recognition is a breakthrough. Try it out at 1-800-GOOG411 with your “Starbucks, San Francisco” example. It’ll ask you, “What city, and state?” I and other speech industry people have been testing this out vs. Tellme and other VS-DA services, and Google is extremely impressive, and their recognition performance is a genuine breakthrough.

    Not to pour cold water on anything, we’re rooting for Microsoft and love the TellMe acquisition. Some of us, in fact, were begging mgmt to buy TellMe back when we worked there. 😉 I just want to point out that there is a lot of work to be done yet and there may be a performance gap between TellMe and Google.

    Like

  7. These systems are relatively old, I remember doing a project at Philips Speech Processing (Philips sold this division) Where they had a system based upon National Language Recognition. It was very good, (i’m talking about late 90’s).

    The problem with speech recognition is that it seems every year is the year of speech recognition, but @ the end of the year you’re still left behind wondering what went wrong.

    By the way, they then already had a system that was able to help you book a hotel by just saying something generic like “I’m looking for a hotel on the via apia in Rome” and it worked.

    In the end it’s all about statistics and processing power..

    Like

  8. These systems are relatively old, I remember doing a project at Philips Speech Processing (Philips sold this division) Where they had a system based upon National Language Recognition. It was very good, (i’m talking about late 90’s).

    The problem with speech recognition is that it seems every year is the year of speech recognition, but @ the end of the year you’re still left behind wondering what went wrong.

    By the way, they then already had a system that was able to help you book a hotel by just saying something generic like “I’m looking for a hotel on the via apia in Rome” and it worked.

    In the end it’s all about statistics and processing power..

    Like

  9. The car rental headquarters I worked at deployed a voice response system… It’s GOOD, REAL GOOD. Fewer abandoned calls. Improved look to book. Improved the critical no-show rate. Overall a BIG success.

    Except, many people WOULD rather talk to another human. WHY!? Beats the hell out of me. I’d rather talk to that thing ANY day of the week than one of the operators from God Knows Where who’s motivated to shake me down (agressively I might add) to buy additional insurance I don’t need/want just to line their pocket for the upsell.

    THAT they’ve not taught the voice response system to do as annoyingly as a human. As C3P0 says, “Thank the maker!”

    Like

  10. The car rental headquarters I worked at deployed a voice response system… It’s GOOD, REAL GOOD. Fewer abandoned calls. Improved look to book. Improved the critical no-show rate. Overall a BIG success.

    Except, many people WOULD rather talk to another human. WHY!? Beats the hell out of me. I’d rather talk to that thing ANY day of the week than one of the operators from God Knows Where who’s motivated to shake me down (agressively I might add) to buy additional insurance I don’t need/want just to line their pocket for the upsell.

    THAT they’ve not taught the voice response system to do as annoyingly as a human. As C3P0 says, “Thank the maker!”

    Like

  11. I wanted to cancel my Rhapsody account last night. It wants me to call someone. Case where I’d rather talk to a machine or click somewhere on a webpage.

    Like

  12. I wanted to cancel my Rhapsody account last night. It wants me to call someone. Case where I’d rather talk to a machine or click somewhere on a webpage.

    Like

  13. Sure, a human is the best.

    But, when comparing the 2 systems, Microsoft’s seems more human-oriented. If it works well, then fine.

    I’ve used voice systems that work not as well and it’s quite infuriating. It’s almost as infuriating as the touch tone systems that bring you to the wrong person.

    Like

  14. Sure, a human is the best.

    But, when comparing the 2 systems, Microsoft’s seems more human-oriented. If it works well, then fine.

    I’ve used voice systems that work not as well and it’s quite infuriating. It’s almost as infuriating as the touch tone systems that bring you to the wrong person.

    Like

  15. Microsoft’s switchboard answering system is one of the embarassing things about its public image.

    Jeff is an easy one, and it doesn’t play with with non-american speakers, like perhaps people with a non-regional English accent *cough*.

    I’d say perhaps only 50% of the time it gets the right person at the beginning.

    Try calling people with non-distinctive names.

    “I’m sorry…, did you mean a, b, c”
    “I’m sorry…, can you spell…. .do you know their email alias…”

    If I knew their email alias, I would have probably emailed them.

    I’m quite inclined to talk gibberish just to get to a real person.

    Also, generally with Microsoft, the fact that you can’t easily ask who is responsible for what in a team, or even get a list of teams without having to hunt down people on blogs, forums, articles, videos, etc., is quite bothersome.

    Like

  16. Microsoft’s switchboard answering system is one of the embarassing things about its public image.

    Jeff is an easy one, and it doesn’t play with with non-american speakers, like perhaps people with a non-regional English accent *cough*.

    I’d say perhaps only 50% of the time it gets the right person at the beginning.

    Try calling people with non-distinctive names.

    “I’m sorry…, did you mean a, b, c”
    “I’m sorry…, can you spell…. .do you know their email alias…”

    If I knew their email alias, I would have probably emailed them.

    I’m quite inclined to talk gibberish just to get to a real person.

    Also, generally with Microsoft, the fact that you can’t easily ask who is responsible for what in a team, or even get a list of teams without having to hunt down people on blogs, forums, articles, videos, etc., is quite bothersome.

    Like

  17. Intel has one of these too – the phonetic operator. Work fantastically – I’ve never had it give me the wrong person. If there are similar names (and with 100,000 employees, there are a lot of those), it disambiguates for you. Don’t know how it compares to MS and Google, but it’s pretty awesome.

    Actually, I think in this case, with a very large directory, a system like this beats a human operator. Quicker, I bet, and probably more accurate.

    Like

  18. Intel has one of these too – the phonetic operator. Work fantastically – I’ve never had it give me the wrong person. If there are similar names (and with 100,000 employees, there are a lot of those), it disambiguates for you. Don’t know how it compares to MS and Google, but it’s pretty awesome.

    Actually, I think in this case, with a very large directory, a system like this beats a human operator. Quicker, I bet, and probably more accurate.

    Like

  19. Gosh, I find Microsoft’s system byzantine, trapped in some Cylonish-speak loop (maybe easier if you have worked there for years), Google’s be just telecom-like impossible. Either way, screwed.

    With Microsoft you need an insider friend, tons of blog feeds (and the time to read them) and might have to watch some long boring videos from shaky cam types, to figure out the daily volleyball rotational team shifts, and even still, you might need subscribe to someone like Directions on Microsoft or mow Mary Jo’s lawn (and wash her car and clean house), to even find out who you need. With Google, you don’t have that problem, they simply don’t tell you, and will make you sign an NDA, circle you with a PR Army, sicc’ing some of that infamous Russian muscle on you if you dare breathe a word. Byzantine or paranoia-Russkie-crazy impossible, take your pick.

    Like

  20. Gosh, I find Microsoft’s system byzantine, trapped in some Cylonish-speak loop (maybe easier if you have worked there for years), Google’s be just telecom-like impossible. Either way, screwed.

    With Microsoft you need an insider friend, tons of blog feeds (and the time to read them) and might have to watch some long boring videos from shaky cam types, to figure out the daily volleyball rotational team shifts, and even still, you might need subscribe to someone like Directions on Microsoft or mow Mary Jo’s lawn (and wash her car and clean house), to even find out who you need. With Google, you don’t have that problem, they simply don’t tell you, and will make you sign an NDA, circle you with a PR Army, sicc’ing some of that infamous Russian muscle on you if you dare breathe a word. Byzantine or paranoia-Russkie-crazy impossible, take your pick.

    Like

  21. This isn’t anything special to Microsoft. Voice recognition has been available on PBX systems for years. Google just needs to buy that option. No big deal really.

    But it still isn’t human. My wireless company has the same kind of system and it is the biggest pain in the butt to get anything done with it.

    Like

  22. This isn’t anything special to Microsoft. Voice recognition has been available on PBX systems for years. Google just needs to buy that option. No big deal really.

    But it still isn’t human. My wireless company has the same kind of system and it is the biggest pain in the butt to get anything done with it.

    Like

  23. This was something that we explored a while back (When I worked for a telco) With things like Enum (DNS for phone/contact details it all gets easier.

    One of the “beauties is the telco “remembers” your network and history of contacts and conversations. THere are probably three or four categories to worry about. YOur known contact list (uploaded, and/or grown over experience.) You local Geography – Pizza Hut (location sensitive as well)and general “whole of world access” – i.e. Scoble in San Franscisco.

    Of course, as unified communication develops, the next step is “voice to text translate and deposit in Scobles email inbox, or onto his fax machine…”

    And also, like “Call Robert” “Is that Scoble or Smith or Jones” based on the three having been in ones “last 1000 calls placed or received…”

    Of course, this points to another can of worms on identity management/verification etc. but that’s another discussion.

    Like

  24. This was something that we explored a while back (When I worked for a telco) With things like Enum (DNS for phone/contact details it all gets easier.

    One of the “beauties is the telco “remembers” your network and history of contacts and conversations. THere are probably three or four categories to worry about. YOur known contact list (uploaded, and/or grown over experience.) You local Geography – Pizza Hut (location sensitive as well)and general “whole of world access” – i.e. Scoble in San Franscisco.

    Of course, as unified communication develops, the next step is “voice to text translate and deposit in Scobles email inbox, or onto his fax machine…”

    And also, like “Call Robert” “Is that Scoble or Smith or Jones” based on the three having been in ones “last 1000 calls placed or received…”

    Of course, this points to another can of worms on identity management/verification etc. but that’s another discussion.

    Like

  25. Another cool feature as I recall was at Cambridge Research Labs back in the Digital Equipment Days…

    Everyone’s ID badge had electronic ID that could be proximity sensed. Walk out of your cube into a conference room, or lab, and your phone would forward to wherever you were… (Obvious pros and cos)

    Like

  26. Another cool feature as I recall was at Cambridge Research Labs back in the Digital Equipment Days…

    Everyone’s ID badge had electronic ID that could be proximity sensed. Walk out of your cube into a conference room, or lab, and your phone would forward to wherever you were… (Obvious pros and cos)

    Like

  27. Google’s 411 service is way better, TBH.

    I just messed around with TellMe and the final outcome was (the prompter goes)- “It’s not working out anymore”

    I called again and started making gunfire sounds like “Doosh Doosh” and to my horror, they narrowed down to *Airline search* and I again went “Dishaaaawww” and the prompter goes “Connecting to Japanese Airlines”

    Holy Moly Guacamole!!

    Like

  28. Google’s 411 service is way better, TBH.

    I just messed around with TellMe and the final outcome was (the prompter goes)- “It’s not working out anymore”

    I called again and started making gunfire sounds like “Doosh Doosh” and to my horror, they narrowed down to *Airline search* and I again went “Dishaaaawww” and the prompter goes “Connecting to Japanese Airlines”

    Holy Moly Guacamole!!

    Like

  29. Rather than calling a number and saying “Starbucks, San Francisco”, wouldn’t it be cooler if you could just pick up the phone and say that without dialling? One day!

    Like

  30. Rather than calling a number and saying “Starbucks, San Francisco”, wouldn’t it be cooler if you could just pick up the phone and say that without dialling? One day!

    Like

  31. Rather than calling a number and saying “Starbucks, San Francisco”, wouldn’t it be cooler if you could just pick up the phone and say that without dialling? One day!

    You mean like the phone system was 40-50 years ago?

    Like

  32. Rather than calling a number and saying “Starbucks, San Francisco”, wouldn’t it be cooler if you could just pick up the phone and say that without dialling? One day!

    You mean like the phone system was 40-50 years ago?

    Like

  33. Lots of companies have voice forwarding PBX systems. This isn’t unique to Microsoft.

    Voice recognition has not yet been working with Asterix PBX, but there are some people working on it I believe.

    Like

  34. Lots of companies have voice forwarding PBX systems. This isn’t unique to Microsoft.

    Voice recognition has not yet been working with Asterix PBX, but there are some people working on it I believe.

    Like

  35. you’re a smart cookie Robert… Microsoft should have never let you out of there.

    Like

  36. I think that much like myself, Google just doesn’t want to talk on the phone. I’ve tried calling them a few times, and it’s pretty difficult to get through to anyone at all. Unless you know exactly who you’re calling, and exactly how to do it, you’re not going to have much luck.

    Like

  37. I think that much like myself, Google just doesn’t want to talk on the phone. I’ve tried calling them a few times, and it’s pretty difficult to get through to anyone at all. Unless you know exactly who you’re calling, and exactly how to do it, you’re not going to have much luck.

    Like

  38. http://www.linuxjournal.com/issue/155

    As you know I am a subscriber of Linux Journal.

    You have to consider that using Asterix is far cheaper than using a proprietary based solution. It’s about usability. Sometimes voice automation does not work as well as a button menu system because the recognition does not work well with different accents, ect..

    Asterix will soon have voice recognition as well.

    Why do people still use B&W instead of color?

    It’s all about saving money, and Microsoft will never have as good of a deal on PBX systems as Asterix on Linux.

    http://www.google.com/search?hl=en&safe=off&q=asterix+pbx&btnG=Search&meta=

    Look at the adverts on the right hand column of a google search and you will see complete PBX solutions under Asterix including hardware for $599. MS can never offer that value.

    For the price of 1 Microsoft based PBX, you could pay a development team to integrate the IBM open source voice recognition into Asterix ahead of it’s official release. That’s how valuable it is.

    Like

  39. http://www.linuxjournal.com/issue/155

    As you know I am a subscriber of Linux Journal.

    You have to consider that using Asterix is far cheaper than using a proprietary based solution. It’s about usability. Sometimes voice automation does not work as well as a button menu system because the recognition does not work well with different accents, ect..

    Asterix will soon have voice recognition as well.

    Why do people still use B&W instead of color?

    It’s all about saving money, and Microsoft will never have as good of a deal on PBX systems as Asterix on Linux.

    http://www.google.com/search?hl=en&safe=off&q=asterix+pbx&btnG=Search&meta=

    Look at the adverts on the right hand column of a google search and you will see complete PBX solutions under Asterix including hardware for $599. MS can never offer that value.

    For the price of 1 Microsoft based PBX, you could pay a development team to integrate the IBM open source voice recognition into Asterix ahead of it’s official release. That’s how valuable it is.

    Like

  40. “you could pay a development team to integrate the IBM open source voice recognition into Asterix ahead of it’s official release. That’s how valuable it is.”

    I of course mean that you could do this then resubmit it to the world for free, so nobody else would have to do it again.

    Like

  41. “you could pay a development team to integrate the IBM open source voice recognition into Asterix ahead of it’s official release. That’s how valuable it is.”

    I of course mean that you could do this then resubmit it to the world for free, so nobody else would have to do it again.

    Like

  42. It’s been mentioned above that it is preferrable to talk to a human as opposed to a computer. Although it would be somewhat deceiptful, what about a system so good that you couldn’t tell the difference?

    I’ve thought about a system at various times that would have a human name and would talk to you like a human. Even better, it would you voice print matching and caller id to determine as best as possible who was calling.

    Initially the system would ask some questions to draw the person out and record the information. If would tell the person ficticional information about it’s own “life.” In calls thereafter, it would be sure to ask the user how so-and-so is doing, etc. and fill the user in on more of what was going on in the system’s said “life.” In addition to simulated empathy and friendship, the system would also be able to answer questions that the caller had in the first place!

    The result of such a system (unless the user found out and it mattered to them) would be great customer service providing “quality” one-on-one time with the caller.

    An autmated system could handle the call load without hiring new call center personal. No need to get through the call quickly to move on to the next caller. The caller could be given complete undivided “attention” for as long as the caller wants to talk.

    What do think? Would this be good? If you found out that “person” on the other end of the line you were talking to was a computer would that bother you?

    Like

  43. It’s been mentioned above that it is preferrable to talk to a human as opposed to a computer. Although it would be somewhat deceiptful, what about a system so good that you couldn’t tell the difference?

    I’ve thought about a system at various times that would have a human name and would talk to you like a human. Even better, it would you voice print matching and caller id to determine as best as possible who was calling.

    Initially the system would ask some questions to draw the person out and record the information. If would tell the person ficticional information about it’s own “life.” In calls thereafter, it would be sure to ask the user how so-and-so is doing, etc. and fill the user in on more of what was going on in the system’s said “life.” In addition to simulated empathy and friendship, the system would also be able to answer questions that the caller had in the first place!

    The result of such a system (unless the user found out and it mattered to them) would be great customer service providing “quality” one-on-one time with the caller.

    An autmated system could handle the call load without hiring new call center personal. No need to get through the call quickly to move on to the next caller. The caller could be given complete undivided “attention” for as long as the caller wants to talk.

    What do think? Would this be good? If you found out that “person” on the other end of the line you were talking to was a computer would that bother you?

    Like

  44. @32 “In addition to simulated empathy and friendship”

    Unless you are an extremely empty souless individual, I am 1000% positive that this would piss people off more than anything else.

    Even a PBX system is supposed to act as a tool for both the caller and callee to efficiently accomplish a task. Sometimes that can be most efficient with a call menu, because it doesn’t take accents and dialects into account.

    Something REALLY cool would be to code an Asterix add on that would tie….

    … nevermind, creating a patent on this now.
    Not to stifle GPL, just to make sure that istanbul has to pay through the nose to use it. It’s the American way. Yay. I’m chalk full of patents now.

    Like

  45. @32 “In addition to simulated empathy and friendship”

    Unless you are an extremely empty souless individual, I am 1000% positive that this would piss people off more than anything else.

    Even a PBX system is supposed to act as a tool for both the caller and callee to efficiently accomplish a task. Sometimes that can be most efficient with a call menu, because it doesn’t take accents and dialects into account.

    Something REALLY cool would be to code an Asterix add on that would tie….

    … nevermind, creating a patent on this now.
    Not to stifle GPL, just to make sure that istanbul has to pay through the nose to use it. It’s the American way. Yay. I’m chalk full of patents now.

    Like

  46. I’ve never called Google but I’ve call Microsoft’s system hundreds of times over the years, and it does work very well. They have tweaked it over time. I don’t need to speak to a human if speaking to a computer system expedites my call faster.

    Like

  47. I’ve never called Google but I’ve call Microsoft’s system hundreds of times over the years, and it does work very well. They have tweaked it over time. I don’t need to speak to a human if speaking to a computer system expedites my call faster.

    Like

  48. I’ve never called Google but I’ve call Microsoft’s system hundreds of times over the years, and it does work very well. They have tweaked it over time. I don’t need to speak to a human if speaking to a computer system expedites my call faster.

    It can, right up until it hits a situation outside of its programming. An automated system is not better than a well-trained human. That’s a myth. It has exactly two advantages: it can handle more calls than a single person, and it is cheaper over time. Note the last one is the driving force here. But a human is still going to be far more flexible, and for this kind of situation, that’s an important feature.

    Like

  49. I’ve never called Google but I’ve call Microsoft’s system hundreds of times over the years, and it does work very well. They have tweaked it over time. I don’t need to speak to a human if speaking to a computer system expedites my call faster.

    It can, right up until it hits a situation outside of its programming. An automated system is not better than a well-trained human. That’s a myth. It has exactly two advantages: it can handle more calls than a single person, and it is cheaper over time. Note the last one is the driving force here. But a human is still going to be far more flexible, and for this kind of situation, that’s an important feature.

    Like

  50. “An automated system is not better than a well-trained human”

    Totally incorrect except in a few corner cases(unless you are talking specifically about phone call routing).

    ex – EMail servers. How would you acheive instant email delivery with manual processing?

    Like

  51. “An automated system is not better than a well-trained human”

    Totally incorrect except in a few corner cases(unless you are talking specifically about phone call routing).

    ex – EMail servers. How would you acheive instant email delivery with manual processing?

    Like

  52. #38 Well, I don’t know about you, but I wouldn’t want automated medical diagnosis. I wouldn’t want automated financial advice. I wouldn’t want automated legal advice. I’d hardly call those corner cases.

    and that automated email server does a great job of delivering viagra ads I don’t ask for.

    Like

  53. #38 Well, I don’t know about you, but I wouldn’t want automated medical diagnosis. I wouldn’t want automated financial advice. I wouldn’t want automated legal advice. I’d hardly call those corner cases.

    and that automated email server does a great job of delivering viagra ads I don’t ask for.

    Like

  54. “and that automated email server does a great job of delivering viagra ads I don’t ask for”

    So does the mailman. And he is slower too…

    Like

  55. “and that automated email server does a great job of delivering viagra ads I don’t ask for”

    So does the mailman. And he is slower too…

    Like

  56. I want to make this post because I got lots of email in response to writing about IBM’s open source voice recognition technology. Firstly, here is a link to an overview: http://www.eclipse.org/vtp/ Most of this speech recognition platform was donated by IBM corporation, as they also donated a lot of the rest of the apache tomcat code. The tricky part is that Asterix is not a java application, it is C++, so to use vtp with asterix, you will have to integrate it as JNI or some other shell call/return, at a lower level than you would regularly use it at with the Eclipse framework. There are also premade solutions such as: http://products.i6net.com for VoiceXML with Asterix. And there are other recognition libraries: http://cmusphinx.sourceforge.net/html/cmusphinx.php All in all if you have a strong development team, they can most likely whip this up for you. You could also code it and keep your product separate from the Asterix download and sell it as a proprietary patch, and then you could really start making some money downselling Microsoft’s solution. Even if you code it and sell it for half price, that’s still going to be far more attractive than paying full price.

    Like

  57. I want to make this post because I got lots of email in response to writing about IBM’s open source voice recognition technology. Firstly, here is a link to an overview: http://www.eclipse.org/vtp/ Most of this speech recognition platform was donated by IBM corporation, as they also donated a lot of the rest of the apache tomcat code. The tricky part is that Asterix is not a java application, it is C++, so to use vtp with asterix, you will have to integrate it as JNI or some other shell call/return, at a lower level than you would regularly use it at with the Eclipse framework. There are also premade solutions such as: http://products.i6net.com for VoiceXML with Asterix. And there are other recognition libraries: http://cmusphinx.sourceforge.net/html/cmusphinx.php All in all if you have a strong development team, they can most likely whip this up for you. You could also code it and keep your product separate from the Asterix download and sell it as a proprietary patch, and then you could really start making some money downselling Microsoft’s solution. Even if you code it and sell it for half price, that’s still going to be far more attractive than paying full price.

    Like

  58. @38-40

    “ex – EMail servers. How would you acheive instant email delivery with manual processing?”

    Email doesn’t try to solve your complex problem. It merely routes a message to an end point server, which delivers to the local recipient.

    Automated phone systems, no matter how good the tree structure for Q&A can be very inefficient and hard to use. Voice or no voice.

    Having “dumb workers”, IE, workers that are trained to read FAQ messages out of a database, often times are no better than the automated systems.

    If you are using PBX for private branch exchange or routing, then fine, otherwise, it can easily be a losing system all together. People are only willing to take so much before they go to your competitor.

    The promise of dumb labor and saving money by means of light AI and automation is deceptive. In the end when your customers leave, you may find out it’s not as good of a deal as you thought.

    Like

  59. @38-40

    “ex – EMail servers. How would you acheive instant email delivery with manual processing?”

    Email doesn’t try to solve your complex problem. It merely routes a message to an end point server, which delivers to the local recipient.

    Automated phone systems, no matter how good the tree structure for Q&A can be very inefficient and hard to use. Voice or no voice.

    Having “dumb workers”, IE, workers that are trained to read FAQ messages out of a database, often times are no better than the automated systems.

    If you are using PBX for private branch exchange or routing, then fine, otherwise, it can easily be a losing system all together. People are only willing to take so much before they go to your competitor.

    The promise of dumb labor and saving money by means of light AI and automation is deceptive. In the end when your customers leave, you may find out it’s not as good of a deal as you thought.

    Like

  60. “An automated system is not better than a well-trained human”

    Totally incorrect except in a few corner cases(unless you are talking specifically about phone call routing).

    Or someone left and didn’t tell you. Or they’re on vacation and didn’t set their voicemail, or any one of a dozen issues that happen daily, like unexpected meetings, et al, that require some kind of human notification for the automated system to work.

    ex – EMail servers. How would you acheive instant email delivery with manual processing?

    How do you suddenly try to compare email to phone calls? As a strawman, that sucks.

    Like

  61. “An automated system is not better than a well-trained human”

    Totally incorrect except in a few corner cases(unless you are talking specifically about phone call routing).

    Or someone left and didn’t tell you. Or they’re on vacation and didn’t set their voicemail, or any one of a dozen issues that happen daily, like unexpected meetings, et al, that require some kind of human notification for the automated system to work.

    ex – EMail servers. How would you acheive instant email delivery with manual processing?

    How do you suddenly try to compare email to phone calls? As a strawman, that sucks.

    Like

  62. I hate those voice recognition systems.

    The other day, I was trying to schedule a package picked up via DHL’s voice recognition system. It asked whether I was commercial or residential (commercial), then identified 5 incorrect addresses for the phone number it automatically identified as my own. As it read each street address off, I would say “yes” or “no, tell me the next one” or words to that effect. By the time I got through the 5 addresses (and realized none of them were correct!), I had spent probably 5x the amount of time it would have taken to complete the task by talking to a damn person. Needless to say, I swore at the machine and dialed 00 to get to a real, live human being who scheduled my package to be picked up the very same day – and it took under a minute!

    Sometimes a technology solution is worse than the alternative, “old way” of doing things or conducting business. It certainly seems like Microsoft has a culture of pushing technology ahead of usable solutions.

    Like

  63. I hate those voice recognition systems.

    The other day, I was trying to schedule a package picked up via DHL’s voice recognition system. It asked whether I was commercial or residential (commercial), then identified 5 incorrect addresses for the phone number it automatically identified as my own. As it read each street address off, I would say “yes” or “no, tell me the next one” or words to that effect. By the time I got through the 5 addresses (and realized none of them were correct!), I had spent probably 5x the amount of time it would have taken to complete the task by talking to a damn person. Needless to say, I swore at the machine and dialed 00 to get to a real, live human being who scheduled my package to be picked up the very same day – and it took under a minute!

    Sometimes a technology solution is worse than the alternative, “old way” of doing things or conducting business. It certainly seems like Microsoft has a culture of pushing technology ahead of usable solutions.

    Like

  64. Microsoft did not develop the speech system on thier PBX. They use Nortel for all of thier speech recognition. Actually most of this software was developed canadian government. Just thought I would set the record strait

    Like

  65. Microsoft did not develop the speech system on thier PBX. They use Nortel for all of thier speech recognition. Actually most of this software was developed canadian government. Just thought I would set the record strait

    Like

Comments are closed.