But you can read a tracking number off a slip of paper safely while driving at 80 and talking on a cell phone? Interesting definition of “safely”.
Apparently the voice recognition systems have been getting some bad feedback as there’s at least one ad on TV making fun of that and pushing their “easy access to customer support” feature.
Sometimes I can get this by just saying something unintelligible (or even one of their choices but they don’t like how I say it). “Ok, I heard you say you want to speak to someone. Is that correct?” YES.
I HATE the system we force our callers to use at my place of work, where we book travel for people who want to redeem points on their credit cards. The callers are prompted for their account information and then the cities and dates on which they wish to fly. Less than half of the calls actually come through with the correct account information, and of those, only about half have their flight information fully entered in. People are already pissed that they spent several minutes wading through several voice prompts, only to have to give this information all over again to a human operator.
Most of the systems I have used will still recognize numeric input from any keypad. If you are asked to give your social security number, for example, it’s likely you could enter the number manually. The systems I’ve used are backwards compatable in that way, even if they don’t explicitly state that they are.
That’s true, but I’ve had problems where any noise would make the system think I tried to speak to it. Recent example:
Phone system: “Please speak your account number.”
Me: <start entering number on keypad>
My dog: “Woof!”
Phone system: “Sorry, that’s not a valid account number. Please enter again…”
/repeat 5 times
Seriously has happened to me with the dog, also the cockatiel. Other culprits, I lose train of thought and say “um” of “ah” which the system doesn’t like. I’ve had my raspy stubbly chin rub across the mouthpiece of the phone and mess up the system also. And good luck trying to do this in a noisy office with phones and conversations going on all around.
These things have been the goal of the telcos for years. I remember over a decade ago calling into an experimental system at Bell Labs Research to help train it.
They’ve worked for me pretty well. If customer service is option 7 on a menu, it’s much easier remembering to say Customer Service than waiting until choice 7 comes up. I’ve used them for directory lookup when calling a company - much better than spelling out a name on a keypad.
I don’t have any accent (coming from New York 
) but my speech is not the best, and it never has a problem understanding me. And they’ll get better.
Isn’t anyone going to say “Hang up and drive” to this fellow? No? Allow me to be the first!
I hate those voice recognition systems, they seem like little more than a novelty at this point. It also always seems like whatever problem I have isn’t covered in their menu options. Verizon’s system is particularly bad.
I just lay on the 0 key until I get a person.
When you turn off the stereo, put down the burger, the hair comb, the driving directions, the electric razor, the makeup, that cigarette, the bong, that beer, stop talking to your passengers and stop looking at that funky looking dog!
Point is, we’re all doing something incredibly dangerous while driving, and those few who don’t usually tend to be horrible drivers and enter the freeway at 30mph and then merge.
If you don’t want other people needlessly endangering your life, you shouldn’t live in a human society, IMO.
The push-button systems used to allow you to avoid the menu system if you had a rotary dial phone. I act like I had a rotary dial, never hit any buttons, then go straight to a person. The voice recognition systems can now keep the “fake” rotary dialers in menu hell.
Um. Yeah. I dragged this one into the Pit.
I’m puzzled. The OP wants to speak to a real live human instead of to a computer, but also dislikes saying things on the phone. Do you expect to call the human operator and then push buttons at him or her?
And it’s been my experience that whenever a computerized phone system wants a number, it will always accept it punched into the keypad. But the voice recognition systems aren’t restricted to just numbers. The first time I encountered one of these, it was calling an airline to find out flight information. When it asked me for a city, I was able to just say “Cleveland”. Would it be better to have to go through ten minutes of menu choices to know what numbers to push for a city? Or recently, a friend of mine called an automated directory service, to find the address for a restaraunt we had heard of. What kind of numbers would you have to press for “Passage to India Restaurant, Westlake, Ohio”? A conventional numbers-only system just couldn’t do it, but the new system could. In either of these cases, the business would have had to hire human operators to handle these menial tasks, and that wouldn’t even be any guarantee of user-friendliness: How many times have you talked to a human operator, and not been able to communicate?
Since these have been under development since before the proliferation of cell phones, I always thought the driver was senior citizens. IME many seniors had a really hard time entering things on the keypad. Not sure if they do better with the voice response, but logically it would be harder for them to complain about.
I remember a particularly horrid system for Sears appliance repair. How exactly is their machine going to take me from “My dryer is broken” to some kind of useful resolution???
Don’t know about anyone else, but I can dial the necessary numbers with my thumb while keeping the phone beside my ear. And plenty of house-phones have the same setup, with the number pad on the handset.
True, but in almost all voice recognition systems, if you immediately shout “representative!” or “customer service,” it will usually get you to a human. Sometimes, however, the human isn’t much better. 
Many people like I, do not have the dextrity to make it through a option menu that requires 20 to 30 digits or more pressed correctly, or you have to start over. Think of a older person that is sick and having to type in the medication refils and verifications for each to get their perscriptions filled. I try for 2 weeks without getting through the damn procedure. I made it 5 minutes into the process a couple times. I finaly got the stuff ordered because in frustration I entered about 30 digits and the system sent me right to an operator the next time I called. Their system monitors the calling phone and the apparent malfunction sent the call to a person, which was not a menu option. Next time I may try the tune from “Close Encounters”, but the random atonal seems to work great.
I happen to work for a company that makes this type of speech recognition system.
A well-designed system will put you through to a live agent if it can’t recognize what you’re saying. If it doesn’t do this the first try, it should by the third try.
Not all speech recognition systems let you jump immediately to a live agent. Most companies do this to help you, and I agree with the logic. Would you rather wait 30 minutes each time you call, or 5 minutes? If the automatic system handles 80% of customer calls, the live agents are then free to handle the other 20%. The result is much shorter waiting times.
You will probably argue that the company could hire more live agents. That costs a lot more money than an automatic system, and you end up paying the difference. Or, the company goes off to some other country to get the people.
In line with the experience others have noted here, I have never had a problem with a voice recognition system. Ever. Perhaps I am editing out previous experiences due to my current job situation? I don’t think so.
Last weekend I used voice systems to check on traffic and to look up a telephone number. In both cases I had no problem whatsoever with recognition. I note this as distinct from navigation. I didn’t have to spell the city name for an operator, or the highway name, or fool with touchtone buttons.
I will follow this post with some hints for using recog systems.
[ul]
[li]Speak in a normal tone of voice[/li][li]If you hit a recognition error, don’t shout. Instead, speak a bit more slowly, and speak more clearly.[/li][li]Listen to the instructions, and try to match your responses as closely as possible.[/li][li]Pretend you’re programming a robot rather than speaking to a person. Unfortunately, speech systems often unconsciously trigger our near-instinctive habit of thinking a human is listening. One isn’t. You can communicate, but you’re not speaking to a human.[/li][li]Don’t rush to get a live agent. Try the system first. Some of them are quite good.[/li][li]If you hit a bad system, blame the programmers, not the recognition.[/li][li]If you’re forced to a live agent because of a system problem, tell the agent of your problem. It’s hard to diagnose recog problems from reports, while actual customer experiences help get the system changed.[/li][/ul]
Some more notes:
[ul]
[li]VUI (Voice User Interface) systems are a special type of program. Many SW engineers make a hash out of them because they’re too used to programming a GUI (Graphical User Interface). The result is a bad user experience, but recog isn’t the real problem.[/li][li]Similarly, recog depends on the grammar that you give the recog engine. A good VUI designer knows this and sets up a good grammar with a high success rate. A plain, GUI-oriented SW engineer may set up a grammar that is too broad or too narrow to yield good results.[/li][/ul]
Too broad a grammar confuses the recognition engine, because it can’t match any choice closely enough. Too narrow a grammar leaves out useful choices. For example, it’s counter-productive to add all 26 letters of the alphabet to a grammar if you only expect the letters “A”, “B”, and “C”. Similarly, leaving the word “highway” out of a grammar for choosing roads is a Bad Mistake.
It’s problems like those that cause recog problems, and not any general problems with the technology.
First, folks, you’re talking about speech recognition, not voice recognition. Systems that recognize voices are used in the security business. Systems that recognize human speech are used by telephone companies. I know–there are probably more people that get it wrong than right these days, but that doesn’t mean we can’t try.
Second, I’ve worked on and off with these systems since the early 1980’s, and they certainly have their place. That place, however, is not in a telephone menuing system that exists solely to keep me from speaking to the person or department I’m trying to call. The recognition quality still stinks. I have to repeat things several times. I can’t use it in a noisy environment.
I detest having to struggle with a speech recognition system for several minutes before I even reach the hold queue I want. I think they’re just hoping I’ll give up and not bother them.
Only part of that hatred goes toward the speech recognition system, though. Most of it is directed at the menuing system itself. I had to call Dish Network the other day, and none of their generic categories fit what I needed. I finally just picked something that got me to a real person and had them transfer me to the appropriate place. I wasted almost ten minutes that time before hitting the hold queue, where I waited another ten.