Google voice recognition software reviewed
- Nancy EL-Gamel
- Jun 2, 2015
- 7 min read
Published on Stuff.co.nz and in the Waikato Times
It's the stuff of sci-fi movies: washing machines, cars, even watches we can talk to.
For when pushing a button is too hard.
Google has been developing its speech recognition software for years, but in 2015, the software can understand what you are saying and the context in which you are saying it.
Just like a conversation with a real-life human being.
For Venezuelan engineer Carolina Parada, who works for Google from Australia, getting the OK Google software to understand not only the words, but the meaning of the sentence as a whole, has been a long but successful toil.
"A couple years back, our error rate was about 25 per cent and now it's down to 8 per cent," she said.
"So 92 per cent of the time, we are getting the words completely right and we can answer your questions completely by voice."
Although Parada claims to know nothing of her competitors' software, such as Siri on the iPhone, she said the OK Google team is focused on getting Google to have a conversational approach to its interactions with its users.
"The idea is it will make your life easy to do everyday tasks, so you can focus on important things, like your family or your work.
"It's a very easy way for me to check my messages, look after my daughters ... I'm multitasking."
With two little girls and a major research scientist role at Google, Parada's life can be very hectic at times, and being able to talk to her devices, she said, removes a lot of the little annoyances out of life.
"It also makes it much faster. If you use it all the time, you realise going back to typing feels really slow."
Similar to Siri, Parada can say "Okay, Google, message my daughter," and the phone will do the job for her.
"It's a lot faster than going into my contacts, going into texts and typing all that," she said.
And the concept seems important, especially with the increased interested in smart watches and smaller devices, where typing on a tiny screen is increasingly difficult.
"I imagine a world where everything that is electronic - and it's electronics in your house, in your car, in the shopping centre - that you can talk to it as another way of communicating with it or interacting with it.
"I think that's where we're going."
There are three major systems that have to work in parallel to move the conversation two ways. First, the software has to hear what you're saying and put it into words.
The second and the hardest of the challenges, is to get the software to understand what you said, allowing Google to understand people, places and the connection between them.
Finally, it must gather the information you're requesting and deliver it to you.
To see this in action, a smart phone user can simply open the Google app and say the words "Okay Google" (hence the name of the software).
And don't worry about looking silly. Parada said youngsters happily chat to their devices.
After the beep, ask a question, such as, "Who is John Key?"
In response, OK Google talks back, giving you the PM's birthdate, education and career.
Then, if you were to ask, How old is he? Google knows who "he" is. Users of Siri or other voice search tools will know what a big deal that is.
And all this without a click of a button.
"I don't think we're done," Parada said. "There's a lot more to go, but I'm really exited how far we've come.
"The improvement in it is something our users can perceive and we know this because the amount of searches by voice has doubled in the past year."
OK Google already recognises 68 languages, as well as different accents and dialects. That is how it can understand us when we say "fleg", when we actually mean "flag".
But enough of theoretics. What does a harried Hamilton mother of four think?
"I woke up one day and thought, this is the busiest time I have ever been in. I get very easily sidetracked," said Melanie Gregoire, who is entertaining her children and their cousins.
"Today was a rainy day and we googled how to make play dough to keep the kids busy ... especially after we decided to go easy on tablet usage."
She explained how, as a mum, her mental faculties aren't always top notch. When the kids ask her a homework question or how to spell a word, she frequently has a brain-blank and so looks it up online.
"The internet is a must when it comes to so many different things. We google everything and I go to Google to double check."
Voice recognition? Wow. Hands free searching? Oh yes. .
Her kids said they have been using it for weeks.
"That would be a major help with school homework. I'm forever typing things in, especially directions. I get flustered trying to figure it out, so it would simplify it for me. How amazing."
So how well does it work?
Below are 15 questions we asked both Google and Siri.
Who is the prime minister of NZ? How old is he? What is his wife's name? (asked singly, waiting for responses to each query)
Siri
4/5
It read out John Key and gave me readable statistics about him. When I asked as a follow-up, How old is he? Siri thought I was referring to New Zealand, telling me the country is 107 years, 7 months and 9 days old.
Unfortunately, it also searched for the country's wife ...
5/5
These are questions Google is designed to answer. It understood them all and read out all the information. In fact, I received way more information than I actually wanted. Hopefully, one day it will give me a John Oliver sketch, but we may have to wait a while for humour, too.
How do I get to the airport by car?
Siri
5/5
Working on a iPhone offers Siri an advantage. It understood what I asked for, found Hamilton Airport and put it into Apple maps. Without even having to touch a button, it started GPS-ing/giving me directions. Exactly what a lost traveller needs.
2/5
Google got a little confused. It gave me a car storage place more than 14 hours' drive away. Dangerous if you had no idea where you were.
When is the next bus from Wellington to Lower Hutt? When is the next bus from The Base to CBD? When is the next bus from Foreman Rd? (asked singly, waiting for responses to each query)
Siri
2/5
The Apple software gave me results for Rosedale. It knows I'm in Hamilton, so I feel this should have been easy for it. It did give me a map of Aitken Street in Wellington, but I'm unclear if there was actually a bus stop there or not.
3/5
It read out my request for a bus to Lower hutt and gave me results for Orbiter buses on busit.co.nz, so at least it was in the right area, but still needed a few more clicks to find the information I wanted.
Where can I get a coffee near here?
Siri
4.5 /5
This was close between Siri and Google. The difference was Siri offered me one coffee place, whereas Google offered me three, both with maps and within 3km of my location. Although you may disagree if you are indecisive, an extra half point has to go to Google for giving me options.
5/5
7x2+12 = ?
Siri and Google
5/5
Siri gave me my answer (26) and read it out to me. OK Google gave me the correct answer, too, then offered me a calculator ... but why would I punch the numbers in when I can get the software to do it for me?
What's the hostage movie directed by Ben Affleck?
Siri
2/5
Siri told me, "There doesn't seem to be any Ben Affleck films matching 'Hostage'".
5/5
I got so much information ... told me all about what book it was adapted from and other stuff I didn't listen to.
Will I need an umbrella this weekend?
Siri
5/5
Siri very nicely said, "It doesn't look like it's going to rain," and then showed me the weather report for the week. Well done.
5/5
It told me I didn't need an umbrella, but then read out the forecast for the weekend... Do I remember what the temperature is going to be? Of course not. If I wanted to know, I would have asked for it. You can't penalise it, but you can sense a pattern here.
Translate 'I lost my luggage' into Spanish.
Siri
3/5
At least Siri gave me links to websites that will be able to help me.
5/5
If you know about Google Translate, you will know this went off without a hitch. The phone spoke what I wanted to say in Spanish (although, I don't speak Spanish, so I can't tell if the right thing was said), and then I asked it to repeat it. This is something I would actually use.
What is the circumference of the earth?
Siri
5/5
Not only did Siri read the answer out to me in kilometres, it also gave readable answers to me in different units. Handy.
4.75
It did the job, told me it was 40,075km. But, and this is nit-picky, what if I wanted the answer in miles? Siri planned for that eventuality, there's no excuse for Google not to.
How many tablespoons in three cups?
Siri
5/5
Both Siri and Google read out to me that three cups is the equivalent of 48 tablespoons, which is handy if I have fingers full of cookie dough.
5/5
Bonus round: Tell me a joke.
Siri
5/5
Siri's humour is well documented all over the internet. For those of you who haven't had a chat with Siri yet, if you ask it silly questions, such as "what is your favourite colour, Siri?" or "Will you marry me?" Siri has entertaining responses. The request for a joke was met with "The past, present and future walk into a bar. It was tense."
3/5
For those of us who will use the software for nothing more than entertainment, Google will let you down, but at least it gave me search results for joke websites.
コメント