You are viewing a single comment's thread from:

RE: The Best Virtual Assistant.

in #hypeitup9 years ago

I've been playing at making Voice Assistants for several years now, mostly scripts. What I have settled on is something that basically announces its messages like a radio DJ. When I am doing something else and the music is on, between songs it announces the song, the time and temperature (randomly), any upcoming appointments, if I am outside, any approaching buses, various other things like that, then asks me if I need anything else. I can ignore or say no thanks, or ask my question, which is sent over to Google Now.

I ended up with that because I simply didn't use anything else. I had a complicated "Hal3000" bot for a long time, but never used it. I use my DJ assistant every day, but rarely ask it anything. It pretty much automatically does the main things I would ask it about, what time is it?, where's the bus?, what are my appointments?

At this point I see the main limiting factor of voice assistants as being the microphone. Unless I am wearing a headset, if I am near the microphone, I am either at the keyboard or have my phone in my hand. It's just as fast to type. When bots can understand language as you walk from one room to another, voice assistants will become a real tool instead of a toy.

Sort:  

What? That sounds awesome! How did you go about scripting it anyway? Any way to see it in action?

The current one is everything in Tasker and on my android phone. There's really not much to see, it's just a pseudorandom script generator for a few hundred actions . For instance :

Addressintrosubjectverbobjectoutro
Excuse meit'stimeto go tothe doctorBill
Pardon meIt isalmost timeto visit-----SIr
----------I believe it isyour appointmentto see--------------

Which gives you 2x3x3x3x1x2=108 random sentences when your calendar has a reminder about your doctor appointment. Usually with a little thought I can beef them up to seem pretty varied, maybe 500 different sentences or so. I just do that for various things, adding to the array of phrases when I run into something I need.

I just read the array and pick (with the equivalent of a case/switch loop) depending on the activity and then randomize the sentence structure to sound less robotic.

There's no real way to demonstrate it unless I'm doing something, I mean, right now it just tells me the name and artist of the song and maybe the time, I don't have anything going on until after Christmas. If I were outside it checks my location and uses the Transit API to find the next bus. Oh wait, it actually just reminded me to take my vitamins (1:00pm), but now that's gone.

I'll write a post about it one day and make some examples.

Transit API? Man that's awesome, I wish we have that here. I don't think even Singapore has something like that! Gonna check Tasker out. Speaking of which, I hope your data is safe using it!

I keep my data safe by making sure no one wants it. I keep it boring. The most someone could charge by stealing my credit card is $300

Woah! Thats pretty impressive! Hope we get the demo/intro first right here on steemit!!

I've been working on it in a Google Plus community (Tasker) for a couple of years. So no premiere! Lol! But if I had a little help getting my people from G+ over to here we might see all sorts of interesting stuff. I seem to be the only Steemer on Google Plus.

A transit A.p.i. huh. What city has that?

Most large cities that I checked, I'm in Lexington KY and I use
http://realtime.lextran.com/InfoPoint/departures.aspx?stopid=5013
To get the departure times at my grocery store