Thursday, October 11, 2012

Better Living Through Speech Recognition

A new feature on Android-based phones (as of version 4.1) is a "personal assistant" app called "Google Now" (see the video introduction of it.)  I've been using it, and I'll report the results as they come in.  But for now, what I want to report on is one aspect of it: a system to respond to simple spoken questions.  (Again, see video introduction in front of hysterical Google fans here.)

Of course, that's a lot like the famous Siri feature on iPhones, which was the big feature added for iPhone4.  (As opposed to iPhone5's big features, which was that it's slightly thinner and the maps don't work.)  So yes Apple fans, you did have this feature first.  But then, Apple is currently running a rather annoying commercial bragging about their panorama photo feature, which Android already had.

Of course, I've been testing the voice search ever since.  One disappointing aspect is that it can only respond in that friendly-yet-formal voice for a limited selection of data.  For instance, I verified that as in the video, it will tell you the height of the Space Needle, along with many other famous towers.  But when I asked, "What's the new name of the Sears Tower?" it just put that phrase into the normal Google search engine, and gave me a list of links related to what is apparently now called the Willis Tower.

Anyway, here are some other experiences:

Q: How old is Ichiro Suzuki?
A: Answered simply and correctly, "Ichiro Suzuki is 38 years old"

Q: How old is A-Rod?
A: I was worried when it transcribed it as, "How old is a rod?"  Yet it answered correctly again, "Alex Rodriguez is 37 years old."

Q: How did the Jays do today?
A: Got confused, spelled it out as, "How did J today," and gave me a list of web sites, strangely starting with John Elway's Wikipedia entry.  I rephrased it to, "Did the Blue Jays win today?" and this time it understood, responding, "No, the Blue Jays lost 3-0 to the Rays."

Q: Who won the Grand Prix today?
A: I was referring to the Singapore Grand Prix earlier in the day, and it gave me a write-up of the results.

Q: Who is Joe Walsh?
A: No speech, but it gave me a quick fact-sheet of his career.

Q: What is the Sub of the Day?
A: It just got confused, even when I specified the "Subway sub of the day."

Q: What did Mitt Romney say about airplanes?
A: I asked this after a bunch of confusing references to a Romney gaffe talking about air travel.  It did indeed give me a list of pages regaling the incident.

Q: Did Villa win today?
A: This was in reference to family soccer team Aston Villa.  Kept interpreting "Villa" as "Phil" or "Philo."  When I specified "Aston Villa," it gave me a table of recent Premier League results, which would have been nice, but I was enquiring about a League Cup game.

Q: Where is Dortmund?
A: Interpreted as either "Where is George mind?" or "Where is Georgia Mountains?"

Q: Did the Orioles win?
A: This was only minutes after the Orioles/Yankees playoff game had ended.  The question was transcribed as "Did the girl win?"  You can imagine the split-second of panic as I saw that this is what it had fed into the search engine.  Fortunately, only benign (but unhelpful) pages were returned.  Upon re-asking the question with superior enunciation, it interpreted it correctly.  However, the Google voice only responded by telling me that the Orioles and Yankees would be playing that night, seemingly oblivious to the fact that the game was over.

Q: Where's the nearest Salvation Army?
A: Successfully showed me three local properties.

Q: Where's the nearest cafe?
A: Gave me several valid answers, though it missed two of the closest ones.

Q: Show me pictures of Japanese Flying Squirrels
A: Just like the Pygmy Marmosets in the demo, I was successfully overloaded with cuteness.

No comments:

Post a Comment