back to article Google widens search net and takes on Siri with iOS app

Google is moving closer to a planned search singularity with the extension of the Knowledge Graph system, a trial to allow personal Gmail search results to be included in generic web searches, and an iOS app that takes voice requests and tries to answer. Knowledge Graph, launched in May for English-speaking users of Google, …


This topic is closed for new posts.
  1. Anonymous Coward
    Anonymous Coward

    Does it come with built in malware and does it have DRM so it crashes?

  2. Oninoshiko

    Isn't competing with an Apple product

    on of the EXPRESSED reasons Apple denies apps? It's not even one of the "well, we don't need a reason" reasons. I though they actually gave that as an example.

    1. Anonymous Coward
      Anonymous Coward

      Re: Isn't competing with an Apple product

      Of course not, plenty of products that compete with Apple are approved, like Sparrow, Kindle, Spotify... There's other voice search apps on the store already, like Evi or Voice Actions.

      The only rule is that apps that look confusingly similar to the bundled apps can be rejected, but from the video I think Google probably scraped that through (knowing Google they probably iterated through less and less similar designs until it finally got accepted)

  3. Anonymous Coward
    Anonymous Coward

    How things change

    I don’t believe that your phone should be an assistant. Your phone is a tool for communicating. You shouldn’t be communicating with the phone; you should be communicating with somebody on the other side of the phone.

    – Andy Rubin, SVP of Mobile at Google and founder of Android,October 19, 2011

    Rubin->reality translation:

    "Shit, we don't have that feature yet"

    1. Andrew Jones 2

      Re: How things change

      Google still don't believe your phone should be an assistant - when was the last time you asked an assistant how tall you have to be to ride space mountain, or what time the sun rises in Japan or how long the runtime is for the movie Ted?

      The Google Voice search is also not new - Android has voice search forever - the only different is that if there is a specific answer to your question that Google knows about it - it can speak the answer instead of just providing a list of results (which it will still do even if it does know the answer).

      Now - on to what I wanted to ask.

      Could the register please find out more about the technology behind the spoken response? On all the Google Now videos on YouTube - the response actually sounds like Majel Barrett - who of course is the voice of the Star Trek computer. Now it's obvious that Majel Barrett probably hasn't sat and recorded every known word in the English Language - so could you find out how they managed this feat?

      There are rumours that Google acquired some tech last year that enables them to record about 200 words of someone's voice and then some sophisticated tech can uses some special algorithm to generate any word you want in the persons voice - but there isn't much information out there about it.

      Even if this tech is sci-fi fantasy - the response does not sound like the artificial voices TTS has always been plagued with - so it would still be interesting to find out more about the tech Google is using to enable a more human sounding voice.

      Many thanks.

      1. Anonymous Coward
        Anonymous Coward

        Re: How things change

        Umm I don't find any resemblance to either Majel's natural voice or Start Trek's computer voice.

        1. Andrew Jones 2

          Re: How things change

          You don't think this sounds at all like the Star Trek computer?

      2. Sean Timarco Baggaley

        Re: How things change

        Modern speech synthesis uses a matrix of phonemes at its core. The "Siri" voice in the UK is the same as the "Daniel" voice available for OS X Lion / Mountain Lion users and is based on the voice of Jon Briggs.

        Jon went into a studio some years ago and recorded 5000 phrases in a monotone voice. From those phrases, a complete set of phonemes was extracted and these are what the synthesiser plays, adding pitch and volume changes to simulate stress and intonation.

        This is why, although Siri sounds a lot better than the older fully synthesised systems of old (think Prof. Stephen Hawking), we can still tell that it's a synthetic voice. When humans modulate their voices, they do so by changing a lot more than mere pitch and volume: they constrict their throat, move their tongue, lips and change mouth-shape, change how they breathe, and so on. Our ears are trained to notice these changes, so we're still in the voice synthesis equivalent of the Uncanny Valley. But we're getting pretty damned close.

        Crossing the valley completely is technically feasible, but will require paying the voice artist to record those 5000 phrases multiple times, in multiple intonations and projection levels. That alone will add days to the recording process, and it'll be very, very boring. Once processed, the resulting voice phoneme set will also be much bigger—the UK "Siri" voice is a 500MB-ish download already—and that's the biggest problem. It'd take much, much longer to process the raw sample data and produce the final phoneme matrices, and it'll take up a whopping great chunk of storage space too. Realistically, computers will need to come with either much faster storage, or a lot more RAM, as standard.

        1. Woodgar

          Re: How things change

          Sounds similar to a vocaloid (page in Japanese), only more sophisticated.

          Vocaloids use the same technique, but for singing rather than speaking, but the principle is the same, in that they use a voice bank of phonemes to construct a full singing voice.

        2. CmdrX3

          Re: How things change

          All very interesting, but it still doesn't stop Siri from not having a fucking clue what I want most of the time. I have a five minute journey to my brothers house. If I try to use Siri to call him and say I'm on my way I'm usually there before El Thicko even manages to figure out who I want to call and on what number. I've tried using it on numerous occasions but it rarely gets things entirely accurate and I give up and do it manually which usually works out quicker. Hopefully IOS 6 will improve it. but I'm not holding my breath.

      3. David 164

        Re: How things change

        It was a UK company, Phonetic Arts, this is a user document for their product.

        It was acquired in 2010.

        1. g e

          Re: How things change

          Was that the bunch that had Tom Baker record a phrase containing all English vocal sounds so that BT could read SMS to you ?

  4. Anonymous Coward
    Anonymous Coward

    Of course would Google want a "Siri" equivalent

    After all, Google collects information. Collecting pristine digital voiceprints of everyone is yet another bit of information they can sell..

  5. QuinnDexter

    "I imagined a future where a starship computer would be able to answer any question I might ask, instantly. Today, we’re closer to that dream than I ever thought possible during my working life."

    Have Google got a starship?! This has been about on Android and Apple for ages, hasn't it?

    1. g e


      It's been in Android (or installable as Google Voice) for ages, think I had it on my old HTC Magic, so up to 3 years ago there. Apple have had it since the 4s phone iteration, so about a year if memory serves. Apple bought it from Nuance (IIRC), dunno where Google got theirs, I guess it's maybe derived from Festival/Sphinx for ASR/TTS, perhaps? Wouldn't have thought Google would bother writing one from scratch as there's so much FOSS groundwork done out there already, maybe they bought someone too.

  6. David Pollard

    Be careful what you wish for

    "I imagined a ... computer ... to answer any question I might ask..."

    To the extent that responses were accurate and truthful, many of us would often find this rather uncomfortable.

    1. nexsphil

      Re: Be careful what you wish for

      This is bloody true actually. Imagine a computer that simply answered your kids' questions about Santa with a flat "No", or labelled your favourite religion a "simplistic control scam". Or reminded you that political corruption and conspiracy is very, very far from "haha impossible" and that your generation will be regarded as morons for believing this for the next 1000 years.

      Yeah, truth has limited use to humans. We tend to favour comfort.

This topic is closed for new posts.

Other stories you might like

Biting the hand that feeds IT © 1998–2022