Google Wave: Natural Language Processing

[JustinNeverSayNever7]

I love that Shirt, Can I have one? :) LOL

[Matriluzviva]

this will be able to create a universal and of easy learn universal language that can make us be able to talk with every nation in the world. I will be able to flert with chinese girls at the internet, sex tourism will increase in CHina lol, thanks Google \o/

[cgarrey]

this is all waste of time .. autocorrect doesnt work anywhere, and making it 5% faster doesnt mean, these people will use 5% time, they will produce 5% more garbage and spam .. hence just dont do this anywhere ... you are just wasting your time ..

[BowDownToSwineflue]

LETS MATE TOMORROW!

[27emman]

how do you add spelly?

[L0VECHILDD]

so, if i want to meAt with my friend tomorrow (meat = eat some meat), this tool will change my spelling, even though i intended to write meat?

[punchthedog]

Meat is not a verb.

[L0VECHILDD]

@punchthedog right, but in our community we use it as a verb all the time. I hope they plan on having settings for different dialects.

[JoseJimeniz]

If you write, "Let's meat tomorrow", then Spelly will smack you in the head.

[MrProgrampro]

@L0VECHILDD I think it only suggests that you change it, like microsoft Word's spellcheck. It doesn't change it automatically.

What would be cooler would be if, instead of statistics, this thing used grammar patterns. eg. "met" is a past participle of "meet"; but the infinitive form "meet" is required following "Let's", so based off of that it'd change the word. Then again, this wouldn't work for general misspellings.

But still, for grammar check, this'd be cool.

[tmarbut]

Looks way cooler in Wave than on his outdoor whiteboard.

[googlwave]

НУ что скажешь угарно

[anxious223]

Watching this work in Wave is really amazing. It's considerably "smarter" than the spell checker in nearly every other app I've used. Much of the time, it just fixes things for you with very few misses.

[WiburTN]

This more than anything might be the most exciting and impressive thing Google has going right now. This is absolutely amazing!

[MarkOates]

This is great and all... just let me know when Google is going to wipe my ass.

[Prepentach]

ask your boss if you can go inside and show us the actual software. Or are you just some homeless guy with a 'google' shirt on.

[qazwsxedcrfvg123]

@Prepentach He is not., because this video is posted on the official Google youtube channel.

[graymalkin1234]

What is the point of the whiteboard?

[MotAdvNSW]

Should make for some interesting conversation, particularly those into Tok Pisin:-)

[snezhiko]

so boring!

[obsidanse]

Yes you're right. It doesn't give you a quick hit of adrenalin... IT'S NOT MEANT TO YOU FOOL!

He's talking about a spell checker, not the latest EA game.

[mauronr]

On the fly? In a browser? In a collaborative tool?

[TrantaLocked]

Actually, though it may exist online, I have not seen it in any browser or online forum. It still has existed for quite some time.

[haaampie]

Take another look...

[sjwashere]

Awesome and 10 points for getting a Sydney boy to present!

[dzgndotinfo]

It all makes sense. Google Voice, Google Chat are probably feeding natural processing data to Google for Wave

[garycalgary]

this is preceisely why people in this world can't spell.

[200609]

On the bright side, you know that it's "I before E except after C."

[deathreape98]

Except in science.

[Neylonx]

or weird

[cmba101]

Think there's a trick to this system.. Teachers could ask to be sent the document created through a wave and review the history.. (ie see if spelly was used)..

A simple school policy, such as, no spelly allowed would solve quite a few problems

[AndyMacProGamer]

speak up faggot.

[garycalgary]

calling people names make you feel better?

[AndyMacProGamer]

simply saying speak up can you please shut up for once and fucking deal with it? Or does everyone have to bitch about the way I comment? FUCK!

[JessClearwater]

I didn't notice the Aussie accent till about 2mins in. Duh. Haha

-Jess (Sydney)

[sroodhorst]

nope. this thing directly check your words and corrects them immediatley if they are wrong typed

[Kevbox2008]

Is this the same thing Microsoft Word does with Spell check and Grammar Check??

[stgpetrovic]

No.

Microsoft let's you add a dictionary/grammar (not to big in size) and then uses the rules provided inside to see if you type correctly.

In wave - no grammar is used, but your text is compared by similar texts in all English websites seen by Google, and then what appears more often is offered first.

[ConflictedMonkey]

Well, it's a really bad clip as far as production value goes (Google is using a side-angle whiteboard on a noisy balcony to tout its technical advances? Does anyone else see the stark irony??) but even as an English teacher and fast typist, I can *still* see the benefit of this.

[mkklmann]

Whue cæres abaut spelling corect, annyways?

[OldProp1941]

Man, I hope we can turn this "feature" off.

[Kaminix]

Did u srsy bellieving that wold be a mandatory spel chekker in Gugle Weive?

[manueltheplebb]

This may already be something Google have already thought of, but what if 'Spelly' remembers your corrections, including the context so in this example it would remember the words; 'Let's', 'met' and 'tomorrow' but it would also remember that you changed met to meet. With this information would it then be possible for 'Spelly' to make corrections for you, or indeed suggest more accurate suggestions??

P.S At this point i havn't watched the full video so don't hate me if he mentions this later on.

[antong20]

Well, I don't think that would make such a big difference since Wave has the whole internet to observe. There is a similiar function in my cellphone, it suggests words based on what I usually type, and it's just annoying. It seldom picks the right words.

I think that Spelly would be less accurate if it used your personal history of spelling errors too much.

[antong20]

Hmm.. That was supposed to be an answer to manueltheplebb's post...

[manueltheplebb]

Cool, lol, i never thought of those things. I'm kinda new to programming too so i don't really know what the restrictions are. Would be cool if they developed one oneday that just knew what you were thinking lol.

[joyerjoyer]

I think the best part of this is the centralized language processing on the server side where you could host a huge database with a very powerful processing.

[benzourry87]

that's graat!!

..opst, mispel..

..opst, and again..

Haaa, wish they could employ NLP here too..

[t3hubern00b]

My question is how long is the delay for the correction after typing? Will it be worth the 5% faster time, if you're going to be slowed down anyways?

[miel69]

there's no noticeable delay, check the google I/O vid :)

[015159978732]

0:59: "ups" :D

[ReverbDP]

Why do i get a thumbs down for just saying he seems nervous... lol DO you all love him or something lol

[smowkmylungs]

Well who isn't nervous when giving a presentation, even actors and singers whos job is 80% presenting get stage fright. Let alone this guy whos job is under 10% presenting.

[IntelArt]

I would like to meat all Google girls...

[liguofu]

@IntelArt you mean..meat? take them as meat in dinner?

[IntelArt]

@liguofu MEAT... Did you ever MEAT a girl? MEAT... I meant MEAT the ALL AMERICAN Google girls...

[lemiffe]

@IntelArt you mean meet not meat.

[ReverbDP]

He seems nervous when making this lol..

[bloodcumulus]

nerd is nerd

[thiebes]

The next step could be for Google to write likely sentences on a topic. Then likely paragraphs. Then have Google's data center write a book about itself.

[cybertpad18]

aren't all these features already available inside Microsoft Word?

[uvmedraco]

No, this is cleverer and free, like he said, it's not a spell checker and doesn't have a dictionary it just suggests words similar

[cybertpad18]

microsoft word does that same thing imho. anyway.

It will be a huge tool to use. specially for collaborative working. hope some companies actually realise and self contain all these possibilities inside their products. Who doesn't wanna see a real time collaborative photoshop? or dreamweaver? I'd love that.

[t3hubern00b]

Do you use Miscrosoft Word as your prefered Aim client?

[cybertpad18]

will you use wave as your im client? - .-

[1847ez]

I appreciate any effort to make our communication clearer and more efficient. (Your shirt not withstanding.)

[llillillillillllill]

this freakin owns

[nottRobin]

it's a great service, but a shite demo

[cybertpad18]

it's not a demo, it's a tech video.

[thehayabusahaloninja]

nice tshirt

[fycat1995]

可否用於其他語言呢?

[NOHAYCHANCES]

如果大多数语言

[NPRCVI]

New applications in text editing systems, Conversational text editing systems, and industrial applications contained symbol manipulation of comprehensive hypertext. What we have dedicated to the 21st Century is streams of symbols, varied translations, to accomplish communication and allow human cognitive behavior become an active processor of information.

[immigration420]

That google office in Sydney looks nice.

[srijansandilya]

Will this work for languages which are processed via IME? The coomon error calculation for chinese, Japanese and other languages would not follow the same mechanism.

[onoken2tokyo]

Context-aware and knowledge -based together with phonetic algorithm is currently mainstream to suggest a correct spell. Goolge speller is based on Hunspell plus trend words including wikipedia and so on...

[webjthefirst]

how does Spelly know that "Let's met tomorrow" is more likely than "Let's meet tomorrow" without knowing the difference [in meaning] of 'met' and 'meet'? Also same question for context (/grammar) "Let's met tomorrow" is grammatically incorrect, but you would only know that if you knew the meaning of the word 'tomorrow'.

[mortenlu]

Say, "Let's met tomorrow", is used 300 times on 6 billion pages while "Let's meet tomorrow" is used 30 000, its not that hard to imagine. Ofcourse, google uses way more advanced formulas.

If you typed "Lets eat meet tomorrow" it would correct it to "meat".

[webjthefirst]

oh awesome! I totally didn't think of that :P . Google really has managed to incorporate their original service into everything.

[AmberwiseTheWise]

Somewhat hard to hear, might I suggest you pick a more quiet location/directional microphone?

[garvasis]

need to work on presentation skills clarity bravity and keep it simple and to the point . WAVE is a npo brainer , doubt its sucess

[exo762]

no no, facebook and twitter are no brainers. Wave is rather impressive _mature_ and _usefull_ product.

[goodsirknight]

shitty tshirt Casey

[evonmixon]

cool

[lunaticculpepper]

lunatic culpepper says thank you

[dejudicibus]

How does it work for other languages, for example, Italian or Russian? And what if you use different languages in the same document, which is quite common today outside USA?

[n00braz]

Well, he said it would work for other *alphabetic* languages, which i suppose should include your examples...I think something like Japanese could be a problem.

Different languages in one document should work if you don't use one foreign word out of context.

[dejudicibus]

Well, since it seems to understand some semantic to selecta world, I wonder how it can manage mixed languages. For example, what if I write:

George' letter begins by «Caro Giovanni, come stai?»

[sanatgersappa]

you might want to get a quieter location for your next video.

[BloodykneeVrykul]

they should work it out going by the tense of the verb and the time: met and tomorrow, which is past and present tense. meet is closer to met than yesterday is to tomorrow so met should be changed to meet. that seems like a nicer algorithm to me.

[adeori1]

Yes, although that is not going to work in all scenarios. They opted for an algorithm that would be applicable in more circumstances. In that specific case, your option is better, but not in all cases. Instead, they simply made it slightly more complex. It is rather similar to yours, though, it seems.

[alienorz]

haha from nover

[jameshehe]

he screws the meet up . . . and not the tomorrow . . . lol

[JamiB92]

LOL

[Dampir]

So, basically all you conversations will be sent to Google, and most probably logged. What about personal privacy? Why are people so easily content with minor conveniences in the exchange of their privacy?

[acorm]

The protocol and server software are open. You could use the service from a trusted source to you.

[typhoon476]

The spellee feature is not open. It's a patented technology. Search for "AUTOMATIC CORRECTION OF USER INPUT BASED ON DICTIONARY " on google patents.

[mod50]

What he's describing is a patented technology too (patented by Microsoft). I know because I co-authored the patent. :-)

[Monchanger]

@typhoon: if they release source code that implements the patent along with a license to use it (this is included in all licenses), then patents are not an issue.

[Dampir]

I do not know whether the protocol or the server software is free software or proprietary software. The video did not say anything on that issue.

What the guy in the video said, is that it would require a server farm and it's not feasible to run this service on your own. Whether the server software is free or not, doesn't matter that much if you can't run it on your computer.

[Dampir]

So, practically it will require you to pass all your conversations through Google, or maybe later on, through some other big corporation (if it's really free software).

If people could run this service at home, on their computers and it is free software, then I may say that there is no concern.

[Monchanger]

@Dampir: watch the Google IO presentation. It should make clear that the server farm pertains strictly to the spell-checker robot and not the core Wave infrastructure.

If you're really hell-bent on running a spell checker on the desktop computer stashed in your garage, you'll just have to implement your own less processor- and data-intensive service (or install one that some open source effort developed).

But yes, letting someone else spell-check you means sharing information.

[Dampir]

Well yes, that's exactly what I said didn't I?

You may be free to run this service "in your garage", but practically - you can't, it is not feasible, it would require a server farm, it is too "data-intensive".

So, you have to "share" all your personal conversations and documents with Google. Of course, I would not use the word "share".

You said exactly what I did, only you used a more positive wording.

[Monchanger]

No, Dampir.

An analogy: compare GMail to running your own mail server. You only need to scale hardware when you scale users.

The Wave technology is not data-intensive until you scale it to the millions of users that Google serves. Your garage machine and home cable networking should have no trouble handling several dozen active waves (and hundreds of inactive ones) with several participants each.

Remember- we're talking Google here, not Microsoft. Lean software is the modus operandi.

[Dampir]

We were discussing the spell-checker up until now, weren't we?

Now, what you did was, you weaselly substituted the spell-checker with the social networking server just because it suites your argument.

Your analogy does not work for the spell-checker. It was said in the video, that the spell-checker server does not "scale with users" and precisely because of this, unlike other spell-checkers, it is an online service.

Watch from 4:00 to 4:30.

[Monchanger]

I didn't discuss scalability of Spelly, I never said you could get it running in your garage. I said you're looking for a "less processor- and data-intensive service" and I meant it. I never suggested Spelly was portable and explained that for a private installation you'd need a different implementation. Period.

Anything else you care to complain about or misrepresent?

[Dampir]

I just watched the Google Wave Preview, and it was irrelevant to our current conversation. Nothing more was said concerning "Spelly", than was said in this video.

I knew already, from this video, that the spell-checker is a standalone server.

And last, I need to clarify, that it does not matter much to which company you would need to send your private conversations and documents to. You should not send it to any of them, that is why they are called "private".

[MoeruzeBurning]

I would say if you want your conversations to be private, then host a wave server yourself, or don't use wave. Where is the problem?

If you can't run spelly on your server then don't. It sounds like it will be trivial to make your own spell checking robot to replace spelly, or not use spelly at all.

I personally would be much more skeptical about hosting my own server than trusting google with my personal conversations.

[Dampir]

MoeruzeBurning, we are discussing the spell-checker (not Wave) and you can't feasibly host Spelly by yourself-for yourself.

Please don't make the typical bullshit "if don't like it, don't use it" argument. Sure, I won't use Spelly, there are other spell-checker that will respect my privacy. Can't I give my critical opinion on something I don't like?

Why would you be even a bit skeptical of something you have complete control on?

[MoeruzeBurning]

Anything I create is far less likely to be secure than something created by a team of people with more web experience than I. If I am going to be communicating over the internet, I assume my privacy to be compromised anyway.

The trick I use is not caring about privacy. If I really cared, I would learn to spell.

[astroquick]

LOL, how stupid can you be?

[Monchanger]

Dear Dampir:

Your hesitance to use Google software and spreading of fear and distrust has been noted. The logging threshold of your personal information has been adjusted to assure our capacity to maintain control of your life.

Sincerely,

Server #0x526E1A

PS- Don't ever ask Johan about girls- that kid is clueless. You should totally ask her out- she digs you too. Good luck!

[youlikame]

Woe is I. I am not a developer, so my comments should be quickly dismissed. Nonetheless, I appreciate an intelligent critique, Dampir, please go on...

[henkvandervelden]

Sounds good!:)

[jim30333]

This is hard to understand. So the program looks at the word in question in context to the sentence? Why didn't he just say that? That would have saved me 70% more time! XD

[dodeja]

I think this is some great stuff. Although they really need to open up this technology and let others use the data set they have built. Google has the resources to share their data, people/developers will even pay for the usage of this data set.

[nikthetrick2]

They more than likely used Google Web 1T which is available from the Linguistic Data Consortium. It's $150 if you weren't a member in 2006. Oh, and there are 1,024,908,267,229 tokens, so I hope you've got plenty of free space!

[dogstylee51]

So Wave is basically a web based rich-text instant message client.....

[adjohnson916]

@dogstylee51: And e-mail, and collaborative document editing with file versioning and history, and file-sharing, and photo-albums, and blog commenting, and hundreds of developer plugins on the way...

[elatedrobot]

This is a brilliant move on Google's part. Differentiated experience has the greatest potential to lock people into using a service. Cool and useful [patented] features like this will ensure that people become long-term users of Wave.

[shaolindreams]

where can i get your tshirt?

[beck2c]

Google "Google Store"

[shaolindreams]

thanks but i was only making a joke :D

[b1asphemer]

the logo looks like the SLAP CHOP blade ;)

[SIelhell]

How does spelly know that "met" is miss spelled?? It is a valid english word.

[ssc4k]

It looks at context

[chongli66]

Try doing a google search for "let's met tomorrow".

You'll see what he means.

[tjerk]

it looks at a statistical language model and the context and concludes that "met" is a lot less probable than "meet".

A lot of math and computer science behind it. There not dumb at google.

[phillysaxon]

Hopefully that model'll be good enough to pick up on the mistake you made there.

[Ezmarii]

I once saw 2-3 pages in the back of a magazine i think, when I was taking a few courses at ITT Tech. They were a series of brain teasers, and the first page said that some of these are very very similar to what a google employee must solve to get a job there. Man they were CRAZY. Like, if you see easy, medium, and hard brain teaser puzzles, these were all medium/hard, and tested you on different things like crosswords, shape/geometry puzzles, number puzzles...crazy.

[adjohnson916]

@SIelhell: He just explained this. It looks up the context of each word and checks to see if a replacement word has a significantly higher probability on Google's database.

[fidler12]

This is going to be helpful for so many people! It will also separate Wave from any would be competitors. Congrats!

[davidalln]

This is very exciting!

[chu121su12]

I was thinking this audio is nice...

it felt like i was there

oh. except the wind sound..

[HandfulOfMinutes]

Interesting stuff but I'm curious why a company that makes the annual GDP of Guatamala every ten seconds couldn't have spotted this poor guy for a decent microphone that would have excluded all that background sound.

[Monchanger]

That's one way of looking at it.

Another is "We're such a cool company, you can work outside if you like."

Yet another could be: "We're professional enough to not let silly things like a little background noise distract us from learning about really interesting problems."

I really wish YouTube had an "ignore shallow comments" filter.

[minimalniemand]

"I really wish YouTube had an "ignore shallow comments" filter." - thanks this language model and youtube owned by google - should be possible!

[Theowest]

Firefox add on......

[GClueandroid]

Too Cool!

[dolphonia]

If it can correct your when it should be you're then Google are geniuses (but then we know they are anyway) and have made the world a better place

Also, loving the t-shirt.

[chkchun]

3rd... yeah. google wave...

[coijoi]

2nd

[namberwuan]

First coment.

Please add captions.