Slashdot Mirror
Browsing the Web, One Sentence at a Time
rtmyers writes "A really simple yet radical idea: break web pages down into sentences, and then have the browser walk through sentences and do useful sentence-level things. This is the paradigm shift behind the product called Infowalker, which unfortunately is implemented as an IE toolbar, but would be fabulous as a feature built into Mozilla or Opera.
Currently implemented features include sentence-level interfaces for TTS, translation, large-type display, and the funkiest of all, dynamic display of an image pulled off the web based on keywords extracted from each sentence -- hey, turn all your web pages into slide shows today! Then there's the feature to show an Amazon product related to the sentence you're reading -- which presumably is the revenue model behind the product, but turns out to also be surprisingly useful.
This might not be for everyone, but it could just be the first real change in the browsing model since the earliest browsers starting throwing text up on the screen more than a decade ago. And apparently, Infowalker's architecture allows for pluggable third-party sentence-level "behaviors", with the potential for the development of a whole ecosystem of sentence-level functionality in browsers. And it seems Infowalker can also be controlled by strategically placed custom CSS tags within the HTML, raising the possibility of a new class of web pages especially tuned for this sentence-based approach."
If you really want to browse the web, try this: The world's best Surf Engine IconSurf.com, where you surf the web using website icons. When you want to explore instead of doing targeted searches or visiting the same old sites, this is the way to do it. Shameless plug of my new toy, but it's pretty revolutionary as well. I'd appreciate any feedback, as the site is just a little over 1 week old.
Because as this shows, breaking coherent paragraphs up into sentences and further mashing them into keywords, makes works far more accesible.
Good precis is a skill, and a creative and demanding one at that. A computer will no more likely do a good summary of a lengthy text, than it could write a novel.
Athletic Scholarships to universities make as much sense as academic scholarships to sports teams.
Is it just me, or did all that make no sense what -so-ever?
My Webcomic: Asylum on 5th Street
That's a good idea. I'm going to make an addon for IE that does spell checking and auto-fixing. Now if we can just get the spelling and grammar trolls to use it so we don't have to listen to them grumble.
I can count to 1023 on my hands. Ask me about #132.
Ever see dashboard? It takes information gathered during IRC, IM, web browsing, e-mail, and more, does a lot of backend cluepacket mojo, and returns a lot of useful information while you work. If "bug 1565" comes up during your work, it'll fetch information in dashboard about the bug without needing you to click on a bug link. Microsoft is working on the same thing, called "implicit query" or some such. Look at the Windows Longhorn screenshots so far... It looks like they are taking the classic IE information sidebar and altering it to work in this way.
Interesting idea. And if you can read the intent behind each sentence then you can build on this to provide personalized or customized content to the reader based on text read off the page and compared to the user profile. I've been thinking about a tangent to this for a long time now, but can never figure out how to actually build it.
It's the idea of each thought being an atom of the content tree, captured in either a sentence, sentence group or paragraph. If each thought is a unique object, then each can be dynamically associated with relevant metadata or hyperlinks, and also manipulated to provide the user with the level of detail or level of expertise that's appropriate for them.
The example I use is a textbook on a common subject...say, microeconomics. The "thought" of price competition could either be encapsulated in a simple way for a survey-level audience or dynamically substituted to a more complex version that provides better information to better education audiences (e.g., "price competition varies by industry structure, where the degree of product homogeneity largely determines the pressures placed on pricing" for mid-level audiences and "price competition is the ability for each firm to impact the price that other firms in the marketplace can charge" for an entry-level audience). In this case, the system would need to understand both how to understand the content object and then how to articulate it back to the user in understandable terms.
We're getting there...but I don't think there's enough base technology to launch this idea yet. There's a lot of AI involved, particularly when it comes to the varied characterstics of the audience (people don't tend to come according to standards...). Maybe I'll do something fabulous with this in 10-20 years or so...
---- Please be nice in case my Slashdot karma ~= my real life karma.
From the site: "Works seamlessly with Internet Explorer--keep browsing just as you do now"
Uh, I use Mozilla Firefox because it embraces the current standards, especially CSS.
"custom CSS tags within the HTML"
I hope this means a custom external stylesheet, not invalid markup within the page; their site isn't exactly using the current standards or embracing CSS either.And, most importantly, at least try to go through the system (W3C) before resorting to custom markup such as this. How does this relate to the Semantic Web? Have they gone through the process of presenting this as a new standard or improving upon a standard? I doubt it.
So far as I'm concerned, this just feeds into the "sound-bite" culture vortex that television has been sucking us into for the last 2 decades. Why do we feel the need to strip the nuance and subtlety from everything?
This study seems to confirm what I've always thought about our soul-less Info-culture. I love technology, but we need to be careful that it doesn't strip away our humanity.
The meek shall inherit the earth, in 3 by 6 plots. - Lazerus Long
A sentence is basically a linear construct. But the way our brain processes a sentence is non-linear. We skip forward, we refer back. We process the sentence in this "looping" fashion until we comprehend it (or not, sometimes we just read it in linear fashion and move on to the next sentence in hopes that it will provide more context for us).
Read any good sonnets lately?
I did a little digging into our poster "rtmyers."
A google search on his email gave me this page, which reveals to us his name.
http://mail.gnome.org/archives/balsa-list/2001-MaThen I looked up who registered NaturallyOpen.com at Register.com.
WHOIS lookup on Register.comSurprise, Surprise... the same guy.
He could have told us in the post that he wrote the thing.
I don't know what these guys are doing, the blurb on their site is funny but I don't want an IE plugin. However, it is possible they are a couple computational linguistics grads..
I did a survey of literature, coming at it from a layman developer's angle, and it seems the one area of natural language recognition (hence their name naturally open?) where computers are trustworthy and even exceed humans, is pronoun extraction. Not semantic recognition where meaning is understood, but just getting the who/where/what of proper nouns and being able to also link pronouns to them correctly. It's somewhere around 95% accurate and apparently better than a human volunteer in average accuracy, in one test.
This is accomplished not by dividing into sentences but looking at passages of multiple sentences. Perhaps theirs does some of this too, but even a very simple product searcher could just look for words not in its dictionary and google them. So it is not obvious what the merits of their approach are. Personally I'm interested in text-based interaction and news retrieval with open NLP tools.