30 comments

[ 3.1 ms ] story [ 59.8 ms ] thread
Pretty interesting. Fully automated text summarization certainly isn't new (libots, Classifier4j, OSX's summarization service, Copernic Summarizer, etc.) but I haven't yet seen any implementations that use a machine learning approach.

EDIT: A better integrated summarizer obviously won't "revolutionize the web", though.

Somehow "May summarize the web up to 30% better!" just doesn't have the same ring to it.
I don't really see how 30% better is going to make a qualitative difference in the utility of text on the web.
I can't even think of any time other than raw search results that I read auto-generated summaries now anyway.

Of course, maybe if the summaries were just a little better..it'd revolutionize the whole web! :-)

does this all work based on picking up key sentences or can it rewrite content as well?
I think 30% more is pretty significant. Especially if applied to email, long articles etc.

However, still not sure if it ensures that i wouldn't miss out on anything important which may be fatal if it was an important email.

PLUS it doesnt say if it picks up on emotional context which is super important in interpersonal communications?

Doesn't seem like like it rewrites content, also seems to often use a sentence as the first one that refers to a previously introduced concept in the article that has no context in the summary.

This doesn't really surprise me as in the time he has been working in ai/ml/nlp I don't think there would have been much time to do more than learn up to and implement some of the current state of the art. The named entity recognition for the keywords seems okay.

I did a bit of reading on summarization during my thesis, enough to know it is a big and difficult problem. I hope he keeps at it, seems like a good start but a long way to go.

Measurement in summarization is very inexact, and there are no standards for measuring it.

I've experimented with machine learning and haven't had great results especially if you want a small summary.

I wrote Classifier4J, and the summarizer in that does pretty well - well enough that it has been ported to C# and as a patch for Python NLTK.

I bet Demand Media would salivate at using something like this. Pages with automated summaries of every page in the internet.
Isnt that what reddit and HN are. metas of info available to see if we want to click the bait.

The number of sites I visit directly over the years has diminished to 4 since reddit, HN and two others.

the issue to be solved for other sites to grow is not that they can summarize, but how well they curate a community around the results.

16 yo get $250k for technology that pushes us even closer to a Farenheit 451 distopian future.
Hypothesis: There is more to be gained intellectually by reading 10 novel synopses on Wikipedia that reading any one of the novels all the way through.

I'm aware of the dangers of soundbyte-ism, but it needn't be inherently evil. In fact, the internet is the best place for summarization to take place; you can't click through on a TV news ticker, but you can on a linkbaiting blog.

"Hypothesis: There is more to be gained intellectually by reading 10 novel synopses on Wikipedia that reading any one of the novels all the way through."

That depends completely on what your goal is. What do you "gain intellectually" when knowing the overall story? Aren't novels more about the experience you get while reading them?

For news I can understand summaries can be as good as the whole thing, especially if you're not that interested but want to know what is going on.

I tried the service and I am sorry to say that it is really just awful. I don't want to be mean, but it seems to pick a random sample of sentences. Great idea, poor execution.

Much of the media attention has probably been around the fact that he is 16. If Google engineers released this, they'd get laughed at.

That seems to be the state of a lot of coverage on tech companies. Age first, service second, quality third.
That sounds great! I'm 17, where can I sign up to get 250k
Unfortunately, I have to agree that it's not very effective. Even using it on Wikipedia articles, on which it has likely been trained, doesn't yield impressive results. It would be more interesting if it worked well.
I use an enhanced version of the summarizer from Classifier4J (mentioned elsewhere in this thread) for http://demo.qontex.com/.

There is no public API to show the summaries, but if you can find a Wikipedia page it doesn't generate reasonable ads for I'd be very interested.

Eg random pages from today's front page: http://demo.qontex.com/?url=http://en.wikipedia.org/wiki/Pre...

http://demo.qontex.com/?url=http://en.wikipedia.org/wiki/Rep...

http://demo.qontex.com/?url=http://en.wikipedia.org/wiki/Bri...

Australia - the homeland of successful emigrants. And Dick Smith.

OK, before I get slammed for being a redditor ...

Australia has had some brilliant inventors and inventions. Not a hugely disproportionate number (though we'd like to think so), but we don't do too badly. But it never goes from prototype to product.

I guess Australians are used to zero-sum thinking. It's all rent seeking, con artists, and conservative hard work - farmers, miners, government, business people (hustlers) and retail. There's no real need for innovative companies.

So while we have a few good ideas, no-one really wants to fund them. There's the occasional Ponzi scheme (generally backed by a good idea, but generally over-hyped), but the actual value created by innovative Australian stuff is relatively insignificant.

I think the problem is more on the side of investors than innovators. Investors here are more likely to be after less risky opportunities and want to see more revenue profit before getting involved. Compared to in SV where it seems like a good team/ idea can land you enough funding to make a real go of it without having to first bootstrap to profitability.
It won't just be investors. It will also be regulators (who will take a dim view of smaller operators), potential employees, suppliers, and so on.
Robots summarizing text produced by other robots. That's the future of the web? Sigh.
Umm .. this feature is already implemented and shipping .. in Mac OSX! Try this: select some text, right click, "Summarize" .. oila! Not news at all.