Summly: intelligent news summarisation
summly.comSorry, this isn't rocket science at all.
Standard clustering algorithms (found in any off-the-shelf natural text processing library) and text summation with libots should suffice for most of the heavy lifting.
http://tldr.it/ http://libots.sourceforge.net/
Further, most news articles' first paragraph is a practical (although you may have not noticed) summary.
Coming from NLP, unless you can influence the source and the source being Web, the story should be an 80%-20% in the best case -- and you'll work VERY hard to correct the remaining 20%, and you WILL remain with a percentage of content you just can't summarize properly.
What would make a difference is a real people-driven summation, not machines (see what voicebunny did for text-to-speech, for example). And yes, it would have been fun to combine the two as well.
I experienced an article in TheVerge which is mainly a video as its content.
What I will be amazed is a good automatic summarization algorithm that is using abstraction and not just extraction.
Also, check out circa (http://cir.ca/). Never tried, but as I read, it uses both human and algorithm to "summarize" articles.
circa is a good idea, actually. From my close experience with this field, when a news article will be published it will be edited and republished many times, over many forms and shapes (Web, RSS, etc.) in many of these steps, a manual, human work is needed -- and this affects the volumes of the published news.
Further, many of the news really originate from relatively limited sources (reuters, etc), so you can plug your solution there as well.
Therefore it should be OK to assume that if you put humans at the same pipeline to summarize news manually, the capacity and efficiency will be reasonable.
The problem in summarizing news manually is that it takes too much effort for a human to do it. The efficiency may be good, but as many news pass by, his efficiency will go down. (assuming that he's only the one summarizing)
True, but my point is people are already doing it at the start of the pipeline. Think what happens when Reuters decide to make a SaaS offering of their summarized content. Even regardless of that, you can hire a battery of professional summarizers instead of PHDs and do it pretty well.
Where this doesn't apply, and where I do think you're completely right is non-news articles: think blogs, tweets (although there's not much to summarize in 140chars), product descriptions, scientific articles, etc. These things are produced in much more volume and much less workflow around them.
From the looks of the website, the video and what the apps does, it looks like the idea got too much funding.
Cmo'n they even have a famous person in their clip!
Most news sites already give a summary in an RSS feed. This is most of the time sufficient.
I think that a good idea should sell itself, not with all this fancy pancy.
-- Facts: 17 year developer got a $250.000 funding http://thenextweb.com/apps/2011/12/17/meet-the-16-year-old-w...
As a nerd I find much of the daily news to be rather trying. Much of it is of little interest and impact on my day to day life. I send a lot of time reading about technology and developments in our industry and almost no time on regular news.
That said, as someone who likes to be well rounded, I think this is a woefully inadequate state for myself. I'd like to be aware of what is happening in the world, financially, politically, and internationally. I yearn for a concise distillation of the torrent of news that most sites provide. I don't have time for nor desire to read all of that.
Right now I've got the following in my feed reader to try and fulfill this:
http://www.themorningnews.org/ - two times daily, kind of a lot of content, no summarization just links http://evening-edition.com/ - dialy, summaries of 4-6 of the days top news, I like this very much http://thebrief.io/ - tech oriented top news, seems to be delivered somewhat irregularly
I'm not overly happy with this, I might like a second take from an alternate source similar to the evening edition. The brief is delivered somewhat irregularly and I find the morning news to be just generally lacking, but enjoy some of the frivolous fun that they have with it. Perhaps an evening edition that has a little fun... hmm.
Given all of this, I'm quite interested in this service.
Sorry to drive this off topic. I run The Brief, and to date I've published every week day by 10am, with the exception of two days where I had to delay it to later in the day due to a hurricane doing its best to destroy my city. What is it about that schedule you find irregular? I welcome any feedback. Feel free to email me (the address is listed on the site).
Orofino, I run http://skimfeed.com. It's a news dump site, currently in beta. Give it a shot.
It looks like the 17-year-old got enough funding to hire a proper AI/NLP lab to do his algorithm - reading between the lines, it's difficult to say how much input the founder had on the new product: my guess would be concept, maybe design, and possibly interface coding.
yes it appears the tech was done by SRI. still doesnt take away from the idea and execution though.
http://gigaom.com/2012/10/31/summly-wants-to-make-news-summa...
Fair. I know just enough AI to realize that this is a hard problem, so the whole news-friendly "genius self-taught teenager builds awesome app in his spare time" seemed a bit incredible - I still wouldn't want to be competing against this kid for a job, though, he's impressively smart.
Agree on everything, just wanted to point out that Stephen Fry said he wasn't paid: https://twitter.com/stephenfry/status/264066858225905664
For a majority of news articles, the first three sentences is usually more than perfect as a summarization method -- feature stories on the other hand are insane (and that's why we have things like Lexical Chains: http://acl.ldc.upenn.edu/W/W97/W97-0703.pdf)
We use lexical chains in our undergraduate thesis. Not a fan of it. AFAIK, not much of the text we summarized are affected by our lexical chain feature. It is a minor feature in our thesis. We combined lexical chains with other summarization techniques.
It's a good concept, but chains maybe a long one and may not be serve its "summary" purpose (maybe that's only for our case). In our thesis, we just extract the 5 most important sentences. Sentences with a lexical chain to a sentence or sentences may consume all the 5 sentences we needed and not leaving a room for other sentences in the text.
Is he doing extraction or abstraction? I'm doing a research about automatic summarization and also creating an app for it. Summly and my idea has a different approach but I'm curious in some aspects of how summly summarizes articles.
I think last year, Summly was pulled out in the app store. Back in those days, the quality of there summary is not that impressive. I'm still downloading the app and will check its performance.
I found the answer to my question. It is still extraction. Doing abstraction is still a difficult problem to solve.
the app's summaries are intelligible sentences, but since they're extracted from articles without providing any other context besides headlines, they often sound disjointed and non-conclusive. http://www.theverge.com/2012/11/1/3583720/summly-nick-d-aloi...
Here's [1] an interview Nick gave just a couple of hours ago on Bloomsberg.
[1] http://www.bloomberg.com/video/slacker-16-year-old-whiz-laun...
Relevant news story about this from December 2011: http://www.bbc.co.uk/news/technology-16306742
I think things have been fairly quiet for a few months pending this re-launch.
Off-topic, but does anyone know what sneakers Nick D'Aloisio is wearing in the video with Stephen Fry?
Looks like they have a slate gray upper, white soles, and a prominent logo on the tongue.
Edit: For anyone who's wondering, they're Gourmet Footwear[1].
Very impressive design / interface. The summaries help me decide which articles to read in full, as opposed to being abridged versions of the full stories. Still useful, but not a substitute for my Times Digest (http://www.timesdigest.com/).
They don't seem to indicate the cost without signing up for this, what do they charge you for the times digest? Also, is there an easy way to get it that isn't a pdf?
Sorry to hit you with a barrage of questions, but this is interesting and I didn't get much info from their site. IT seems that this is intended for larger entities to distribute to clients for the most part. They do have an individuals section, but without much info.
The Summly blog mentions about a demo to "let people enter a web site’s URL and receive a summary of the content."
Anyway to still use that? The idea sounds very exciting and an online bookmarklet can super-useful to TL;DR anything.
That would be really awesome. Overall I like this landing page, but my main question is "how good is the summary?" and "Can I trust the summary generated?"
Maybe non-technical people are just like, "Yeah! Awesome!" and that is all that matters. My thoughts are "How good is it? This is a hard problem and there is probably a lot of money in it. Why haven't other companies developed this?"
The app itself and the site look nice.
Very interesting. I'm on Android, so I can't actually try out the app, but I'd be interested to see what kind of compression ratio they get, how understandable it is, and if they hit all the major points. I'm also curious if they can create an intelligent summary of a cluster of related articles (say, the explosion of news recently about Disney buying out Lucasfilm), which seems like a problem that would be almost as hard as "general AI".
A friend of mine is doing a research about this. Clustering articles through the use of neural networks. But of course, it can only cluster articles that is in their collection or corpus.
Great application. similar to Circa in a way but I think Circa's user experience and notifications are better.
That is a seriously large iPhone at the bottom right of the landing page.
"To use Summly pls use Chrome, Firefox or Safari"