Site owner here. A few thoughts:
- HN has been pretty hostile towards Summly and how it is "trivial" or "basic" to create an app that summarizes the news. I've always wanted to try. All of the negativity actually motivated me to think "Why can't I make something similar."
- This has been an incredible learning experience for me.
- Feedback is more than welcome.
Isn't it always going to be more efficient for the producer of said text to produce the summary him/herself? Couldn't resources be better spent trying to influence the production process of various news outlets to provide summaries?<p>In the legal community (and I'm sure many others), there is tremendous benefit to writing concise introductory and concluding paragraphs, as well as tables of contents that act as excellent skeletons for much longer documents. In policy-land, the one-pager is king...but I digress.<p>I guess I'm just kind of lost as to how or when the cost of <i>accurately</i> summarizing text by computer is cheaper than basically "asking" the author to provide one. Will the quality of a computer-generated summary ever be >= the quality of an author-generated summary?
Speaking only for myself, I want a few different things out of a service like this.<p>For one, I want an article fingerprinting technology. One that can tell that multiple different sites are talking about the same original post, and not really saying anything that is materially different. Maybe they all just cut-n-paste (which something like Churnalism would hopefully address), but I also want to catch the sites that add a little unique content to an article, but not enough to make a real difference. Link analysis would have to be factored into this, based on the full expanded URLs -- Sometimes there are new articles that come out with additional information on a topic that has previously been discussed, and I wouldn't want to miss those.<p>Second, once you have the fingerprint for each article from each site, you need a fuzzy way to compare them for uniqueness -- I want to do a "sort -u" on all news articles, based on the fingerprint.<p>Third, I need a way to tweak the scores and settings, so that articles from a high quality site like Ars Technica gets rated better than a lower-quality site. Of course, a certain amount of automation can be used here to generate default scores and settings, but I may have a different idea of exactly what scores and settings I want to use as compared to someone else.<p>I do like the idea of taking input from sites like HN as an additional variable for the positive or negative weighting of a news story (or a particular news site), if the article in question is one that has recently been discussed there.<p>Of course, you also need the concept of pluggable modules, so that when the next new thing comes out (like Churnalism), it can be quickly and easily added to the mix.<p>I don't suppose this sounds remotely familiar to anyone? I've got a bunch of feeds that I watch, but there's a lot of duplication and I would dearly love to be able to filter out that chaff while still allowing through the occasional unique article from those sites that usually just jump on the bandwagon long after the horses have escaped the barn.<p>Thanks!
I'm interested in how you decide what sources to use, and then what subject to summarise, and then how each story is summarised.<p>Please don't take this the wrong way but: it's a list of sites that I dislike so I wouldn't use this service. I can, however, see the value, and I would use it if it was sites that I was more interested in.<p>I guess for a tech crowd I'm a bit confused about this and RSS: Why don't people just use a better RSS reader?<p>But for non-technical people who need to keep up with a few different websites this could be great. Once you get v1 sorted you could think about adding some kind of voting for v2. "Useful [y][n]" "important [y][n]" etc.<p>Good luck with it if you do decide to do any more with it.
<a href="http://news.thetechblock.com" rel="nofollow">http://news.thetechblock.com</a> does an outstanding job with relevant tech news in my opinion. It's hand curated, though. There's also no summary. UI is great.
Looks good but tldr.io exists. Feedly now has, what comes down to, a rating based on how many times an article was saved. It just needs a tldr.io layer over it.