I know why that's happening, but I unfortunately can't fix it due to Rails Rumble rules. I'll be overhauling the content extraction over the next little while using what I've learned from stuff you guys have submitted. Example: Did you know a lot of sites don't actually use <p> tags? Or that some sites have a metric ton of JS (many times more than the HTML in the <body>!)? It's awesome! :(
Unfortunately there isn't anything near perfect as far as I've found.
There are a lot of clever things you can do, but markup is incredibly hard to parse (in the "What's actually in here?" sense) and it's even harder to determine the original author's intent for elements. HTML 5 sites makes this much easier, but that's like 0.00000001% of pages on the Internet at this point. :/
I made a similar site a year or so ago called synopit.com but that relied on user generated content rather than algorithmic summarizers. I had some good traction early on, covered by readwriteweb and a few other decent publications. Me and a buddy were generating most of the content, usually of popular stories on reddit or digg, to which we would link in the comments. That generated a good amount of traffic, but we kind of got busy and then let the idea fall by the way side.
I still think user generated content is the way to go. A good example is this summary posted in the comments of the Ars Technica review of the Windows 7 phone: http://news.ycombinator.com/item?id=1818782 - that kind of summary just can't be programmatically generated (yet, anyway). I'm glad tlrd.it is having such early success, it kind of validates the market for me and may inspire me to pursue my idea further.
5 comments
[ 460 ms ] story [ 485 ms ] thread->
http://tldr.it/summaries/7744
Short, Medium, and Long show the same short piece of insignificant text.
I know why that's happening, but I unfortunately can't fix it due to Rails Rumble rules. I'll be overhauling the content extraction over the next little while using what I've learned from stuff you guys have submitted. Example: Did you know a lot of sites don't actually use <p> tags? Or that some sites have a metric ton of JS (many times more than the HTML in the <body>!)? It's awesome! :(
There are a lot of clever things you can do, but markup is incredibly hard to parse (in the "What's actually in here?" sense) and it's even harder to determine the original author's intent for elements. HTML 5 sites makes this much easier, but that's like 0.00000001% of pages on the Internet at this point. :/
I still think user generated content is the way to go. A good example is this summary posted in the comments of the Ars Technica review of the Windows 7 phone: http://news.ycombinator.com/item?id=1818782 - that kind of summary just can't be programmatically generated (yet, anyway). I'm glad tlrd.it is having such early success, it kind of validates the market for me and may inspire me to pursue my idea further.