e diel, 3 maj 2009

Comment system

Comments are a fast-growing part of web content. Often, the comments on an article are more interesting than the article itself. Crappy comments and spam crowd out good comments, though. Comment software often allows people to rate comments and mark spam, but this still leaves a lot of duplicates. Usually, there are about five points made in all the comments for a given article, even when there are 1,000 comments. We need CommentMeme. Actually, we need a new way to write on the internet.

You should be able to highlight any part of a comment and rank it up or down. higher ranked parts would become more prominent. This way, you would see the best comments first. This would still leave a lot of duplicate comments, though. You would need software to recognize that the comments are all about the same thing. Or you could let people merge comments that are the same. For example, you could drag one comment onto another to show they're part of the same meme. You could also hide lower-ranked comments. For example, if one comment said "Firefox is the best because it is open source," and another said "Firefox is the best because it is the most customizable," readers would recognize that they are both about the same thing. They could vote up whichever one was a better way of putting the idea. Then the higher-voted comment would become more visually prominent than the lower-ranked one. One way to do this would be to show each group of comments on a certain topic, like Firefox's extensibility, as a stack, like a plain 3d cube. All the topic-cubes would be laid out like a city. The largest cubes would be the ones with that the most people agreed on, as signified by the number of "likes" and duplicate comments on that same topic.

The other important fix is to make comment "liking" more granular, so you can like a paragraph, a sentence or even a phrase by highlighting it and

you would end up with a kind of automated, crowdsourced editorial. for example, the website buzzilions.com aggregates customer reviews and extracts phrases that are often used in reviews of a certain product. For example, the review for one monitor lists "sharp image" (said 11 times) "vivid colors" (also said 11 times.) In this way, you get an automatically-generated summary as well as more details in the customer reviews themselves, which are ranked from most helpful to least helpful. So right now, the best CE reviews are written by consumers and edited by software.

I got this idea from looking at whitehouse2.org, where policy suggestions, called priorities, are listed and ranked like articles and pictures are on digg. I said "Consolidating priorities could not only eliminate duplicates but could also be used to make each priority more complete and interesting. For example, there is "Invest in clean energy and create 5 million new green jobs and "Subsidize Green Energy like Solar and Wind" These are asking for the same thing, but in different ways. It's interesting that the first one is endorsed 3:1 and the second one is split. You should merge the two priorities and rank the descriptions, so there would be one priority called "Invest in clean energy and create 5 million new green jobs""

Nuk ka komente: