Hacker News new | comments | show | ask | jobs | submit login

One idea I've had to help mitigate this problem is to develop a sort of "intelligence" score, where you weigh votes from high scoring individuals over votes from low scoring individuals.

The easiest way I can think of to do this is to use a Flesch-Kincaid grade level test, which uses sentence length and word length to estimate reading level. A quick and dirty calculation for any post might look like: (character count / word count) + (word count / sentence count) = Writing Level.

The problem, of course, is that it's only a tiny subset of users that actually comment with any frequency so you can quickly kill the democratic appeal of the site, but I'm frankly less concerned that my vote is actually counted and more concerned with seeing quality content on the front page.

* wiki - http://en.wikipedia.org/wiki/Flesch%E2%80%93Kincaid_readabil...




I actually like this idea. The idea of Flesch-Kincaid tests. There was a guy on reddit who analyzed the words on different subreddits and he came to the conclusion that the average word was 'because', and average word length was around 6.7

I would think we could modify the scoring system a bit where common English words are penalized, and uncommon words are weighted a bit more. The only problem I see is, if someone plans to post code, that's going to wreck their score. Another thing would be that most of our programming names/terminologies contain very few letters. So, an exclusion/inclusion list would definitely be needed.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: