3 ms·

I am a data scientist and right now I am working on exactly this problem: automatically identifying good content; in my case among piles of user generated conte

by ma2rten 13y ago

I am a data scientist and right now I am working on exactly this problem: automatically identifying good content; in my case among piles of user generated content. The question is how do you define that? It's not what most people like (or at least say they like), it's not what most people share or reply to. I wonder if anyone here has a good idea.

mjklin 13y ago
HN and Reddit seem to work well with the voting model
- ma2rten 13y ago
  I think the success of both is based on a) niche audience b) manual moderation.
QuantumGood 13y ago
Good content depends on • Highlighting (usually, a great headline/share title); • The audience; • Content form; • The KPI. Conversions or comments? Visits or shares? Etc. • The content itself; • Incentivizing distribution. If you say that good content is "not what most people share or reply to" then what KPI (Key performance indicator) are you measuring by? For example, with a good influencer strategy, shares can be more important than outright clicks/visits, because getting your work to spread results in more clicks/visits down the line.
corresation 13y ago
Most people do not vote - a front page link on HN or Reddit will see 10 - 100x more impressions than votes (my own experience has been 100x+ on the low side). The more complex the material, and the slower it is to parse, the response rate drops dramatically further still because it will likely either be backed out of (ain't got time for that), or more likely when the user is done with it they have lost all context. Those magical tabs you find sitting in the background hours later. So you're left with the highest response rate being towards short, sweet material that ideally panders to a bias, allowing the user to shortcut even more. This is true even on HN where trends come and go, people trying to evangelize some bias based on nothing more than a title or a skim summary. All of this doesn't even count astroturfing and sock puppets - the lowest common denominator ascension of content already makes enough noise.

by ma2rten 13y ago

mjklin 13y ago

HN and Reddit seem to work well with the voting model

ma2rten 13y ago
I think the success of both is based on a) niche audience b) manual moderation.

QuantumGood 13y ago

Good content depends on • Highlighting (usually, a great headline/share title); • The audience; • Content form; • The KPI. Conversions or comments? Visits or shares? Etc. • The content itself; • Incentivizing distribution. If you say that good content is "not what most people share or reply to" then what KPI (Key performance indicator) are you measuring by? For example, with a good influencer strategy, shares can be more important than outright clicks/visits, because getting your work to spread results in more clicks/visits down the line.

corresation 13y ago

Most people do not vote - a front page link on HN or Reddit will see 10 - 100x more impressions than votes (my own experience has been 100x+ on the low side). The more complex the material, and the slower it is to parse, the response rate drops dramatically further still because it will likely either be backed out of (ain't got time for that), or more likely when the user is done with it they have lost all context. Those magical tabs you find sitting in the background hours later. So you're left with the highest response rate being towards short, sweet material that ideally panders to a bias, allowing the user to shortcut even more. This is true even on HN where trends come and go, people trying to evangelize some bias based on nothing more than a title or a skim summary. All of this doesn't even count astroturfing and sock puppets - the lowest common denominator ascension of content already makes enough noise.