Dataminingandusergeneratedcontent - renepickhardt/related-work.net GitHub Wiki

get the prediction verified.

Many things can be predicted using data mining. The problem is that even the best algorithms don't achieve 100% precision. One good strategy is to ask the user for feedback in a very light way. Assume for example we want to predict which author supervised who. The prediction function would basically be a tanimoto coefficient. Instead of posting this to the wesite we could ask one of the to authors on the next login or during the next session (did you supervise...? / have you been supervised by...?) This principle can be used in many places.

user generated content

should be allowed by anybody but needs to be verfied by trusted users. users win trust by reputation and will be allowed to do more tasks the higher the trust is. Especially new content is displayed right away. Only if a user produces way to much content in a short time we change the method. we will still need a method for versioning.