Ranking

From P2P-Fusion

Jump to: navigation, search
Ranking
Scenarios: all
Communities: all
Importance: medium

Ranking is a relative evaluation tool, which is slightly different from rating.

Contents

Description

This tool is used for getting feedback from the user. While ratings need to be given on a fixed absolute scale, ranking is only interpreted relatively. This is much easier for the user to decide the relative value of the content instead of the absolute value.

Input

User can create a ranked list of instances in increasing or decreasing order. Sometimes the list only includes 2 objects, so the only information given is the object evaluated higher.

Ranking (or at least partial ranking) can be computed based on ratings or other known attributes of the object.

Storage

The explicit rankings can be stored or we can use some internal variables connected to objects and modify them after each ranking done. It is at least as hard to store in a decentralized way as ratings and comments.

Output

The final output is a huge list of ranked items, which is usually not displayed for the user, but it can be the input for recommendation, a basic profile, toplists, or automatically generated favourites.

Interface

Usually not displayed, only an input for other tools.

Prevalence

It can be used in areas where absolute value is very hard to decide, and so rating doesn't work perfectly. For example it is used in dating sites for deciding which person's profile is more interesting for us.

Technical problems

It's really hard to get useful information from rankings, because other users' rankings cannot be used as a person's own. For perfect results, a user should rank every piece of content at least once. Although, for group toplists more people's rankings can be used.

Social problems

It is usually used to decide personal preferences, so there are no social problems in this case. Using at a global or group level, it has all the problems of other social tools.

Implementation

Binary rankings can be used to build a decision tree about user preference, and there are similar mathematical solutions for the multiple case. On a user level it is centralized, but on group level it can be hard to handle in a decentralized way.

Existing examples

TODO

Application in Fusion

Simple binary ranking can be used to fastly learn user preferences, which is a good starting point for recommendations, profile building, and many more personalized solutions. It works the best when multiple items can be reviewed at the same time, like pictures and profiles.

Related tools

Personal tools