Normalizing ratings

As platforms and organisations use a wide variation of rating systems we need to normalize ratings to get consistent rating values over all sources.

To get this done we have a very simple rating schema. This means indexed content needs to be translated into the following schema:

value

label

description / example

3

No issues: Content is rated as not problematic. Credibility high or very high.

The reviewed item is rated as mostly true or true.

2

Neutral: Whether positive nor negative rating. Neutral credibility.

The reviewed item is not rated as true, mostly true, mostly false or false.

1

Questionable: Content is rated as questionable (at least for some aspects).Credibility low or very low.

The reviewed item is rated as mostly false or false.

n/a

No rating

There is no rating.

Mapping example for climatefeedback

reviewRating.alternateName climatefeedback

Hoaxly normalized rating

> 0.5

3

-0.5 - 0.5

2

< - 0.5

1

n/a

n/a

So if an article on climatefeedback is rated with 0.5 we want to have it saved in hoax.ly index with an normalized rating of 2 (neutral).

Rating json schema

See the following schema for an description how ratings should be presented in the hoax.ly index:

"ReviewRating": {
"type": "object",
"properties": {
// THE ORIGINAL RATING
"originalRatingValue": {
"type": "integer"
},
// THE ORIGINAL ALTERNATE NAME
"originalAlternateName": {
"type": "string"
},
// THE LABEL FOR THE NORMALIZED RATING (Questionable, Neutral, No Issues)
"alternateName": {
"type": "integer"
},
// THE BEST RATING VALUE (IN OUR CASE ALWAYS 3)
"bestRating": {
"type": "integer"
},
// THE WORST RATING VALUE (IN OUR CASE ALWAYS 1)
"worstRating": {
"type": "integer"
},
// THE NORMALIZED RATING VALUE
"ratingValue": {
"type": "integer"
}
}
},