The categories detected into the speech
The emotions detected into the speech
Any filter chosen by the customer
To search for the main causes of evolution (in positive or negative fields) between two periods for the key indicators.
One of the main difficulties we encountered is that the notion of comment and categories are different :
Each comment may have several categories, but only one e-index, one CSAT (Customer satisfaction), one Recommendation score.
When we calculate the comment e-index average AND the category e-index average, the calculation cannot be separated (some categories may have one or more comments which have one e-index).
Therefore how to extract which categories have more impact one the results ?