Leaf-level Category Scoring
Leaf-level categories are categories at the lowest level of an application. Leaf-level category scores are called leaf scores.
When leaf-level categories contain include phrases, the leaf score is equal to the total number of matched phrases in the transcript minus the number of matched exclude phrases. The first round of scoring is done on each individual speaker turn, and each turn's score has a floor of 0 even if the number of exclude phrases exceeds the number of matched phrases. These turn scores are then combined for an overall transcript score. Leaf-level categories with no include phrases receive a leaf score of 1.
Once calculated, leaf scores are stored in the V‑Spark database and are converted into binary Match/No Match values each time Coverage scores need to be calculated for upper-level categories:
-
Leaf-level categories that contain no matches (and therefore have Leaf Scores of 0) maintain a Match/No Match value of 0 when the upper-level category scores are calculated.
-
Leaf-level categories that contain at least one match (and therefore have Leaf Scores greater than 0) are given a Match/No Match value of 1 when the upper-level category scores are calculated.
Exclude phrases
Exclude phrases in leaf-level categories work differently than exclude phrases in Category Filters sections. Exclude phrases added to Call Filters sections act as filters for entire calls, but exclude phrases in leaf-level categories simply cancel out matches for include phrases.
When searching for application matches in leaf-level categories, V‑Spark searches first for all include phrases, and takes note of which speaker turns contained a match, and how many matches were contained in each of those turns. V‑Spark then searches those saved speaker turns for the exclude phrases and subtracts the number of exclude phrases from the number of include phrases. This result is then entered as the Leaf Score.
Because any exclude phrase can cancel out an include phrase, regardless of whether the words used are related, it is important to use exclude phrases with care. We recommend that you design your exclude phrases to eliminate very specific false positives that might result from a more general include phrase (e.g. “-thank you for calling” vs. “+thank you”).
