zenloop's sentiment analysis helps you to understand your customer's mood by identifying if the overall comments and Topics are viewed positively, negatively or neutrally.
But how does it work?
The flow is as follows;
1. Customer leaves a comment:
"Super product, super quality, but too high price"
Keyword: price → Price
Keyword: quality → Quality
3. The full comment is analysed for sentiment. This results in the Sentiment per comment:
4. The comment is then checked for where the keywords can be found and then uses our separator rules to break up the sentence. For the below example, the separator is the comma;
Super product, [super quality], [but too high price]
5. These sections are individually analysed to form the Topic Sentiment:
The Separator Rules are:
If comment contains one sentence → we use the comma as a separator
If comment contains multiple sentences → we use the following punctuation marks as separators;
Additionally, we use the following stopwords as separators:
In the cases of
, dass → the comma won't be treated as a separator.
The comment and Topic sentiment is analysed independently of the score. For example, an NPS score of 9 has no bearing on the sentiment analysis
If none of the separator rules can apply, no Topic sentiment will apply (e.g. no punctuation marks)
There are some words or phrases that will result in a default sentiment being assigned to the relevant Topic.
This means that when these expressions are detected within the comments, the sentiment is automatically assigned as follows:
leider etwas teuer
wie erwartet angekommen
super vielfältige Auswahl
The expression exceptions will still apply, even if there are no separators present (e.g. when there is no punctuation in the sentence)
The assignment of the automatic sentiment is based on the word stem. For example, cheap price=positive would also work for cheaper price
For more information on Topics, stems and keywords - click here.