In this article
You perform text analysis to discover your respondents' feelings or sentiments concerning your product or service. The table uses a reusable 11-point scale, so the codes are 1 through to 11. Create the scale in the Reusable scale/list folder and call the scale SentimentScale.
It is recommended that the reporting is set up using a 3-point scale (-5 to -1 = Negative, 0 = Neutral and 1 to 5 =Positive) or a 5-point scale (-5 to -3 = Negative, -2 to -1 = Slightly Negative, 0 = Neutral, 1 to 2 = Slightly Positive and 3 to 5 = Positive). This is set up in the report configuration.
You can set up the sentiment scale for both of the sentiment configurations (go to About Sentiment Analysis for more information) in the same way. The older sentiment configuration will give you scores for all 11 codes (i.e. from -5 through 0 to +5) - . The newer Deep Learning configuration will only give results at -5, -3, 0, +3 and +5, as this works on a 5-point scale. We recommend trying the Deep Learning configuration first and only going to the original sentiment configuration if your results are not good enough.
Tip
If the question you are analyzing contains sentiment (for example "What do you like about our company?" or "What went wrong with your transaction?"), the sentiment will not work well for your comments. To achieve more accurate sentiment results, questions should be neutral (for example "Tell us why you gave us that score").
Figure 1 - The SentimentScale reusable scale
This SentimentScale reusable list will be used as the answer list in the OverallSentiment (Single) variable and the Category/Sentiment (single) variable in the LoopReference variable .