Sentiment detection
This action node uses the 'Vader' sentiment detection. It was created by Hutto and Gilbert (2015) and is a dictionary based method that offers different scoring options for text analysis26. It makes use of a text and an emoji dictionary and offers negation checks and sentiment modification by certain keywords such as ’but’. The result is improved further by score dampening or amplification given punctuation or all-caps information. Vader is unable to detect stance, during text analysis the system maintains scores for positive and negative sentiment that may be merged to a compound score.
The results and sentiment metrics where obtained by using Vaders’ compound score which maps to a value in the interval [−1..1] for an input text.
This compound score is used to determine whether the input text is considered negative (lesser than or equal to -0.05),
positive (≥ 0.05) or neutral, according to that the corresponding outgoing edge is taken.
Variables​
The Vader scores are set as variables for further processing.
| Variable | Description |
|---|---|
| sentiment.polarity.compound | the aforementioned combined sentiment |
| sentiment.polarity.positive | positive portion of sentiment ≥ 0.0 |
| sentiment.polarity.negative | detected negative sentiment ≥ 0.0 |
| sentiment.polarity.neutral | neutral portion ≥ 0.0 |