Text Sentiment

Description

The Text Sentiment step scores the sentiment of a text column. It is the equivalent of the matching Alteryx tool and runs in the workflow engine (the ML / NLP / document libraries run in the executor job); results are an approximation of the Alteryx tool pending validation.

The Text Sentiment step scores each row of a text column, adding a numeric sentiment score and, optionally, a categorical positive/negative label.

Source & Options

• Source / Output tables — the input data and where the result is written.
• Text column, sentiment thresholds, and whether to emit a categorical label.

Sentence-Level Scoring

Find sentiment at sentence level splits each record’s text into sentences on its punctuation, scores each sentence, and returns the mean of those scores instead of one score for the whole body. It is the equivalent of the matching Alteryx option, and it is off by default.

This does not change the number of rows — one record in, one record out, with an averaged score. It only changes the result where a record holds more than one sentence: on a single-sentence record the average over sentences is the whole-body score, so the two settings agree. Text with no punctuation is scored as one body, because there are no sentence boundaries to split on.

Columns

The Columns tab maps the input columns passed through to the output.