Speech analytics and quality · L2 · 15 min

Sentiment is a label

Learn in five minutes

Wikipedia defines sentiment analysis as identifying affective states and subjective information.
A basic task it describes is polarity: positive, negative, or neutral, for a document, a sentence, or an aspect.
The page says telling subjective text from objective text can be harder than polarity, and that context can change the class.
It lists challenges that include metaphor, differences in writing, and context.
A label on a call is a classifier output. It is not a fact the caller stated.

If the sentence is a fact about a balance, do not store a polarity word as if the caller had said it.

Take the sentence 'the card has no fee' and mark it as a fact, not a mood.
Take 'I hate this hold music' and mark a negative polarity, then write 'classifier label'.
Add the Wikipedia caution: context can change the class.
Do not average the two into a customer-satisfaction score.

Polarity: A positive, negative, or neutral class. Wikipedia lists it as a basic sentiment task.

Sisters: Automatic speech recognition, Sentiment analysis, Quality rubrics, Personal data.

The live voice pipeline answers the person. Analytics reads a recording later. Keep those clocks apart.

A later review can add a rubric you wrote. It cannot borrow a customer's tape.

The answer looks fluent and is wrong. Check whether a source passage actually contains the claim.
The loop never stops. Set a step cap and stop when the tool result repeats.