# How Chattermill Defines Overall Sentiment

Written by Kesi Kagbala  
October 16, 2025

## 🧠 How Sentiment Is Calculated

Sentiment in Chattermill is generated by a **machine learning model** trained on **labelled feedback text**.

This model looks only at the **written comments**, not the numeric scores (NPS, CSAT, star ratings, etc.), and assigns sentiment to **each theme mentioned** in the text — for example, _Delivery_, _Customer Service_, _Pricing_, etc.

That means a single comment can include a **mix of positive, neutral, and negative sentiments** depending on what’s being said about each theme.

There isn’t one “unified sentiment” per feedback item.

## **🌟 How Review Scores Are Interpreted**

Numeric ratings such as scores or NPS values are treated separately from text sentiment.

They’re often used for reporting or comparison, and are typically grouped into these tiers:

|     |     |     |
| --- | --- | --- |
| **Rating** | **Sentiment Tier** | **Typical Mapping** |
| ⭐ 1–2 | Negative | Dissatisfied or critical |
| ⭐ 3 | Neutral | Mixed or ambivalent |
| ⭐ 4–5 | Positive | Satisfied or enthusiastic |

This mapping is standard across most datasets unless a client has custom rules.

It helps make **ratings and text sentiment** easier to compare in dashboards, even though they are calculated independently.

## **⚖️ Handling Rating–Comment Mismatches**

Each theme gets its own sentiment; the system doesn’t try to create a single “overall” sentiment label.

Instead, it’s possible to calculate a **Net Sentiment** value — the **average** of all theme-level sentiments, where:

- Negative = −100
- Neutral = 0
- Positive = +100

So, a piece of feedback with both positive and negative mentions might end up with a Net Sentiment close to 0 (neutral overall).