---
title: Topic-Based Sentiment Analysis
category: Text Analysis
slug: topic-based-sentiment
blurb: Analyze open-text responses by topic and sentiment so you can see what people praise or criticize most.
order: 7
---
# Topic-Based Sentiment Analysis

Topic-based sentiment analysis helps you understand not just overall sentiment, but **which topics** are driving positive, neutral, or negative feedback.

Example: Sentiment might be strongly positive for product quality but negative for delivery speed.

![Topic-based sentiment walkthrough](https://player.mux.com/oQH1NdJy9J5wsxVCT8Kt3Ru3Vf15R7imJlqdwyEPhdM)

---

## Where to find it

1. Open a **Text** column from the dashboard.
2. Click **AI Coding / Tagging**.
3. Click **Get started**.
4. Select **I want to analyze themes by sentiment**.

---

## How to run topic-based sentiment

1. Choose how to define topics:
   - Let AddMaple generate themes automatically, or
   - Enter your own themes.
2. Set the number of themes (for example, 5 themes for a broader summary).
3. (Optional) add instructions to guide the model toward your research focus.
4. Run the coding process.
5. When complete, click **View Topic Sentiment**.

---

## How to read the results

In **View Topic Sentiment**, each topic is broken down by sentiment:
- **Positive**
- **Neutral**
- **Negative**

Use this view to quickly spot:
- Topics with mostly positive reactions (strengths)
- Topics with concentrated negatives (friction points)
- Mixed topics that may need deeper review

---

## Review evidence in responses

After the chart view, open the response table to inspect the underlying comments.  
You can check where sentiment was assigned in each record and verify that the coding matches the respondent’s wording.

This is especially useful before sharing results with stakeholders.

---

## Tips for better output

- Start with fewer topics (5-8) for executive summaries.
- Use more topics when you need detailed operational insights.
- If themes are too broad, rerun with clearer instructions.
- If themes overlap, provide your own custom themes.

---

## Related guides

- For a single overall sentiment metric: [Single Sentiment Score from Text](single-sentiment-score)
- For iterative theme coding workflows: [Thematic Coding (Iterative)](thematic-coding)
- For the broader AI coding workflow: [How to analyze text data thematically or categorically with AI](ai-codes)
- For choosing model depth: [AI Models](ai-models)

---

## Key points

- Topic-based sentiment lets you see how people feel about each theme, not just the overall sentiment.
- It helps you spot what is working well and what is causing frustration.
- You can move from the summary chart to individual responses to double-check the results.

---