How to run a sentiment analysis in ATLAS.ti

Who this article is for

This article is for ATLAS.ti Web and ATLAS.ti Desktop users who want to automatically identify positive, negative, or neutral sentiment in text data and use the results to support qualitative analysis.

What is Sentiment Analysis?

Sentiment Analysis is an AI-supported tool that helps identify whether text expresses positive, negative, or neutral sentiment.

You can use Sentiment Analysis to explore tone, reactions, opinions, and evaluations in your data. It can be useful when working with:

• open-ended survey responses
• customer feedback
• product or service evaluations
• interview transcripts
• online conversations
• student evaluations
• participant reflections

Sentiment Analysis creates or suggests sentiment-based codings that you can review, edit, and use in later analysis.

Sentiment Analysis can help speed up your analysis, but it does not replace researcher interpretation. Always review the results and make corrections where needed.

Use Sentiment Analysis in ATLAS.ti Web

In ATLAS.ti Web, Sentiment Analysis is available from an open document.

Step 1: Prepare sentiment codes

Before running Sentiment Analysis, you may want to create codes for each sentiment.

Examples:

• Sentiment: Positive
• Sentiment: Neutral
• Sentiment: Negative

1. Open your ATLAS.ti Web project.
2. Go to the Code Manager.
3. Create a code for each sentiment you want to use.
4. Choose a color for each code if needed.
5. Add a code comment if you want to explain how each sentiment code should be used.

Step 2: Open a document

1. Open your ATLAS.ti Web project.
2. Open the document you want to analyze.

Step 3: Open Sentiment Analysis

1. Click Tools.
2. Select Sentiment Analysis.

Step 4: Choose sentence or paragraph analysis

1. Select whether ATLAS.ti should analyze text based on:
   • Sentence
   • Paragraph
2. Choose the option that best fits your analysis.

Use sentence analysis for more focused results. Use paragraph analysis when you want broader context.

Step 5: Run the analysis

1. Click Start the sentiment analysis.

Step 6: Select sentiment codes

1. Choose which code should be associated with positive sentiment.
2. Choose which code should be associated with neutral sentiment.
3. Choose which code should be associated with negative sentiment.

Step 7: Auto Code the results

1. Click Code Results to apply the selected sentiment codes.

Step 8: Review and edit the results

After the results are coded, review the codings in the document margin.

You can:

• delete irrelevant quotations
• remove incorrect codes
• add additional codes
• change the assigned sentiment code
• add comments to quotations

To delete an irrelevant quotation:

1. Click the quotation in the margin area.
2. Click the trash icon.
3. Confirm the deletion if prompted.

To edit a coding:

1. Click the quotation in the margin area.
2. Add or remove codes as needed.
3. Save your changes.

Use Sentiment Analysis in ATLAS.ti Desktop

ATLAS.ti Desktop includes ATLAS.ti Windows and ATLAS.ti Mac. In Desktop, Sentiment Analysis can be run on selected documents or document groups, codes or code groups.

Sentiment Analysis can identify:

• positive sentiment
• neutral sentiment
• negative sentiment

It can code either:

• sentences
• paragraphs

Step 1: Open Sentiment Analysis

• In ATLAS.ti Windows

1. Open your project.
2. Go to the Search & Code tab.
3. Select Sentiment Analysis.

• In ATLAS.ti Mac

1. Open your project.
2. Go to Analysis
3. Select Sentiment Analysis.

Step 2: Select the documents, document groups, code, or code groups.

Step 3: Choose the base unit

Choose whether ATLAS.ti should analyze and code:

• paragraphs
• sentences
• In ATLAS.ti Windows and Mac

1. Select Paragraphs or Sentences.
2. Continue to the sentiment options.

Step 4: Manage language models if needed

If you want to improve the results, you can install a more comprehensive language model.

• In ATLAS.ti Windows and Mac

1. Click Manage Models.
2. Install or uninstall the available extended model if needed.

Step 5: Run the analysis

• In ATLAS.ti Windows

1. Click Show Results.
2. Wait for ATLAS.ti to process the selected data.

• In ATLAS.ti Mac

1. Click Continue.
2. Wait for ATLAS.ti to process the selected data.

Step 6: Review the results

The results page shows quotations in the Quotation Reader.

• In ATLAS.ti Windows and Mac

1. Review each result.
2. Use the preview controls to switch between smaller or larger previews if needed.
3. Check whether the proposed sentiment matches the text.

Step 7: Apply sentiment codes

You can code all results at once or review each result individually.

• In ATLAS.ti Windows and Mac

1. To code all results, use the option to apply all proposed codes.
2. To code results for one sentiment only, use the option for that sentiment.
3. To code individual results, click the plus icon next to the proposed code.
4. Use the regular coding dialog if you want to add or remove codes manually.

When to use Sentiment Analysis

Sentiment Analysis can be useful when you want to understand how people feel about a topic, experience, product, service, or event.

Common use cases include:

• analyzing customer feedback
• reviewing student evaluations
• identifying participant reactions
• analyzing open-ended survey responses
• comparing positive and negative experiences
• exploring tone in interviews or online conversations

Sentiment Analysis works best with structured text data such as survey responses, evaluations, online conversations, and feedback.

Best practices for Sentiment Analysis

Review the results carefully before using them in your final analysis.

AI-supported sentiment analysis may not always interpret:

• sarcasm
• irony
• humor
• mixed opinions
• context-dependent statements
• discipline-specific wording

Use Sentiment Analysis as a support tool alongside manual review, memo writing, and interpretation.

Common issues and mistakes

• Treating Sentiment Analysis results as final
  • Sentiment Analysis can support your analysis, but the results should always be reviewed and refined by the researcher.
• Choosing the wrong base unit
  • Sentence-based analysis provides more focused results. Paragraph-based analysis includes more context but may create broader codings.
• Applying all proposed codes without review
  • Review the quotations before applying all proposed sentiment codes, especially when working with complex or nuanced data.
• Expecting AI to detect all tone and context
  • Sentiment Analysis may not fully capture sarcasm, irony, mixed emotions, or context-specific meanings.
• Forgetting to revise codes and quotations after analysis
  • After applying sentiment codes, review the results and adjust incorrect codings, delete irrelevant quotations, or add comments where needed.

When to contact support

Contact ATLAS.ti Support if:

• Sentiment Analysis does not open
• selected documents or document groups do not appear
• results fail to generate
• sentiment codes are not applied correctly
• the Manage Models option does not work as expected
• quotations do not appear in the results
• coding options are missing
• the tool behaves differently from the documentation

When contacting support, include:

• your platform: Web, Windows, or Mac
• your ATLAS.ti version if using Desktop
• screenshots or error messages
• the type of data you are analyzing
• whether you selected sentence or paragraph analysis
• a short description of your workflow