How to analyze text data thematically or categorically with AI
Below is a guide for how to work with an AI research Co-Pilot to analyze text.
Analyzing text with AddMaple's AI Co-Pilot is built to follow the same process as working without AI or collaborating with a human research partner.
The process in brief is described below:
When analyzing text, we read the text data and decide on codes or categories to group the text records into. From there we might refine the codes, merge a few, get rid of some, add new codes and then apply the new codes to all records. You can use AddMaple's AI co-pilot achieve this.
This guide takes you through the steps to analyze text by coding text records with AI:
Set up & getting started
Generate codes with AI or add your codes
Refine codes and apply them to individual records with AI
Review the analysis, add & remove codes to records as necessary
Use AI to apply newly added codes to the rest of the records
Explore statistical Insights
Share and export your findings
To analyze text, you need to find the column containing the text data using the search bar on the dashboard. Check that it is detected as a Text column. See the tip below for what to do if it isn't. Click on the chart to expand it.
When you expand a text column, you will see an Interactive Word Cloud. To analyze the text data you need to click on ✨AI Coding / Tagging
Tip: Changing the column type to 'text'
If your text column isn't automatically detected as a text column, you need to first change the 'column type' from the Chart Dashboard. See here for details on how to do that. Why does AddMaple occasionally show text columns as a 'multiple choice' columns? This is to automatically show you duplicates in the text data, highlighting potential copy/paste responses to help you detect potentially fraudulent or fake entries during cleaning. Also we pull out those mid dot responses!
The start screen for AI text analysis
Next you can choose whether to use your own codes or to let our AI generate codes for you. Using your own codes is useful for when you have an existing codebook you've used for a previous project.
If you choose "Get codes with AI," you'll be taken to a screen where you can guide our AI on the type of codes you'd like to generate. Enter your custom instructions in the white box, or select one of the suggested prompts on the right.
While this step isn't mandatory, it can be helpful in ensuring the generated codes align with your research question. For example, you might want to focus only on bugs or extract all the brands mentioned.
Tip: AI-generated codes will include descriptions and examples that you can refine in the next step. These help improve accuracy when applying codes - descriptions clarify when to use each code, while verbatim examples ground the AI in your actual data language.
With AddMaple, you can create multiple analysis columns from a single text column. This flexibility is especially useful when coding a column from different perspectives to get richer insights.
On this screen, you can also specify the number of codes you'd like to generate. Choosing a smaller number, such as 5, will produce broad themes, while selecting a higher number, like 30, will generate more detailed, fine-grained codes.
When you are happy with your settings, click "Generate with AI".
If you choose the "I have codes" option, you will be taken to this screen where you can type or paste in your existing codes.
You can enter up to 50 codes - each should be on it's own line.
When you are ready, click "Continue" and you will be taken to the review page where you will still have a chance to edit your codes before applying them to your data.
Once your codes have been generated (or manually added), you can review and refine them.
Generated codes now include descriptions and examples that you can edit to improve accuracy:
- Code Title: Edit the name of the code (keep it to 5 words or less)
- Code Description: Add a 1-3 sentence description to clarify when to apply this code
- Verbatim Examples: Include short example snippets from your actual data to ground each code in real user language
You can also merge similar codes, delete any that aren't relevant, or add new codes by clicking "Add code."
After refining your codes, name your analysis column. We'll suggest a name based on your codes, but you can edit it or keep the suggested one.
Finally, choose the type of operation:
**Apply One Code Per Record: **For categorical operations, e.g. sentiment analysis
Apply Multiple Codes per Record: For thematic coding operations - you can also set a maximum number of codes per record.
6Review codes
To merge codes:
Select the codes to merge (you can hold down shift to select multiple codes in one go)
Click the "Merge Selected" button
Select the name to use for the merged codes (you can use one of the existing names or enter a new name)
Click "Save"
7Merge codes
Once you're satisfied with these settings, click "Apply Codes" to complete the process.
We'll now start applying your codes to each record, which may take anywhere from 5 seconds to 5 minutes depending on the number of records.
You can close the pop-up and continue using AddMaple while this process runs in the background, or wait for the results to complete.
Once coding is complete, you'll see a summary of the results. From there, you can choose to either explore the counts of each code in our pivot chart or review the data row by row, with highlighted text showing where each code was applied.
Each time you perform an AI Coding operation, we will create a new column for your results. You can explore this column, like any other in AddMaple.
This is the Pivot Chart view and from here you can:
View a Pivot Table
Get an AI Explanation
View related columns
Rename, merge or delete codes
To view how the codes were assigned row by row, click the "View text and codes" button on the left.
In this view you can explore how AddMaple has assigned codes to each text record.
AddMaple highlights the relevant section of the text response that relates to the assigned code.
To add a code, simply select the relevant text and a pop up box will appear. Select an existing code, or enter a new code, then click "Add Code".
To delete a code from a text record, hover over the code and click on the trash icon. The code will now be removed from that particular record.
You can delete a code from all records (or merge it with another code) from the chart view.
If while reviewing the text and codes, you discover that there is a missing code you don't need to manually assign it to all your records. You can head back to the chart view by clicking the chart icon at the top of your coded column and use AI to apply your new codes to the rest of your records.
Your existing codes won't be changed, AddMaple will simply check if your new codes are relevant to any of the other records in the dataset.