Skip to main content
Chat with us

Search

Find your bearings

How to Run Sentiment Analysis on Open-Ended Responses

Emily Kelley
  • Updated

Sentiment Analysis leverages a Natural Language Processing (NLP) API or a Large Language Model (LLM) API to interpret and apply sentiment to unstructured data in open-ended responses. All clients have access to the NLP-based version by default; to unlock the enhanced LLM-powered sentiment analysis, AI and Skipper must be enabled for the account.

In This Article

1. Interpreting Sentiments

  1. From the Results page, click the Verbatim icon at the top to view all responses to the open-ended questions included in your survey. The colored bands along the left side of each comment box indicate the sentiment related to the response.
  2. Hover anywhere inside a response to see the corresponding emoticon, and click the assigned emoticon to view the spectrum of sentiments.
    • Green (smiling face) represents a positive response.
    • Orange (unsure face) represents a mixed response.
    • Red (frowning face) represents a negative response.
    • Blue (neutral face) represents a neutral response.
    • Gray (no mouth face) represents none.
  3. Click on the icon assigned to a response, and select a different emoticon to change the assigned sentiment.
  4. Click the gray icon to remove a sentiment from a response and assign it none.
  5. Click the ellipses in the gray filter and search menu at the top of the page and uncheck the Show sentiment box to hide sentiments from the open-ended responses.

Verbatim responses with colored sentiment bands and emoticons 

2. Globally Filtering with Sentiments

Sentiments can be used as filters for open-ended responses.

  1. Click in the search bar and select a sentiment from the pop-up menu.
  2. Click the search button to isolate sentiments of one type across all open-ended responses in the survey.

Selecting a sentiment from the search bar to filter verbatim responses 

3. Question Level Sentiment Filters

  1. Click the Filter tab in the left menu, and scroll to the open-ended question you'd like to evaluate.
  2. Click on the Sentiments drop-down that appears below the question, and select the sentiments you'd like to see.

⚠️ Note: This menu only populates with the sentiments that have been applied to that question—e.g., if there are no Mixed sentiments found in the responses to that question, Mixed will not appear as an option.

filter sentiment.png 

4. Coding and Bulk Actions

Sentiments also appear as codes in the Skipper autocoding menu.

  1. Click the Coding tab in the left menu to view any codebooks you have created, as well as codes associated with each sentiment.
  2. Enter search terms into the search bar, and with the filtered results in view, click the gear icon to the right of the desired sentiment.
  3. Select Remove code from comments on page, or Remove code from all found comments, to remove sentiments in bulk.
  4. Select Assign to multiple comments and then choose Apply to the number of comments displayed on page, Apply to all, or Cancel.

coding sents.png 

5. Custom Virtual Questions

  1. Add a Virtual question, and click Switch to advanced mode above the field with the A1 index.

    adding vq.gif

  2. Enter the logic tag.sentiment Q# = 'positive/negative/mixed/neutral/none' into the first text field. Then add a label to the second field. In this way, you can layer sentiments to create a custom Virtual Question.

    To ensure you are using the correct logic for a sentiment code, go to the Filter tab, enable the filter for the specific question and sentiment, and click Advanced mode. Copy the logic that appears in the Custom filters text box, and paste it into your Virtual Question. Remember to delete the text before switching back to basic mode, to find the logic for a different sentiment code.

    advanced sentiments vq.gif

⚠️ Note: It is important to reference the question number in this logic string, since sentiments are applied to all open-ended responses in your survey.

6. Visualizing Sentiments

To see a quantified view of Sentiment Analysis:

  1. From the Quant Report of the Results page, scroll to an open-ended question.
  2. Click the first drop-down menu, and select Sentiments. A bar graph will appear, displaying the distribution of sentiment across responses for that question.