QInsights currently offers sentiment analysis for semi-structured data in Excel format, making it ideal for open-ended survey responses, customer feedback, or social media comments. If you want to conduct a sentiment analysis for text or PDF documents, you can still do this by using a prompt in Conversational Analysis. See below for an example prompt you can use.
Traditional sentiment analysis often categorizes feedback as positive, neutral, or negative, but with the power of generative AI, it becomes much more dynamic. QInsights allows you to define custom dimensions that align with your data, such as "satisfied" vs. "dissatisfied" or "rational" vs. "emotional". Additionally, you can refine your analysis by focusing on specific topics and engaging interactively through follow-up questions, also making sentiment analysis a collaborative process between human researcher and AI-assistant.
How It Works
Select Your Data & the Sentiment Analysis Option
- Select an Excel file.
- Next, choose the column you want to analyse. Sentiment analysis evaluates data case by case (= row by row) and click Select.
- When in analysis menu, select Sentiment Analysis.
Define Sentiment Categories
- Enter the sentiment categories you want to explore.
You can use standard sentiment categories, such as positive, neutral, and negative, or customize dimensions based on your needs. Examples are:
- Satisfied, neutral, dissatisfied
- Rational, emotional
- Planned, impulsive
- Supportive, critical, neutral
It is also possible to narrow the scope to a specific topic. This is especially useful for datasets like social media comments or product feedback. Examples are:
- Social media: "Classify responses that directly refer to the post’s content as like, neutral or dislike.”
- Customer feedback: "Disappointment or satisfaction with regard to usability"
Review Results
QInsights produces a pie chart showing the distribution of your defined dimensions.
- Hover over the chart to see the percentage distribution.
A summary of the analysis is displayed below the pie chart, written by Q, your AI assistant.
- Click View AI Summary to edit it.
Below the summary, you can inspect how each response was classified. Q also provides an explanation for each classification (see screenshot below). The table with the classification results can be exported as Excel file.
Ask follow-up questions
In addition to the summary, you can ask follow-up questions to delve deeper.
- Like in Topic or Grid Analysis, click on the Q button to start a follow-up chat.
You can for example ask Q for a detailed list of dimensions and their corresponding distributions: Here are two example prompts:
- "List all dimensions with their percentage distribution."
- “Check all rational reasons for similarity, group all similar reasons under a category and give each category a name and a description. Output all categories with description and provide some example quotes.”
Best Practices for Effective Sentiment Analysis
Select one question in the Excel file at a time. E.g., if your dataset includes separate questions for "benefits" and "challenges," analyse these columns individually to avoid conflicting results.
Keep prompts simple. Instead of overly complex or multi-layered queries, focus on straightforward questions:
- rational vs. emotional regarding product design
- positive, neutral, or negative regarding pricing
Results like the following might be interesting for smaller data sets, but become quickly difficult to read. If the pie chart looks similar to the screenshot below, simplify your prompt.
Tailor Dimensions to Your Needs: Examples
- Happy, frustrated (e.g. for customer feedback)
- Supportive, critical, neutral (e.g. for evaluating user responses to a policy.)
- Disappointed, neutral, excited
- If you want to work with an established framework of dimensions, add those including their descriptions to the project description. If you use these dimensions in the analysis, Q will take the description of what each dimension means into account.
Prompt for Sentiment Analysis for Word / PDF documents
You can use the following prompt in Conversational Analysis if you want to conduct a sentiment analysis for Word or PDF files, e.g. for an interview, a focus group or a document:
For each document, classify sentiments at a reasonably granular level (e.g., by paragraph, speaker turn, or thematic section) into positive, negative, or neutral categories (or define other dimensions that better suit your analysis purpose).
Document-Level Summary
- Describe the overall balance of positive, negative, and neutral sentiments (e.g., “mostly positive with some negative remarks”).
- Cite short quotes or examples that illustrate each sentiment.
Cross-Document Analysis
- Summarize common topic areas in sentiment across all documents.
- Identify any significant outliers, contradictions, or unexpected expressions of sentiment.
Relative Weighting (instead of counts)
- Indicate for each document whether positive, negative, or neutral sentiments are dominant, secondary, or minor.
- For the full set, summarize the relative distribution of sentiments (e.g., “positive sentiments dominate, but negative remarks are frequent in relation to [topic or theme X]”).
Output Format
Per-Document Analysis
- Document title or identifier
- Sentiment balance (dominant / secondary / minor)
- Representative quotes or excerpts
- Brief summary of key sentiment findings
Cross-Document Summary
- Overall sentiment balance across all documents
- Overarching themes or patterns
- Conflicting, unusual, or surprising findings
Refine the prompt to fit your specific needs!