BlockSurvey’s AI Data Analysis helps you understand your survey data better. Through Thematic Analysis and Sentiment Analysis, we can extract key themes, detect emotions, and generate actionable insights from open-ended responses. Below, we’ll walk you through how the AI processes and analyzes your survey data.
Thematic Analysis Workflow
Our Thematic Analysis process identifies emotional context and key themes from open-ended responses. The goal is to provide a deeper understanding of the emotions behind respondents’ answers and generate themes based on these insights. Here is how it works:
File Ingestion & Preprocessing
The survey data is uploaded in CSV format.
Text responses are extracted from the data column.
The text is then preprocessed to:
Remove non-alphabetical characters.
Normalize text to lowercase.
Tokenize the responses and remove stopwords (e.g., "the", "and").
Discard invalid or empty responses.
Emotion Detection Using Transformer Models
We classify each response's emotion using the DistilBERT-based multilingual sentiment classification model.
The model categorizes responses into:
Positive
Neutral
Negative
Each response gets a probability score for each sentiment, and the dominant emotion is assigned to the response.
Keyword Extraction by Emotion Group
After responses are categorized into emotional groups.
Responses within each emotion group (Happy, Neutral, Sad) are combined for keyword extraction.
Depending on the language of the responses, we use:
Term Frequency-Inverse Document Frequency (TF-IDF for English text)
Rapid Automatic Keyword Extraction (RAKE for non-English text)
The most frequent and meaningful keywords are extracted for each emotion group.
Theme Generation Using Large Language Model (LLM)
We use GPT-4-o to generate themes from the extracted keywords.
The model receives:
Survey details (goal, target audience)
Keywords grouped by emotion
The output includes:
Theme Titles: High-level descriptions of the emerging themes.
Detailed Descriptions: Insights that explain the emotional context and relevance to the survey’s goals, along with actionable interpretations.
Sentiment Analysis Workflow
Our Sentiment Analysis workflow helps classify the emotional tone of your survey responses. This provides valuable insights into how respondents feel about specific topics and helps you understand the overall sentiment. Here is the step-by-step breakdown:
Preprocessing Responses
Responses are processed by:
Converting text to lowercase.
Removing stopwords.
Filtering out invalid or short responses (less than 3 words, generic phrases).
Sentiment Classification
We use the DistilBERT-based multilingual sentiment classification model.
The model assigns each response a sentiment label (Positive, Neutral, or Negative) and computes a compound score (ranging from -1 to 1).
Categorization by Sentiment
Responses are grouped into three categories:
Positive
Neutral
Negative
For each category:
We calculate the average sentiment score.
We extract the top keywords using tokenization and frequency distribution.
A sample of up to 10 representative responses is selected.
Theme Generation (Per Sentiment)
For each sentiment category (Positive, Neutral, Negative), we generate meaningful themes using GPT-4-o.
The input to the LLM includes Sentiment category, extracted keywords, and survey details.
The output includes:
Theme Titles: A concise description of the theme.
Insights: Emotional tone, relevance to the survey’s goals, and actionable recommendations.
How to Use AI Data Analysis in BlockSurvey
You can analyze open-ended survey responses using AI Data Analysis in two ways:
From the Dashboard
Navigate to the AI Data Analysis section.
Upload your survey data in CSV format.
Choose between Thematic Analysis, Sentiment Analysis, or both.
From the Analytics Screen
If your survey contains open-ended responses, navigate to the Analytics -> Analyze Responses screen of the survey.
Click the Generate AI Analysis button for the open-ended question.
Once the analysis is complete, you’ll receive key themes, sentiment classifications, and actionable insights to support your decision-making.
BlockSurvey’s AI Data Analysis provides a powerful way to analyze open-ended responses, identify key themes, and understand the emotional context behind your data. Whether you're conducting market research, gathering feedback, or analyzing customer satisfaction, BlockSurvey’s AI Data Analysis can help you gain valuable insights and make data-driven decisions.