Text Analysis
An AI-driven analysis that compiles, categorises, and identifies patterns and sentiment in text comments.
Open-ended questions often generate large amounts of unstructured text data that can be difficult to sort, interpret, and visualize clearly and efficiently. Instead, a time-consuming manual process is required to read, interpret, sort, categorize, and analyze the comments in order to draw meaningful conclusions.
Brilliant’s text analysis simplifies this by transforming unstructured text into clear visual graphs that are easy to understand.
Method-Guided Artificial Intelligence (AI)
The analysis is AI-based and uses OpenAI’s latest GPT model combined with custom triggers based on Brilliant’s methodology and model. These triggers (or prompts) guide the AI in the right direction and ensure relevant insights.
Topics
A topic can be compared to a category or theme. The AI categorizes comments based on pre-defined topics that appear most frequently in Brilliant’s database and presents them in a way that provides an instant overview of what the comments are about — with the ability to track changes over time.
Example
A text comment can describe a topic without using specific words that describe the topic. For example, a comment can be about information without the word information being included in the text. Below are two comments that both concern information.
"We are notified in time and able to adapt to changes."
vs
"Changes are implemented without us knowing about it which causes stress"
Both comments talk about the same topic, but the sentiments are different. One is positive, one is negative.
Sentiment
Sentiment analysis identifies how respondents feel about the topics they mention in their comments. By calculating a sentiment score, you gain immediate insight into which areas of the employee experience are working well — and which ones need improvement.
A single comment may contain both positive and negative sentiments, either about the same topic or about different topics. If the content related to a specific topic is both positive and negative, it is categorised as neutral sentiment.
Sentiment Score
Brilliant’s sentiment score shows the balance between positive, neutral, and negative expressions in comments and is reported as a value between -100 and +100. A score above 0 means the majority of comments are positive, while a score below 0 indicates mostly negative feedback. Neutral applies when a topic includes both positive and negative expressions.
Example
In a survey, information is the most discussed topic in the text comments. 54% of the comments are positive, and 23% are negative.
Calculation
- Percentage of positive comments minus percentage of negative comments = Sentiment score
- 54% – 23% = 31 Sentiment score
The text analysis is based on pre-defined topics that are most common in Brilliant’s database. This means that in some cases, open-ended questions may contain responses that do not fully match these predefined topics, resulting in partial or no categorisation.