AI Sentiment and Category Analysis
The Client
Creating Communities Australia (CCA) is a prominent organisation committed to fostering vibrant and sustainable communities across Australia, with a mission to build flourishing people and communities.
With a 30-year legacy of innovation, CCA is on a mission to enhance community wellness, equity, and liveability using AI to turn their wealth of insights into actionable solutions.
Specialists in community engagement, CCA excels at diving deep into community insights through comprehensive surveys. Yet, the manual categorisation of responses into 28 categories was time-consuming and costly. Now, with AI, they're cutting through the complexity to deliver smarter solutions faster, making a real impact where it matters most.
The Goal
The goal is to use CCA models and data to build a survey and analytics tool for the resource sector and other sectors to accommodate people.
CCA urgently needed to modernise and streamline their survey analysis process by replacing their manual categorisation method with an automated system to save time and reduce costs.
The longer-term hope is that this solution can be adapted to residential communities and local governments, prisons, and refugee camps amongst others.
The Journey
Enlighten Designs embarked on a project with CCA to develop a cutting-edge solution. The initiative began with in-depth research into machine learning models for text classification and OpenAI's advanced language models for automated text segmentation and sentiment analysis.
The Solution
The Enlighten Designs team crafted an innovative solution comprising:
- A static web application for survey file uploads.
- Microsoft Azure Functions for a seamless serverless backend experience.
- Utilisation of OpenAI's latest GPT model to intelligently analyse, segment, and assess sentiment within user comments.
- A sophisticated machine learning model, trained on synthetic data generated via OpenAI, to classify responses into 28 categories.
- An interactive, custom Power BI report for dynamic data visualisation and analysis, updated in real-time as new data is processed.
Continuous Collaboration and Refinement
Throughout project development, Enlighten Designs maintained close collaboration with CCA, fine-tuning the solution through rigorous testing and prompt engineering. This led to a highly accurate system for both sentiment analysis and category classification.
The Results
The final system delivered impressive results:
- The language model excelled in segmenting thoughts, summarising sentiments, and contextually interpreting user comments, including the translation of non-English remarks, understanding of Australian slang, and identifying central ideas.
- The machine learning model demonstrated exceptional accuracy in categorising survey responses, thanks to the robust synthetic training data.
- CCA achieved significant cost and time savings, gaining insights previously unattainable through manual analysis.
"Enlighten Designs exceeded our expectations with a powerful, precise solution for analyzing our survey data. The insights we now obtain are invaluable, and we're excited to see how further refinements will enhance our processes."