What is Gemma 3?
It is an LLM (Large Language Model) developed by Google. It’s an open-weights AI trained on trillions of data points to understand and generate both text and images.
If you're curious about trying it out, I recently wrote a guide on how to install it on Linux and other systems: https://byandrev.dev/es/blog/ollama-in-linux/
Benefits of using Gemma 3 in this project
In the context of my educational feedback analysis project, here is how I'm applying it:
Smart Executive Summary
Using Gemma, we can analyze "X" number of student comments and process them into a single, coherent paragraph. For example:
- Input: "The teacher explains well but arrives late," "The slides are outdated," "He is very patient," "Doesn't reply to emails."
- Output: "The instructor stands out for his teaching methodology and patience; however, there are areas for improvement regarding punctuality and updating instructional materials."
Suggested Feedback Generation
This is, for me, the most powerful part of the project. We are not just summarizing what students say; we are using Gemma to draft constructive recommendations for the teacher.
Gemma's Suggestion: "It is recommended that the instructor digitize support resources and establish a more agile communication channel to resolve doubts outside the classroom."
Why is this valuable?
- Scalability: Manually processing a large volume of comments is a complex task. With AI, it’s done in seconds.
- Objectivity: The AI helps eliminate personal bias during the summarization process by focusing on recurring patterns.
- Privacy: By using Gemma (Open Source/Open Weights), there's no need to send data to OpenAI, Anthropic, or any other third-party servers. Everything stays in our hands.
