If we think of education as an industry, it relies a lot on textual content, digitally recorded in the form of PDFs or other documents. As per conventional learning, students are expected to read and understand concepts to be further able to solve real world problems.
But what if reading is not the style for a particular learner? Maybe they would like to travel to a Harry Potter universe where they can talk with a textbook to learn what it has to offer.
With that idea in mind, let’s jot down some hypothesis,
- Learner finds it uninspiring to open a textbook and start reading, they would want to make the process more curious
- Learner procrastinates when they are too many action items to begin with. Opening a book, going to a chapter, finding where they left, reading a few paragraphs until you reach the main content… there’s always too many action items
- Learner has a specific query that makes them most curious on a topic and they would like to start their learning from that individual query. Every learner has a specific learning path.
- Learners are good a questions, and hence conversational learning brings much better outcomes
How can we test our hypothesis listed above?
- We will parse some chapters from regular textbooks and make them available for conversational understanding.
- Learner will land on the webpage, choose a chapter and begin conversing.
- Post every conversation they have an option to stop and practice
- For practice, questions are generated based on the concept nodes they have studied and conversations they have had
What do we need to build to achieve this?
- Upload certain chapters - parse and index them
- A UI where learner can select a chapter and start the conversation
- A system where customer can request practice
- Ability to generate practice based on what they have already learnt
Before starting on the first task, let’s understand the Retrieval Augmented Generation