–
Devlog 13 (AI UPDATE)

• Updated sessionShortcuts to use a factory function pattern which allows me to create an object using request as a parameter once, and then access session data easily without having to submit request as a parameter each time. e.g const questionId = sessionShortcuts.getCurrentQuestionId() no request parameter is required.

• Updated AI logic to track what the ai’s suggestion was for each answer and see how that suggestion affected improvement, previously the AI would look at the most recent proposal and credit that the improvement, the new logic tracks the ai’s predicted optimum ratio for each question and tracks improvement individually rather than the last system which would use gross improvement.

To view the website, Click Here

Devlog 12

What is UpRevise? It’s a revision website which adapts to how the user learns best to improve efficiency when studying

Changelog:

• General refactoring
• Added sessionShortcuts which includes helper functions involving session data
• Code now uses sessionShortcuts to get data rather than hardcoding session data keys throughout multiple files
• Fixed bug where the same question would appear multiple times in a row
• Fixed normaliser function so all punctuation is replaced from answers before comparing to correct answers
• Updated short answer and mark feedback styling

To view the website, click here

Devlog 2

Project Description: Uses a custom made cloudflare worker to return details on limited items in Roblox such as RAP (Recent Average Price) and Demand (How many of the item was sold within a time period compared to other months/weeks).

Changelog:

• Refactored most of the code to use reusable functions so expandability is easier and it’s simpler to edit certain functionality that many scripts use
• Details about UGC Limited items are now displayed correctly

Devlog 1

Project Description: A Roblox game which is used for tracking data on Roblox UGC Limiteds (Similar to stocks) and allows users to use an in-game currency to simulate investing in these items and view how much they could be earning if they made real investments.

Changelog:

• Added an explorer for viewing UGC Limited items
• Added basic UI so tokens can later be added for making the fake investments
• Added a basic frame showing analytics of items
• Added system for retrieving the analytic data using a cloudflare worker I created

Devlog 11

Recap: A revision website that uses AI to adapt to the most effective revision methods, personalised for each user

Changelog:

• States for finding the best combination of question formats are now displayed via ratio format and is updated for every question answered
• Fixed marking logic for GapFill questions which would cause false negatives
• Switched the question topics for more general ones to give a better experience for anyone who wants to try it out

VIEW SITE

NOTE: Questions must be answered multiple times before the ratios can adapt

Devlog 10

Project: AI focused revision website to learn how you learn, and use that to improve revision

Changelog:

• Tested two main AI models, used basic versions of Bayesian Bandits and Reinforcement learning
• Reinforcement learning performed the best and was chosen to be used as the model to learn a user’s preference of question formats (e.g GapFill, MultipleChoice, ShortAnswer)
• Add reward file for converting variables to a 0-1 score so it can be easily interpreted by ai models
• Data for models is saved in session data using express as a temporary fix, I plan to move data including past marks on questions and AI weights to a database further down the line.

Media: Screenshot of the state and proposal of the reinforcement learning model after answering a few questions where only ShortAnswer question formats were answered correctly and the other formats were purposely answered wrong

VIEW SITE

(Mini) Devlog 9

Project Recap: A website to help students revise faster and more efficiently using simple but effective AI

Changelog:

• Fixed some of questionsBank.json because the mark points were written incorrectly
• Added colour coding to help users understand where they went wrong/right in their answer

TRY OUT MY SITE

Devlog 8

Project Aim: Build a revision website that uses machine learning to help students learn new topics even faster

Changelog:

• Users now receive additional feedback which specifies which of the “keywords” in their answers were correct, incorrect or ignored (Due to repetition etc)
• Updated marking logic so a user cannot gain multiple marks for the same marking point (E.g NOT getting 2 marks for saying “4 and four” in response to “2 + 2”)
• Reinforced old logic that had been lost to prevent keywords in an answer being detected as filler words (Such as “and” etc)

VIEW SITE

Devlog 7

Project description: An ai-focused revision website to improve the efficiency in learning for students

Changelog

• Edited structure of questions in questionBank.json to allow for synonym answers
• Reconfigured marking logic to suit new structure
• Reconstructed the mark feedback visuals to clearly separate each marking point

CLICK TO TRY SITE

Devlog 6

RECAP: UpRevise is a revision website that will eventually use machine learning to adapt to how different students learn and progress fastest

Changelog

• Added fill in the blank/gap question types
• Fixed marking logic causing answers to be case sensitive
• Debugged some things like input boxes being too small

CLICK TO VIEW SITE

Devlog 5

RECAP: A revision website that aims to adapt to how a user learns to help them learn faster [Machine learning hasn’t been added YET]

Changelog:

• Added the ability to select multiple options at once in multiple choice
• Updated marking logic to be more accurate when marking answers (Like punishing for extra, incorrect answers)
• Fixed some minor ui stuff like centering of feedback text
• Made it easier to add more question formats by giving more control to jsx files managing how different question formats display

TLDR: Added multi-select to multiple choice questions and made the system mark more accurately

CLICK TO TRY SITE

Devlog 4

Recap: UpRevise is a website for revising and will in the future, hopefully, use machine learning to help students learn more efficiently

Changelog

• The enter key can now be used to easily cycle through questions instead of manually clicking submit and next
• Responses are colour coded, varying the visuals to reflect full marks, partial marks, and no marks.
• There is now a system for adding different question formats, currently the website includes: Short Answers and Multiple Choice.
• Refactored some of home.jsx so api calls are handled in client/src/services

HIGHLIGHT: Made an easily expandable system to add multiple question formats like text input answers and multiple choice questions

CLICK TO VIEW THE SITE

Devlog 3

Recap: Revision website to try and increase efficiency in learning so students can carry on spending the same amount of time revising, but with better results.

Changelog

• Refactored loads of my Home.jsx and some other files to become more modular

• Switched to tailwindcss

• Updated UI

• Removed some unnecessary things like dark/light mode to focus on functionality rather than visuals

• Reduced the amount of react states I was using by like 50% whilst maintaining the same functionality

TLDR: I focused mostly on improving my structure so I can greatly increase my future rates of progress

Click To View The Site!

DevLog #2

About

UpRevise - A revision website that will focus on adapting to how the user learns best

Changes

• Shows you some of the answers you could have given to get full marks.

• Site is now publicly accessible thanks to Nest

• Updated logging and API to make debugging easier

Access site: View Site

Constructive criticism IS APPRECIATED

Devlog #1

Introduction

For several years I’ve been exploring programming and, more specifically, machine learning. After a miserable attempt where Claude became the main contributor, I have decided to start from scratch and begin making my ai powered revision website.

What is the aim?

Reducing wasted time - Everyone learns differently, the intent is to make a website that can adapt to whoever is using it.

Progress so far (Humble beginnings)

Since the start of June I’ve:

• Polished a simple question prompting and answering system
• Created an easily expandable filtering system which currently uses subjects and topics
• Added sessions to track user data such as current question to create a server authoritative secure system.

Attachments:

• A demo of answering questions with a filter of Subject: General and Topic: Academic