# Noteaid Proposal
### Georgia Mandell, Nick Irwin, Giorgie McCombe, Gwen Wattenmaker, Jacob Donoghue, Phoebe Ford | CS98 22F


## How Might We Question / Problem Statement
HMW give students in recorded lecture-based classes better study resources without creating extra work for teachers?

## Stakeholders
* Users: college students, maybe high school teachers, society
* Partners: schools / universities, third party education games like Kahoot (?), Dartmouth Accessibility Services
* *Society*
* Founders: us

## Proposed Solution
*Briefly describe your proposed solution. How does it work and what does it do?*
A web or mobile app aimed at helping students master class material. It will include the following main features:
* allow students to upload lecture recordings (& lecture notes)
* the app will do the following:
* identify key concepts from lecture transcript
* compare these to the student's notes and identifies missed concepts
* use NLP question answering to create intelligent questions + answers about lecture concepts
* gives student feedback

## Prior Work
*Briefly recap any particularly relevant sources from your state of the art research.*

* A company called Knowt which already implements generating Multiple Choice, Fill in the Blank, and Explanation/Open-Ended questions from notes
* Audio Note
* records lectures and allows you to take notes at the same time, syncs recorded lecture with student's notes
* demonstrates that auto-generated questions are helpful, but only works for students that already have good notes (which our target users have trouble making)
* Otter Voice Notes
* record lecture and displays transcript as you take notes
* example of live transcription for aiding note taking
* Stuvise
* web app that generates flashcards from your notes, includes metrics for tracking progress
* example of auto generating study materials from notes
* Whisper -- Open AI's automatic speech transcription software
* Will be helpful in transcribing lecture recordings
* Why taking notes is important
* more notes and instruction on note taking leads to better comprehension
* "Question Answering" is a task of automatically answering a correct answer to the questions asked by humans in natural language using either a pre-structured database or a collection of natural language docs
* This current technique is flawed because it presents only the requested information instead of searching and indexing full documents like a search engine
* In education, closed domain QA systems for handling documents related to education subjects promise more precise answers using NLP techniques

## Impact
*Explain how a solution to this problem meets a real-world need. Does it have a broad potential impact?*

* For content-area classes that are lecture-based, students rely heavily on listening and notetaking--teachers told researchers that their class lectures were the major source for developing secondary science and social studies tests. Writing down notes during fast-paced lectures can be a daunting task, especially for students with learning disabilities such as attention deficit disorders or physical disabilities that affect handwriting legibility.
* Our product would provide students with study material based on the lecture content *without* adding any extra work for the teacher. It would level the playing field for students who don't have the organizational skills to decide what the most important information is and take meaningful notes, and facilitate better retention of lecture information for all students.

## Success Metrics
*What are your ultimate success metrics? This could be changing one person's life for the better, or getting 1k downloads, or some performance threshold.*
* A few students finding it helpful without it being extra work for teachers

## Minimum Viable Product (MVP)
*Briefly describe your MVP features. Don't get bogged down in details, this should be a very terse list. What is the critical thing your product needs to do to be special? Build this from your User Personas.*
* Our MVP would be a website that allows a student to upload a lecture recording and generates questions/flashcards corresponding to important terms and concepts from the lecture.

## Validation
*Explain how you plan to test/validate your product. Propose a term 1 and a term 2 validation idea. This can be different things depending on your project: user testing, or usage analytics, or performance characteristics. It can be done via surveys or in product analytics, or other performance measurements.*
* Term 1
* development focus: the algorithm that turns the transcipt into questions and answers
* validation:
* Test auto-generated questions on students taking a given class (w/recordings)
* unit testing for each part of the process
* Term 2
* development focus:
* fine-tuning and expanding upon core functionality
* creating a full-stack web application to make product accessible to users
* validation:
* UI/UX testing (how easy is the product to use?)
* continued functionality testing with students
* Have students test our application and see if they like it and if it helps them with their material and learning and if they would use it more

## Implementation Challenges
*What development challenges have you thought about so far? Identify any risks and briefly discuss possible workarounds / alternatives.*
* Harnessing various open-source libraries (Whisper, NLP library)
* Alternative: build model from the ground up (not ideal)
* Generating intelligent questions and answers from class transcripts
* Evaluating quality of generated questions and correctness of given answer