Skip to content

Latest commit

 

History

History
110 lines (85 loc) · 5.17 KB

PROPOSALS.md

File metadata and controls

110 lines (85 loc) · 5.17 KB

cs0320 Term Project 2020

Team Members: Shalin Patel, Kyle Reyes, Derick Toth, Nick Young

Team Strengths and Weaknesses: Shalin

  • S: Algorithms, Debugging, Data Science, Architecture
  • W: Front-end, Style
  • Stack: Python, R, C++, SQL, Mathematica, Java Kyle
  • S: Algorithms, Testing, Project/Team Management
  • W: Front-end, Optimization
  • Stack: Java, Racket, Pyret Derick
  • S: Visual Art, Optimization, Implementing Functionality, Code Style
  • W: Front-end, Project Structure, Testing
  • Stack: Java, Python, Pyret, Racket, SQL, C++, MATLAB Nick
  • S: Front-end, Design, Project Management
  • W: Algorithms, OOP, Optimization, Memory management
  • Stack: Java, Javascript, HTML, CSS, React, React Native, Python, MATLAB, MongoDB, Node.js, Flutter, Dart TEAM
  • S: Algorithms, Architecting, Managing teams and projects
  • W: Front-end, Style
  • OVERALL: Balanced!

Project Ideas:

Idea 1 - Election/Voting Tool

The Problem

  • Tallying votes in a political election securely with active protection mechanisms is nearly impossible, and physical ballots often introduce a layer of error. A secure mechanism to gather and count ballots in an election could help prevent fraud and speed up the democratic process. Our Solution
  • An open-source application integrating a paper ballot reader, with analysis for detecting fraudulent activity. Critical Features
  • A way to register new elections and generate ballots based on candidates and proposals
  • A tool to read in ballot data reliably and cost-ebased on photo of ballot to integrate a paper trail into the election.
  • Keep backups of original ballot and metadata associated with the ballot like time submitted, what precinct.
  • Fast, intuitive visualization of election results. Live Results?
  • Integrated sharing of results to social media sites/news networks.
  • Security around scanning in ballots to prevent bad actors from adding in falsified or modified ballots.
  • Analysis of ballot metadata to provide an automated fraud checking mechanism. Challenges
  • Reliable reading through computer vision based on picture of ballot
  • Data analysis and developing a useful fraud metric
  • Visualization of Data

Rejected - Not enough "good" data and if there were enough data we could just throw an ML or CV algorithm at this

Idea 2 - Automated Travel Planner

The Problem

  • Planning a fun and interesting trip within a certain budget effectively can be difficult in a busy lifestyle. Moreover, many of the parts of booking are tedious and difficult to keep track of. Our Solution
  • A web application that creates a detailed planner/itinerary for a certain area given a few keywords. Critical Features
  • The web application takes in a certain region (i.e. city) and timeframe
  • It also takes in various constraints, such as budget, modes of transportations, hours of operations, specific locations. This will allow for a realistic planner that can be used by the tourist.
  • Weighs specific interests higher than others (e.g. historic sites, food, nature, etc.) in order to provide a more customized experience compared to other existing solutions
  • Outputs a day-by-day itinerary
  • Minimizes travel time and other cost metrics automatically
  • Displays an interactive map for ease of use Challenges
  • Implementing a generic approximate Traveling Salesman solution that is robust under user input
  • Where to find all the data and how to consolidate them
  • Dealing with each restriction given by the user
  • Scraping data to match with user inputs

Approved - algorithm for computing "score" of trip and ranking should be the main component.

Idea 3 - Parseltongue (Efficient Academic Paper Parsing)

The Problem

  • Research is difficult, and having to read through pages of repetitive content is a waste of time and resources. Especially considering that the most “useful” part of any given paper is generally buried amidst a mass of context and other unrelated content, it would be helpful to be able to parse and filter papers for their most salient content. Our Solution
  • Parseltongue is a paper parser that, when given a list of key words to search for, highlights the portions of a grouping of papers that are most relevant to those topics. Critical Features
  • Uploading papers for parsing.
  • View uploaded paper with annotations for the most relevant information.
  • Maybe an automated abstract generator, based on the content needed.
  • Page-rank type algorithm for determining importance Challenges
  • Natural Language Processing?
  • Dealing with non-parsable documents.
  • Creating an intuitive and beautiful UI.

Approved - lots of interesting parsing algorithms. Should avoid just throwing an ML algorithm at a large dataset

Note: You do not need to resubmit your final project ideas.

Mentor TA: Put your mentor TA's name and email here once you're assigned one!

Meetings

On your first meeting with your mentor TA, you should plan dates for at least the following meetings:

Specs, Mockup, and Design Meeting: (Schedule for on or before March 13)

4-Way Checkpoint: (Schedule for on or before April 23)

Adversary Checkpoint: (Schedule for on or before April 29 once you are assigned an adversary TA)

How to Build and Run

A necessary part of any README!