Fall Demo Day 2020
In a whirlwind of a year that could’ve gotten to the best of us, Demo Day was a reminder that the Spark! community is just as resilient as ever. Every one of the participants demonstrated perseverance in the face of adversity. How else do you explain how they spent an entire semester developing fully functioning prototypes during a global pandemic and the biggest social justice movement since Civil Rights?
Our Spark! Fellows and X-Lab students competed for prizes, prestige, and a coveted Audience Choice award. Let’s face it though- they’re all winners since their projects are guaranteed to rock the marketplace.
Spark! mentors, our partners, and the entire Terrier Town community deserve an overwhelming thanks – there would be no us without you!
Innovation Fellow Project Gallery
LORELAD, Audience Choice Winner
Jonathan Hall, Thuy-An Nguyen, Yuhao He, Tsubasa Morita, Laurina Saint Fleur
The number of languages is quickly decreasing. By 2150, there will be in estimates 600 languages still actively spoken from an estimated 7000 active languages today. The languages at risk of dying are low resource languages. LORELAD aims to be an intelligent storage device for low resource language translations – a place to aggregate existing data and accumulate more as they’re available.
Contact: jshall@bu.edu
View LORELAD at Demo Day HERE
B Scanner, Judges’ Choice Winner
Shateva Long, Lucy Baik, Juan Almazar, Ikechukwu Okoye
The internet is filled with tons of biased sources and misinformation. B SCANNER’s goal is to create an online tool that will scan online sources for bias and educate users on why/how the source was deemed bias, so that they can make an educated decision on whether they want to use a source or not.
Contact: shateval@bu.edu
View B Scanner at Demo Day HERE
FitStart
Mark Tony, Mahnoor Butt, Will Paarz, Israel Ramirez, Minglan Zheng
97% of college students interviewed say their college lifestyles can be healthier. FitStart is a fun and affordable way for college students to engage in the healthy lifestyle they deserve during their four years on Comm Ave. An excellent alternative to the hard to follow, unorganized, and unrealistic health information available online, FitStart offers their users a way to gamify healthy living amongst their peers. Users earn XP points and level up by accomplishing their personal goals, eating a healthier meal at the dining hall, and by completing their favorite work out. Students can add their friends and battle each other out to be the top of their social circle’s FitStart leaderboard, bragging rights included.
Contact: marktony@bu.edu
View FitStart at Demo Day HERE
Sources.fyi
Jason Cooper, Spencer Vilicic, Thachathum Amornkasemwong
Since Instagram blocks links in captions, informational infographic creators often cite sources in their content by providing a photo of multiple links as the last slide of a carousel post. However, since it is still a picture, followers of these creators cannot currently click these links after engaging with an “Instagraphic” to research more. Sources.fyi lets creators attach a list of sources to each post on their Instagram at their own personalized page: sources.fyi/username
Contact: jcoop88@bu.edu
View Sources.fyi at Demo Day HERE
Go Off!
Glo Robinson, Stephanie Lieu, Eric Chao
Go Off! is the first and only emerging media platform that hosts small live-chat discussions that are moderated to encourage smart debate and discussions around current events and news that is happening now. At Go Off, we have developed a software to collect the qualitative analytics of these text based conversations and provide in-depth insights about Gen-Z trends to small to mid-size businesses.
Contact: grobins@bu.edu
View Go Off! at Demo Day HERE
Spot
Savannah Cardenas, Priya Kumari, Melissa Lopez, Nick Ni, Chrissy Casavant
Spot aims to provide crowd and density monitoring tools across the BU campus, allowing students to gauge crowd size and building occupancy in certain areas so they can more effectively evaluate their risks and make more informed decisions on campus this fall and beyond as we adapt to COVID. These tools may be expanded to other areas besides the BU campus.
Contact: scarden@bu.edu, pkumari@bu.edu, mlopez99@bu.edu, ni@bu.edu
View Spot at Demo Day HERE
MP Health
Christian Soderberg, Kaela Gobencion, Sohini Mukherjee, Asif Rahman, Kevin Tu
MP Health is a scheduling platform that helps Medical Clinics jump towards the next step in the future. By automating one the main jobs of a modern day receptionist, MP Health will give the opportunity for any clinic to become technologically advanced and efficient.
Contact: cfsode@bu.edu, kaelaag@bu.edu
View MPHealth at Demo Day HERE
Practicum Project Gallery
BU CFA | A Random Act
Gonzalo Rosales, Ben Sui, Zhou Fang, Seyun Om
The goal of this project is to combine computer vision and graphic generative technologies in order to create an application that can be used to follow and track actors on the play stage. This process will generate graphics that will help in telling the story of the play. The students have been working with technologies that allow them to interpret the data that is being created from the camera, recognize human forms once it comes in the frame shot, and then create visual effects coming from the human skeleton inside the shot in order to simulate complimentary effects. The expected end result is a desktop application that has the functionality described above, along with the user having the ability to modify the generation of effects as they see fit.
View BU CFA | A Random Act at Demo Day HERE
City of Haverhill Constituent Services
Dan Katz, Jin Young Bang, Tej Mulchandani
The City of Haverhill uses ArcGIS/MapGeo to view and analyze Haverhill’s map assets. This includes dozens of city assets such as roads, bridges, sidewalks, fire hydrants, drainage systems, dams,etc. Information includes the physical location, asset dimensions, asset condition, geographical information such as voting districts, watershed boundaries, census blocks, etc. The students have been developing a working pipeline to take the data from the QAlert system as they come in and dynamically update a GIS layer that can be added to the City’s ArcGIS mapping platform. This will allow for a visualization of the data and allow for enhanced analysis in the process.
View City of Haverhill Constituent Services at Demo Day HERE
Framingham Public Schools
Jayden Tayag, Jiwon (Tanner) Park, Yernur Alimkhanov, Ziyu Shen, Zheng Hui
Each year, the Framingham School district receives thousands of applications for new students to be assigned to the various schools in their districts. The assignment process is a complex one that is based on an array of factors. The district looks to assign students to schools that match their preferences, while maintaining thresholds that they have for each individual school in their district. This semester, the students in this group have been working on the implementation of the assignment algorithm that takes an excel sheet that contains the relevant information regarding every student, assigns values to those variables, and outputs an excel file that has each student and their school assignment. A process that presently takes months, can now be completed in a matter of minutes.
View Framingham Public Schools at Demo Day HERE
Mindful Applicant
The Mindful Applicant is a project that represents a shift in the way the college application process is structured. Through the application, users would be incentivized to become aware and develop further their social-emotional skills, something which is an important aspect of development, yet there is very little incentive to be aware of and further develop. The students in the team have been working on developing the web application that will put the gears in motion. The students have developed the initial structure that has the initial functionality of the application. Users can login, take a quiz, and get their results on the personality traits they possess.
View Mindful Applicant at Demo Day HERE
Savor, Audience Choice Award
Yunhan (Hannah) Huang, Nathalie Ye, Alex Doval, Andrew Eramo, Kari Everson
Savor is an application that is meant to guide users in their meal plan development by taking into account elements like family size, budget and dietary preferences. The application will allow users to pick the meals that they intend to make over a certain amount of time, and it will show users the exact portions and ingredients that will be required, in addition to cost. This will allow users to make informed decisions about their meals. The students have been working to put together the initial web application by leveraging modern technologies such as React and Firebase.
View Savor at Demo Day HERE
Suffolk County DA’s Office
Victor Figueroa, Eesha Gholap, Wail Attauabi, Neilkaran Rawal, Dingjie Chen, Sloane Shuchman, Kari Everson
The Suffolk County DA’s office has many processes in which data has to be shared in between different entities. Presently, many of these processes are manual, where documents are created by hand and need to be physically shared between the involved entities. The goal of this project is to create an iOS app that can digitze this process. The students have been working towards developing an iOS app, creating an infrastructure that allows for login and allows the user to capture images of their documents that can be sent to an endpoint for OCR processing. Essentially, the process extracts the data it sees in images and formats it to a machine readable format.
View Suffolk County DA’s Office at Demo Day HERE
FOIA Project – Defense Data / Outsourcing Security
Zhaoguo Zhu, Piotr Nojszewski
This project has four main goals: to provide a comprehensive history of US security outsourcing; explain the reasons for outsourcing, whether due to cost efficiency or managing financial or political risk; identify the shifting balance between private markets and public goods in security over time; and to draw lessons for policy makers for future oversight and regulation. These goals can be disaggregated into several research questions: 1) When did security outsourcing begin? 2) What are the patterns over time of security outsourcing, how do they differ by US agencies, and what topics and policies got outsourced when? 3) Are citizen or constitutional transparency demands, capacity burdens, and financial pressures related to subsequent decisions to contract functions to firms?
View FOIA Project at Demo Day HERE
Harvard Herbarium – Handwriting Transcription Project, Judges’ Choice Award
Shubhangi Jain, Kaihao Jin
In concert with the Harvard University Herbarium, we will continue improving an intelligent word recognition module used to digitize herbarium specimens. Herbarium specimens are pressed plant samples stored on paper. They are an invaluable source of information for phenologists. Digitized specimens will facilitate easier dissemination of information and allow more people access to data. Each specimen contains a label that contains the name of the curator, their institution, the species and genus, and the date the specimen was collected. A sizable number of these labels are handwritten and date back to the early 1900s. We have built a bidirectional LSTM-RNN to transcribe specimen labels, with an accuracy of ~85% across 5000 samples. This project will involve identifying potentially novel approaches to improve the accuracy of the model, or potentially developing a new model from scratch.
View Harvard Herbarium at Demo Day HERE
Liberator Newspaper: Segmentation and Analysis
Ian Saucy, Alex Thomas
Boston Public Library has digitized a very significant collection of Abolitionist newspapers from the mid 1800s called the Liberator, which was the leading anti-slavery paper of its time. This collection has images and issue-level OCR text (around 4,000 pages), but almost no subject analysis. In order to provide more granular access to this material for researchers, we want to segment the issues into article-level objects, and extract information about titles, names, topics, and locations from within each article’s text. This project will involve analyzing the digitized pages, identifying article segments as coordinates/polygons in the image file, performing optical character recognition on each article segment, formatting the word-coordinate data using the ALTO XML standard and extracting article titles and prominent topical features from the text. (The images can be downloaded from digitalcommonwealth.org using available APIs.) An additional challenge will be matching the extracted terms against a controlled vocabulary. For topical subjects, this would be Library of Congress subject headings; for names, this would be Library of Congress Name Authority File; while for locations our preferred source is the Getty Thesaurus of Geographic Names, though we also use Geonames as well.
View Liberator Newspaper: Segmentation and Analysis HERE