LocalR is a platform that analyzes news articles for local content using entity extraction and concept expansion, then shows relevant articles for local journalists to explore contexts beyond ones already mentioned in the articles.
Overall Winner, trip to London to finish the product with the Press Association
MoMA Through Time looks through MoMA's exhibition history between 1929 and 1990, with the exhibition dataset kindly provided by the museum for the event.
culpa sentiment performs sentiment analysis on the reviews posted on CULPA to give students a quick summary of the many reviews available for classes at Columbia University. Reviews are collected through the CULPA API, then analyzed for overall sentiment and keywords using the AlchemyAPI.
Winner of the Best Use of CULPA API Award
A simple single-serving website to show you your exact birthday in decimals. It incorporates time zone information with the birthday to determine the most historically accurate time (with Moment Timezone).
This project uses Markov chains to generate almost convincing sentences. New words are selected based on two preceding words, so the grammatical structures are kept well for the most part. The start and end of sentences are preserved as well, so in general, sentences begin and end as they normally would. With vocabs and grammatical style of the authors and topics in the corpus, it provides strange sentences with much of the overall feeling and style still intact. Check out some example paragraphs!
Built using Python
Scholr is a web-application platform to promote and advertise research projects and scholarship/grants, while keeping the project progress accountable. Scholarship and grant providers can provide information about their programs on the platform for users to search, while the recipients are given a chance to provide updates on their projects on the website.
Built using Flask, Python, SQLite, jQuery, and Bootstrap at Lionhack 2015
An interactive visualization analyzing over 220,000 Foursquare check-ins in New York City. This project lets you explore check-in behaviour through different time frames and geographic locations, as well as how the check-in pattern differs with different categories.
Built using D3.js, jQuery, Bootstrap, QGIS, Python for Storytelling with Data
A series of visualizations created for the D3.js-based visualization class at Columbia, covering various interesting topics like ATP tennis ranking history, GDPs of Canadian provinces, and my NYC half-marathon route!
Built using D3.js and QGIS for Storytelling with Data
Visualizr predicts residence of NYC residents based on gender, age, race, employment, income, and commute time with a two-class boosted decision tree model. It also provides a simple solution to visualize data available in the 2010 United States Census, American Community Survey, and NYC Open Data to further explore with other methods.
Third place finish, $1,000 prize
Built using CartoDB, Python, Microsoft Azure ML, and PostgreSQL at the Columbia Data Science Student Challenge Hackathon
OrthoList allows researchers to search orthologous genes between C. elegans and humans. Providing a list of genes from either species generates a personalized list of genes of interest for use in RNAi screens and other purposes (link to publication).
Built using PHP and MySQL for Greenwald Lab