Projects
Temporally Distant Trained Language Models: A Computational Approach to the Analysis of Language Change
As a final project for LING 28610 Computational Linguistics, I (Federico Tondolo) and fellow students Masaki Makitani and Krisya Louie set out to ask the question: If natural language models were trained on an antiquated corpus could the statistical ties and undercurrents established during training translate to contemporary corpora? That is, if a LM were trained on Shakespeare could it perform acceptably on modern language texts? My contributions to the project consist of a series of n-gram models including bigrams, trigrams, 4-grams, and 5-grams and analysis of their performance in addition to extrapolation of trends and examination of possible underlying motivations behind observed patterns in the final article.
Random Walk - Dev
An R program created by Jai Jeffryes, contributed heavily to the codebase during my time at the Sulzer Lab at Columbia to provide significant speedups to the data pipeline of the program allowing for vastly faster analysis of experimental data in addition to debugging and optimizing code through numerous design iterations working alongside researches to provide input on experimental design to better extrapolate meaningful data during experimental procedure.
AbsolutelyBarmyFlorist
An attempt to include the entirety of the pyABF library and more, this is a program that reads all the ABF files in its directory and creates a series of CSV files for each of the ABF files' channels and sweeps with some minor data analysis mixed in. Used as a demonstration tool to demonstrate the capabilities, both illustrational and manipulatory, that can be automated through Python on .ABF files so that researchers may better understand the breadth of possibilities. Code solidly based on pyABF sample code.AbsolutelyBarmyFlorist is, as you can probably guess, insane. It has a massive amount of features, however, this makes it unimaginably slow. A variation called SimplisticFlorist does away with all the visualizations and console interactions thus keeping the runtime (relatively) in check and can be tailored to execute solely the minimal analysis needed for a given experiment.
SUSTAIN
This was an educational game created with the help of contributors Caroline Apple and Camila Arcentales tailored to teach children between the ages of 11-13 about environmental sustainability through an open-world atmospheric simulator that allowed for players to grow their own gardens while keeping under close watch properties reflecting our own climate crisis such as CO2 levels, temperature and aridity while attempting to ingrain an understanding of how these factors interact to make our world habitable for all living things. A product of Agile and TDD with extensive market research and forethought in its educational approach.
