Celine Lee

Hi! I am a PhD student at Cornell University working with Sasha Rush on questions in natural language and programming language semantics and reasoning. Previously, I studied at the University of Pennsylvania, where I worked with Dan Roth in the Cognitive Computation group and obtained my Bachelor's degree in Electrical Engineering and Computer Science as well as my Master's degree in Embedded Systems. My CV.


Recent News

October 2021 - I was interviewed by Dr. Justin Gottschlich on the Machine Programming & Technology channel. Video.

August 2021 - I presented 'A Survey on Semantic Parsing for Code Generation', co-authored with Justin Gottschlich and Dan Roth, at the 2021 KDD Workshop on Programming Language Processing.

April 2021 - I have committed to Cornell University, where I will pursue my PhD in Computer Science starting Fall 2021!

May 2020 - I am honored to be the recipient the 2020 Walter Korn Award, granted by the Moore School at the University of Pennsylvania. "It is awarded annually to an outstanding senior in the Moore School who will be continuing on at the Moore School for a graduate degree. Selection of the recipient is made by the Moore School chairs and faculty, in conjunction with the Associate Dean for Undergraduate Education."

May 2020 - Our senior design team won the Fred Ketterer Award for our work in TITO - Tune In, Tune Out. "The Fred Ketterer Memorial Award is given annually to the senior design team in the Department of Electrical and Systems Engineering who demonstrates outstanding creativity in an engineering design incorporating hardware." We have been interviewed by the department and by the university, check out the articles here: ESE, Penn Today

Publications and Patents

A Survey on Semantic Parsing for Machine Programming

Celine Lee, Justin Gottschlich, Dan Roth

Presented at the 2021 KDD Workshop on Programming Language Processing. arXiv PLP Workshop

Pending patents

USPTO PreGrant Publication Database.

METHODS AND APPARATUS TO IMPROVE DATA QUALITY FOR ARTIFICIAL INTELLIGENCE (USPTO App. No. 17/540050)

METHODS AND APPARATUS TO TRAIN MODELS FOR PROGRAM SYNTHESIS (USPTO App. No. 17/551170)

METHODS AND APPARATUS TO DETERMINE REFINED CONTEXT FOR SOFTWARE BUG DETECTION AND CORRECTION (USPTO App. No. 17/554918)

METHODS, APPARATUS, AND ARTICLES OF MANUFACTURE TO GENERATE COMMAND LISTS TO BE OFFLOADED TO ACCELERATOR CIRCUITRY (USPTO App. No. 17/559556)


Software

Semantic Role Labeling (English)

Semantic Role Labeling (Spanish)

Blog

Smelly Colors

The Vowel Trapezoid

My first study of "grounding" natural language in code on StackOverflow


Experience

Research Scientist & Software Engineer

Merly.ai
January 2022 - September 2022

Research Scientist Intern

Intel Labs, Machine Programming Research

PI: Dr. Justin Gottschlich

March 2021 - December 2021

Research Assistant

Cognitive Computation Group, University of Pennsylvania

PI: Dr. Dan Roth, Eduardo D. Glandt Distinguished Professor

Development of semantic role labeling models across different predicates and languages.

January 2020 - March 2021

Product Development Intern

VMware
May 2019 - August 2019

Research Assistant

Autonomous Vehicles Group
January 2019 - October 2019

Electrical Engineering Intern

UTC Aerospace Systems - ISR & Space Systems
May 2018 - August 2018

Research Assistant

Singh Nanotechnology Center
January 2017 - January 2018

Research Assistant

The Santiago Research Group
May 2017 - August 2017

Projects

Program Transformation via Natural Language Instructions

An exploration of using pre-trained language models to perform code editing from natural language instructions.

I examined two methods of performing code editing from natural language instructions: a Codex-based system, and a synchronous grammar over a custom-written Python transformation DSL.

PDF attached here

Fall 2021

Toward Code Generation: A Survey and Lessons from Semantic Parsing

Advisors: Dr. Dan Roth, Dr. Justin Gottschlich

Presented at the 2021 KDD Workshop on Programming Language Processing. arXiv

2020 / 2021

FPGA Accelerated Conv2d

Optimized 2d convolution operation on FPGA.

I implemented and optimized a 2D convolution operation on FPGA, with the end goal of accelerating inference on a pre-trained VGG-16 model.

PDF attached here

November 2020 - December 2020

Tune In, Tune Out (TITO)

A snap-on selective noise isolation system.
Winner of the 2020 UPenn Ketterer Prize

Our team developed a mixed-signal approach to selective noise isolation. By leveraging microphone directionality and designing tunable fourth-order bandpass filters, we created TITO. TITO isolates and mixes sounds selected by the user: somebody speaking to them, the cars on the road, safety alarms, a baby crying, etc.

See our video here and our devpost here

August 2019 - May 2020

DRAM Memory Array

4x4 3T DRAM Array

I designed, simulated, and created a partial layout for a full 3T DRAM memory array, with relevant peripheral circuitry. All designs were created and sized from the transistor level.

PDF attached here

April 2020

Musical Style Transfer

An exploration into deep learning approaches to changing music genre

Our team explored various deep learning methods to apply style transfer to music. A CNN with Gram matrix, an RNN, a BiLSTM, and a Cycle-GAN inspired generator are compared.

Video attached here and PDF attached here.

February 2020 - May 2020

PennOS

In the capstone project for our operating systems course, our team built PennOS, a complete operating system with a fully operational scheduler, shell, and file system.

Report not public, in consideration of students still taking the class.

October 2019 - December 2019

Speaker Profiling Using Machine Learning Methods

An exploration into machine learning approaches to identifying speaker metadata.

Our team sought to determine the best method to profile spcific characteristics of a speaker, given an audio sample. We pre-processed the data to create an even distribution across labels, normalized, and computed fourier transformations as well as F0 fundamental frequency. We then tested different models to determine the most effective classifier.

PDF attached here

February 2020 - May 2020

CPU

A superscalar-pipelined CPU performing "LC4" instructions.

In the final project of our computer architecture class, my partner and I created a full superscalar-pipelined CPU with registers, ALU, and ability to process "LC4" instructions, a UPenn variation of LC-3 ISA. The entire project was written in Verilog.

February 2020 - May 2020

Metal Detector

I designed, simulated, and built a metal detector comprised of a frequency oscillator and frequency mixer, from analog circuitry components. A featured hand-wrapped coil acts as the sensing inductor that produces the change in frequency of the circuitry. Capstone project for analog circuitry class.

PDF attached here

April 2020

Audio Docking Station

Separate music based on frequency.

In this project, my partner and I designed, simulated, and built an audio docking station. The project is aided by an AC/DC power converter that we also designed. An input aux port takes audio and passes it through a treble and a bass filter, which separate the audio signals and feed them to the output. Tunable-gain amplifiers allow the user to determine how much bass and how much treble to listen to. Capstone project for circuit theory class.

PDF attached here

November 2017 - December 2017

Extras

I grew up playing volleyball, and feel so lucky to have had talented teammates and coaches that have shaped me into who I am today. Being at university allows me to continue to play on a team today. On a free weekend, I am wandering around as much as possible and if I'm lucky, I'm trying something new. Recently (for 2022), this has been squash.

I also dabble in miscellaneous arts and crafts.