Researcher & Data Scientist
Hello, my name is Julia! I am a Graduate Researcher with the Stanford NLP
Group
under Prof. Diyi Yang, Research Affiliate at the
Georgia Institute of
Technology,
and Data Scientist III at Bombora.
My research interests include Computational Social Science, Multimodality, Coded Language and Social Norms. I
enjoy working at the intersection of Machine Learning, NLP and Computer Vision to study human phenomena and
develop AI technologies that reflect human values.
On a personal note - I have a background in Physics and the Visual Arts. Where as Physics naturally finds its way into my professional career, I pratice analog photography and oil painting as a creative outlet! Ask me about my work :)
Projects & Publications
NeurIPS 2024
A novel method of augmenting images which provides control over what is changed and how, with limited human supervision. We present Semi-Truths: a dataset of ~1.5 million images for evaluation and training of AI-Generated Image Detectors.
ACL 2024
Dog whistles are coded communication that is often weaponized for racial and socioeconomic discrimination. We present an approach for word-sense disambiguation of dog whistles using LLMs, and a dataset of 16,550 high-confidence coded examples of dog whistles.
EMNLP 2023
A novel dataset of photojournalism through which to investigate the semiotics of images, and how specific visual features and design choices can elicit specific emotions, thoughts and beliefs.
EMNLP 2019
Computing author intent from multimodal data requires modeling a complex relationship between text and image. Neither caption nor image is a mere transcript of the other. We present 3 orthogonal taxonoies that describe image-text relationships, and show that encorporating image and text can improve document intent classification.
CalTech REU 2016
Developed software for detecting ambient environmental noise in the interferometer, as well as determines the affected area. The program has shown to decrease false dismissals of gravitational waves by up to a factor of 10, and continues to be used in the vetting process.
AAS 2015
We present an updated and modernized catalog of Cepheid Variable Stars in the Magellanic Clouds originally constructed by Henrietta Swan Leavitt during her tenure at the Harvard College Observatory (HCO) in the early 1900s.
Patents
Machine Learning Techniques for Internet Protocol Address to Domain Resolution Systems
2023 [ App: US20230252324A1 ]
Spatial-temporal anomaly and event detection using night vision sensors
2023 [ US20240212350A1 ]
Determining intent from multimodal content embedded in a common geometric space
2020 [ US20200134398A1 ]
Embedding multimodal content in a common non-euclidean geometric space
2019 [ US20190325342A1 ]