Christos Theodoropoulos

Ph.D. Candidate in CS @ KU Leuven

Hello, world!


I am a fourth-year Ph.D. Candidate in Computer Science at KU Leuven, under the esteemed guidance of Prof. Marie-Francine Moens. My intellectual pursuits orbit at the captivating confluence of Natural Language Processing, Information Extraction, Deep Learning, and the fascinating world of Continual Learning.

In my current academic journey, I spearhead the creation of an end‑to‑end open‑source knowledge extraction system from the ground up. This endeavor, meticulously architected around the expert-in-the-loop paradigm, seamlessly aligns with the tenets of lifelong learning. Recognizing the challenge of navigating vast volumes of research text daily, my unwavering focus has centered on the realm of biomedical text. Within this domain, I've embarked on a mission with far-reaching implications by tackling the intricate complexities of two formidable yet unresolved diseases: Alzheimer's disease, encompassing a staggering trove of over 500,000 PubMed papers, and Rett syndrome, spanning a comprehensive collection of more than 4,000 PubMed papers.

However, my aspirations don't stop at disease comprehension through seamless knowledge extraction. My research gaze extends to the frontier of personalized medicine, where I harness the potential of person-centric knowledge graphs to decode the enigmatic progression of ALS disease. My ambition lies in bridging the gap between cutting-edge medical insights and the noble endeavor of personalized patient care, creating a synergy that holds the promise of transforming lives.

A notable accomplishment that adorns my research journey is the creation of a contrastive learning framework, ingeniously integrating both graph and text modalities to instill relation-oriented structure into Language Models like BERT and RoBERTa. This novel approach has empowered me to transcend the constraints of traditional text encoders, ushering in a new era of possibilities. By translating the acquired embedding spaces into tangible solutions, I've ventured into the realms of relation extraction and named entity recognition, solving these intricate tasks using deceptively simple yet effective classifiers like KNN.

Beyond academia, my expedition led me to the vanguard of innovation as a research scientist intern at IBM Research. Here, I had the privilege of immersing myself in projects that not only stoked the flames of my curiosity but also enriched my repertoire of skills. A significant chapter in this journey involved insights from Electronic Health Records (EHR) to construct person-centric knowledge graphs. This endeavor saw me wield the power of the HSPO ontology and harness the prowess of Graph Neural Networks (GNNs) to tackle the formidable task of predicting intensive care unit readmissions. Embracing the spirit of versatility, my explorations expanded to encompass the fascinating domain of embedding learning for heterogeneous graphs. These explorations are on the brink of yielding results in the shape of upcoming research publications. The team's influence also extends into the realm of open-source collaboration. I've been instrumental in fortifying IBM's commitment to knowledge dissemination, playing a pivotal role in bolstering their open-source initiative, exemplified by the repository. In a testament to my innovation, and inventiveness, I proudly stand as the primary inventor in a patent filing.

In my earlier ventures, I've delved into a diverse array of research domains. One path led me to the intricacies of denoising resting-state fMRI scans through the application of deep learning techniques. Another trail took me into the realm of gaze localization and eye-tracking during driving, where I sought to unravel the mysteries of human attention and behavior on the road. Additionally, I immersed myself in the captivating world of human emotion recognition from conversational video data, uncovering the nuanced interplay of emotions in dynamic interactions.

As for my roots, I hail from Tripolis, the capital of Arcadia, steeped in history and surrounded by breathtaking beauty. My formative years were then shaped in Athens, where seven years of studying, working, and immersing myself in the city's energy left an indelible mark. Today, I'm nestled in Brussels, a cosmopolitan hub pulsating at the heart of Europe, characterized by its multicultural essence and lively spirit.

For deeper insights into my journey and potential collaborations, I invite you to explore my CV or reach out to me directly. And if the allure of Greece beckons, I stand ready to offer recommendations. Here's to embracing the journey of knowledge, discovery, and shared experiences!



Education

2020-now
KU Leuven

Ph.D. in Computer Science

2019-2020
KU Leuven

Advanced M.Sc. in Artificial Intelligence

2012-2018
National Technical University of Athens

M.Eng. in Electrical and Computer Engineering

Professional Experience

2022
IBM Research (ACCELERATED DISCOVERY TEAM)

Research Scientist Intern

2019
I-SENSE Group, ICCS

Researcher, Deep Learning Engineer

2018
Sentio Solutions Inc.

Data Scientist




For more details, please see my full CV (PDF).

News



06/2024
Our follow-up paper "Evaluating the Predictive Features of Person-Centric Knowledge Graph Embeddings: Unfolding Ablation Studies" with Natasha Mulligan and dr. Joao Bettencourt-Silva that unfolds the ablation study based on the framework presented in "Representation Learning for Person or Entity-Centric Knowledge Graphs: An Application in Healthcare" got accepted into MIE '24.

06/2024
Our paper "GADePo: Graph-Assisted Declarative Pooling Transformers for Document-Level Relation Extraction" with with Andrei Cătălin Coman, prof. Marie-Francine Moens and dr. James Henderson got accepted into KnowledgeNLP-ACL '24.

10/2023
Our paper "Representation Learning for Person or Entity-Centric Knowledge Graphs: An Application in Healthcare" with Natasha Mulligan, prof. Thaddeus Stappenbeck, dr. Joao Bettencourt-Silva got accepted into K-CAP '23.


03/2023
Our paper "An Information Extraction Study: Take In Mind the Tokenization!" with prof. Marie-Francine Moens got accepted into EUSFLAT '23.

11/2022
Gave a talk on Person-Centric Knowledge Graph Extraction from Electronic Health Records to solve downstream tasks in "Show & Tell" session of IBM Research.

10/2022
Submitted my first patent as the primary inventor, working in IBM research.

10/2022
Heavily contributed to IBM's open-source project (HSPO ontology).

01/2022
Accepted an offer to join IBM Research (Human-centric AI team) for a research scientist internship.



09/2021
Our paper "Imposing Relation Structure in Language-Model Embeddings Using Contrastive Learning" with dr. James Henderson, Andrei Cătălin Coman and prof. Marie-Francine Moens got accepted into CoNLL '21.


07/2021
Became member of ContinualAI organization.

04/2021
Our work with dr. Christos Chatzichristos and prof. Sabine Van Huffel on automatic denoising of resting-state fMRI using deep learning got accepted into EUSIPCO '21.

09/2020
Received FWO Ph.D. fellowship.

08/2020
Graduated from KU Leuven, M.Sc. in Artificial Intelligence, with Magna Cum Laude.

07/2019
Received scholarship for postgraduate studies by Eugenides Foundation for the academic year of 2019-2020.

06/2019
Received scholarship for postgraduate studies by Bodossaki Foundation for the academic year of 2019-2020.

10/2018
Graduated from National and Technical University of Athens, M.Eng. in Electrical and Computer Engineering, with great distinction.


Papers




2024


Evaluating the Predictive Features of Person-Centric Knowledge Graph Embeddings: Unfolding Ablation Studies

Christos Theodoropoulos, Natasha Mulligan, Joao Bettencourt-Silva
MIE 2024, 34th Medical Informatics Europe Conference
[Code]


2023


GADePo: Graph-Assisted Declarative Pooling Transformers for Document-Level Relation Extraction

Andrei Cătălin Coman, Christos Theodoropoulos, Marie-Francine Moens, James Henderson

Representation Learning for Person or Entity-Centric Knowledge Graphs: An Application in Healthcare

Christos Theodoropoulos, Natasha Mulligan, Thaddeus Stappenbeck, Joao Bettencourt-Silva
K-CAP 2023, 12th Knowledge Capture Conference

An Information Extraction Study: Take In Mind the Tokenization!

Christos Theodoropoulos, Marie-Francine Moens
EUSFLAT 2023, 13th Conference of the European Society for Fuzzy Logic and Technology

2021


Imposing Relation Structure in Language-Model Embeddings Using Contrastive Learning

Christos Theodoropoulos, James Henderson, Andrei Cătălin Coman, Marie-Francine Moens
CoNLL 2021, SIGNLL Conference on Computational Natural Language Learning

Automatic artifact removal of resting-state fMRI with Deep Neural Networks

Christos Theodoropoulos, Christos Chatzichristos, Sabine Van Huffel
EUSIPCO 2021, 29th European Signal Processing Conference

Contact


Χρήστος Θεοδωρόπουλος
E-mail: <x>@kuleuven.be, where x=christos.theodoropoulos
Office: Celestijnenlaan 200A (Room 04.34)(map).


Extra: My first name (Christos) is pronounced as: