Hi, I am Eva!

Curriculum Vitae
Contact me: spilieva@amazon.com


Evangelia Spiliopoulou



I am an Applied Scientist at Amazon, AWS, where my work focuses on Natural Language Processing (NLP) research and developing NLP applications for AWS customers. I am currently part of the Bedrock Evaluation group, where our goal is to develop novel methodologies and tools to assist automatic evaluation of LLMs.

Prior to Amazon, I did my Ph.D. at Language Technologies Institute, Carnegie Mellon University, where I was advised by Eduard Hovy. My Ph.D. topic was on multi-faceted representations for entities and events, focusing specifically on modeling events and their outcomes in low-resource scenarios.

Milestones:

  • Jan 2024 Joined the Bedrock Evaluation group
  • Aug 2023 Joined the Titan RAG group
  • Dec 2022 Super excited to present our paper EvEntS ReaLM: Event Reasoning of Entity States via Language Models at EMNLP 2022!
  • Oct 2022 Joined Amazon, AWS AI labs
  • Aug 2022 Defended my thesis! Now you can call me Dr. Spiliopoulou ;)
  • May 2021 I passed my Thesis Proposal, under the title Modeling Event Implications via Multi-faceted Entity Representations! Many thanks to my committee for their valuable feedback: Yonatan Bisk, Lori Levin and Alan Ritter.
  • Dec 2020 Presenting our work on Definition Frames at COLING'2020!
  • Sep 2020 Our work on Event-Specific Bias Removal was accepted at EMNLP Findings!
  • May 2020 Excited to start my internship with Dataminr!


    • Selected Research Projects

    Event Implications
    Thesis, 2022

  • How do events impact an entity’s state?
  • How can we teach LLMs what to learn, and apply it in unseen entities?


    • Detecting important sub-events from large-scale crisis events
    Completed in May 2021, internship with Dataminr

  • Detect sub-events from social media that may help in humanitarian aid efforts
  • Information is scattered across multiple posts, need to aggregate & combine
  • Importance of temporality and sub-event dependencies (i.e., each sub-event is dependent, on other sub-events)


    • Event-Specific Bias Removal
    Completed in May 2020

  • Detect critical tweets in real-time crisis scenarios
  • No data for the current event, few previous events of similar nature annotated
  • Data bias problem: use adversarial techniques to retain only the useful information for the task


    • Definition Frames
    Completed in March 2020

  • Design explainable entity representations via the use of definitions
  • Representations with structure help us identify which information is important: less data needed


    • Other Projects

    World Modelers, Sofia
    Funded by DARPA, Advisor: Eduard Hovy
  • Design & Implementation of SOFIA (code) and Project Management
  • Detect causal links & events from noisy documents, real-time


    • SAFT, DAFT
    Funded by DARPA, Advisor: Eduard Hovy
  • Detecting events & arguments from news articles
  • System integration for extraction of entities, relations, events and sentiments into ColdStart++ KBs
  • Final report and code



    • Publications

    * = equal contribution

    Detecting Training Data of Large Language Models via Expectation Maximization.
    Gyuwan Kim, Yang Li, Evangelia Spiliopoulou, Jie Ma, Miguel Ballesteros, and William Yang Wang.
    Under review at ICLR 2025.
    PDF

    General Purpose Verification for Chain of Thought Prompting.
    Vacareanu, Robert, Anurag Pratik, Evangelia Spiliopoulou, Zheng Qi, Giovanni Paolini, Neha Anna John, Jie Ma, Yassine Benajiba, and Miguel Ballesteros.
    Under review.
    PDF

    Few-Shot Data-to-Text Generation via Unified Representation and Multi-Source Learning
    Alexander Hanbo Li, Mingyue Shang, Evangelia Spiliopoulou, Jie Ma, Patrick Ng, Zhiguo Wang, Bonan Min, William Wang, Kathleen McKeown, Vittorio Castelli, Dan Roth, and Bing Xiang
    In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Jul 2023.
    PDF

    EvEntS ReaLM: Event Reasoning of Entity States via Language Models
    In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pp. 1982-1997. 2022.
    Evangelia Spiliopoulou, Artidoro Pagnoni, Yonatan Bisk, and Eduard Hovy.
    PDF Code/Data

    SD^2G: A Novel Framework for Detecting Important Subevents from Crisis Events via Dynamic Semantic Graphs.
    In Proceedings of the Seventh Workshop on Noisy User-gene.rated Text (W-NUT 2021).
    Evangelia Spiliopoulou, Tanay K. Saha, Joel Tetreault, Alejandro Jaimes.
    PDF

    Event-Related Bias Removal for Real-time Disaster Events.
    In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings.
    Evangelia Spiliopoulou*, Salvador Medina Maza*, Eduard Hovy, Alexander Hauptmann.
    PDF & Code/Data

    Definition Frames: Using Definitions for Hybrid Concept Representations
    In Proceedings of the 28th International Conference on Computational Linguistics (COLING 2020).
    Evangelia Spiliopoulou, Artidoro Pagnoni and Eduard Hovy.
    PDF & Code/Data

    Event Detection Using Frame-Semantic Parser.
    In Proceedings of the Events and Stories in the News Workshop.
    Evangelia Spiliopoulou, Eduard Hovy and Teruko Mitamura.
    PDF & Code/Data


              Last updated Dec 8th 2024