About me

Hello there! I'm Silvia Terragni and I'm a Senior NLP Research Engineer at Telepathy Labs, previously a Ph.D. Student at the University of Milano-Bicocca. My research as a PhD Student focused on Topic Models and NLP. My work appeared in the top-ranked NLP venues (ACL, EACL, EMNLP) and Q1 Journals in Artificial Intelligence. I usually open-sourced the code I produced during my PhD. These are my main projects:
  • I collaborated with MilaNLP Lab on the combination of neural topic models and language models' representations. I am a maintainer and contributor of the python package contextualized-topic-models.
  • I designed and developed OCTIS, an open-source and comprehensive python framework for training Topic Models. It uses hyper-parameter Bayesian Optimization to ensure a fairer comparison between the models.

If you want to know more about my projects and stuff, here you can currently find a small selection of my projects, my my talks and my publications . Otherwise we can get in touch on Twitter or by email :)

News and updates

  • I am officially a PhD in Computer Science! (21/02/22)
  • I moved to Zurich to work as a Senior NLP Research Engineer at Telepathy Labs :) (15/11/21)
  • Our paper "Word Embedding-based Topic Similarity Measures" got the Best Paper Award at NLDB 2021! (17/06/21)
  • So excited that our python package contextualized-topic-models got over 430 stars on Github and over 31k downloads! And also OCTIS got over 170 Github stars! (10/06/21)
  • Happy to announce that our paper "Pre-training is a Hot Topic: Contextualized Document Embeddings Improve Topic Coherence" has been accepted to ACL!! (06/05/21)
  • Our first demo paper "OCTIS: Comparing and Optimizing Topic Models is Simple!" has been accepted at EACL2021! We also released a python package :) Click here! (01/03/2021)
  • So excited to announce that our paper "Cross-lingual Zero-shot Topic Models with Contextualized Embeddings" has been accepted at EACL2021! Click here to read the pre-print. (12/01/2021)
  • I delivered a virtual talk about Topic Modeling and recent advances in NLP at AllianceBernstein. Click here for the slides (30/10/20)
  • Happy to share that our paper "Which Matters Most? Comparing the Impact of Concept and Document Relationships in Topic Models" has been accepted at the workshop Insights from Negative Results in NLP! (30/09/20)