Lea Frermann

Research Associate, ILCC, University of Edinburgh

About

I am a research associate at the Institute of Language, Logic and Cognition (ILCC), University of Edinburgh, working with Mirella Lapata and Shay Cohen.

From July to September 2017 was a visiting scholar in Michael Frank's Language and Cognition Lab at Stanford University.

I use machine learning methods and computational models to gain a deeper understanding of the structure and dynamics of meaning representations both in language and in humans. I develop Bayesian models of how children learn categories and their structured featural representations in a joint and incremental process. I am also interested in the dynamics of meaning in natural language. I have developed models of word meaning change over centuries. Most recently I have used deep learning methods to model the structure and development of plots in books and films.


Contact

Informatics Forum, Room 3.38
10 Crichton Street
Edinburgh EH89AB

l.frermann [at] ed [dot] ac [dot] uk



Mini Bio

2017 - now Research associate at ILCC, University of Edinburgh (collaborators Mirella Lapata and Shay Cohen)
2017 Visiting scholar at Language and Cognition Lab, Stanford University (host Michael Frank)
2013 - 2017 PhD at ILCC, University of Edinburgh (supervisors Mirella Lapata and Charles Sutton)
2016 Machine Learning Internship with Amazon Berlin (3 months)
2010 - 2013 MSc in Language Science and Technology from Saarland University (supervisors Ivan Titov and Manfred Pinkal)
2012 Erasmus Mundus Research exchange to NTU Singapore. Research project with Francis Bond.
2007 - 2010 BA in Linguistics, University of Bremen, Germany.


Invited Talks and Presentations


10 / 2017 Saarland University, Saarbruecken, Germany.
Title: Whodunnit? Crime Drama as a Case for Natural Language Understanding
10 / 2017 Alan Turing Institute, London, UK.
Title: Structured dynamic models of meaning for understanding language change and representing book plots
09 / 2017 CoAStaL Copenhagen Natural Language Processing Group, Copenhagen, Denmark.
Title: Structure and Dynamics of Meaning in Humans and in Language
08 / 2017 Stanford NLP Seminar Series, Stanford, California, USA.
Title: Of Space Piracy and Secret Baby Romances: Deep Multi-View Book Representations and a Scalable Evaluation Framework
11 / 2016 Keynote talk at the Drift-a-LOD workshop (co-located with EKAW), Bologna.
Title: Modelling fine-grained Change in Word Meaning over centuries from Large Collections of Unstructured Text
12 / 2015 Heriot-Watt University, Edinburgh, Scotland, UK.
Title: Incremental Bayesian Learning of Semantic Categories and their Features
09 / 2015 Google NLP PhD Summit, Z├╝rich.
07 / 2012 Invited Paper at the First Workshop on Multilingual Modeling (in conjunction with ACL 2012), Jeju, Korea.
Title: Cross-lingual Parse Disambiguation based on Semantic Correspondence


Papers


  • Lea Frermann, Shay B. Cohen and Mirella Lapata, (to appear) Whodunnit? Crime Drama as a Case for Natural Language Understanding, Transactions of the Association for Computational Linguistics (TACL).arXivData

  • Lea Frermann and Michael C. Frank, (2017) Prosodic Features from Large Corpora of Child-Directed Speech as Predictors of the Age of Acquisition of Words, arXiv cs.CL 1709.09443.repo

  • Lea Frermann and Gyuri Szarvas, (2017) Inducing Semantic Micro-Clusters from Deep Multi-View Representations of Novels, In Proceedings of the Conference on Empirical Methods on Natural Language Processing (EMNLP), Copenhagen, Denmark.BibTeXPoster

  • Lea Frermann, (2017) Bayesian Models of Category Acquisition and Meaning Development (Ph.D Thesis Abstract). The IEEE Intelligent Informatics Bulletin, 18, (1), 23.

  • Lea Frermann, (2017) Bayesian Models of Category Acquisition and Meaning Development , Ph.D Thesis, University of Edinburgh, Scotland, UK.

  • Lea Frermann and Mirella Lapata, (2016)A Bayesian Model of Diachronic Meaning Change , Transactions of the Association for Computational Linguistics (TACL).BibTeXSlides

  • Lea Frermann and Mirella Lapata, (2016) Incremental Bayesian Category Learning from Natural Language, Cognitive Science.BibTeX

  • Lea Frermann, (2016)A Bayesian Model of Joint Category and Feature Learning 11th Workshop for Women in Machine Learning (WiML) in conj. with NIPS, Barcelona, Spain.Poster

  • Lea Frermann and Mirella Lapata, (2015) A Bayesian Model for Joint Learning of Categories and their Features, In Proceedings of NAACL-HLT 2015, Denver, Colorado, USA.BibTeXSlides

  • Lea Frermann, (2015) A Bayesian Model of the Temporal Dynamics of Word Meaning 10th Workshop for Women in Machine Learning (WiML) in conj. with NIPS, Montreal, Canada.Poster

  • Lea Frermann and Mirella Lapata, (2014) Incremental Bayesian Learning of Semantic Categories, In Proceedings of EACL 2014, Gothenburg, Sweden.BibTeXDataPoster

  • Lea Frermann, Ivan Titov and Manfred Pinkal, (2014) A Hierarchical Bayesian Model for Unsupervised Induction of Script Knowledge , In Proceedings of EACL 2014, Gothenburg, Sweden.BibTeXSlides

  • Lea Frermann, (2013) A Hierarchical Bayesian Model for Unsupervised Learning of Script Knowledge , MSc Thesis, Saarland University, Germany.

  • Lea Frermann and Francis Bond, (2012) Cross-lingual Parse Disambiguation based on Semantic Correspondence, In Proceedings of ACL 2012, Jeju, Republic of Korea.BibTeX

  • Lea Frermann, (2010) Information Extraction from Written Natural Language Input to Interactive Wayfinding Systems BA Thesis, University of Bremen, Germany.


  • Media Coverage


    Our work on training LSTMs to 'watch' episodes of CSI:Crime Scene Investigation and incrementally guess the perpetrator attracted some wider attention in the media, including articles in the BBC, the New Scientist, The Register and FAZ (in German); and an interview in BBC Scotland Radio's Newsdrive!