School of Electronic Engineering and Computer Science

people menu

Dr Mehrnoosh Sadrzadeh

Lecturer and EPSRC Career Acceleration Fellow

Research Interests

My current research is focused on developing mathematical and logical models and tools to reason about large scale natural language data, collected by counting co-occurrences of words within documents. This data is  turned into vector representations and  acted upon by  linear and multi-linear algebraic operations. The goal is to use the grammar to compose  word vectors and build vector representations for larger language units such as phrases and sentences. The models I work with are based on high level categorical models of grammar and of vector spaces;  they are also inspired by  categorical  models of quantum mechanics.  Tenor spaces and entangled states therein have proved to be invaluable resources for linguistic tasks and have resulted in better experimental outcomes. The tasks I experiment with include word, phrase, and sentence level similarity, entailment, dismabiguation, paraphrasing, and classification. 

Previously,  I worked on algebraic and categorical models of multi-agent information flow. In particular, this work resulted in a sound and complete algebraic semantics and a cut-free sequent calculus with adjoint modalities for a positive and an intuitionistic fragment of dynamic epistemic logic. I also applied these logics to reasoning about navigation.

The above have been joint work with: Abramsky, Casadio, Clark, Coecke, Grefenstette, Kartsaklisk, Milajevs, Pulman, Purver, Preller (for the linguistics part) and Baltag, Coecke, Dyckhoff, Panagaden, Philips, Precup, Truffault (for the dynamic epistemic part).

I am a member of  the following groups:

Topics of Interest
Compositional Distributional Semantics
Category Theoretical Models of Natural Language
Pregroup Grammars, Vector Space Models
Dynamic Epistemic Logic, Algebra and Proof Theory

  • ALCOP,  15-16 May 2014, QMUL.
  • IWCS, 14-17 April 2015, QMUL.
  • ADS, 14 April 2015, QMUL.
  • Clothes Network Ltd: Compositional vector methods for request classification, with FayiMora Femi-Balogun.
  • AFSOR project: Algorithmic and Logical Aspects when Computing Meanings.
  • EPSRC project:  New Foundational Structures for Engineering Verified multi-UAV's.
  • EPSRC Career Acceleration Fellowship: Foundational Structures for Compositional Meaning.
  • Done: EPSRC Postdoctoral Fellowship: (AlCoSeK) Alegbraic and Coalgebraic Semantics for Knowledge.
Invited Talks
Edited Volume

2014 Papers (for the rest see my google scholar page)
  1. Evaluating Neural Word Representations in Tensor-Based Compositional Settings, Dmitrijs Milajevs, Dimitri Kartsaklis, Mehrnoosh Sadrzadeh, Matthew Purver, 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), Qatar, October 2014.
  2. Resolving Lexical Ambiguity in Tensor Regression Models of Meaning, Dimitri Kartsaklis, Nal Kalchbrenner and Mehrnoosh Sadrzadeh, In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (ACL): Short Papers. Baltimore‚ USA, June, 2014. 
  3. Concrete Models and Empirical Evaluations for the Categorical Compositional Distributional Model of Meaning, Ed Grefenstette, Mehrnoosh Sadrzadeh, Journal of Computational Linguistics, to appear.
  4. A Low-Level Treatment of Generalized Quantifiers in Categorical Compositional Distributional Semantics, Ondrej Rypacek and Mehrnoosh Sadrzadeh, Second Workshop on Natural Language and Computer Science, A workshop affiliated with the joint meeting of Computer Science Logic (CSL) and Logic in Computer Science (LiCS), Vienna, July 2014. 
  5. Word Order Alternations in Sanskrit via Precyclicity in Pregroup Grammars, Claudia Casadio and Mehrnoosh Sadrzadeh, In Horizons of the Mind, Essays dedicated to Prakash Panangaden on the occasion of his 60th birthday, LNCS 8464, May 2014. 
  6. A Study of Entanglement in a Categorical Model of Natural Language, Dimitri Kartsaklis and Mehrnoosh Sadrzadeh,  11th workshop on Quantum Logic and Physics (QPL), Kyoto University, June 2014. arXiv:1405.2874.
  7. Semantic Unification: a sheaf theoretic approach to natural language, Samson Abramsky and Mehrnoosh Sadrzadeh, In: Catgories and Types in Logic, Language, and Physics, Essays dedicated to Jim Lambek on the occasion of his 90th birthday, Casadio et al (eds.), LNCS 8222, pages 1- 13, 2014. arXiv:1403.3351
  8. Algebraic Semantics and Model Completeness for Intuitionistic Public Announcement Logic, Minghui Ma, Alessandra Palmigiano and Mehrnoosh Sadrzadeh, Annals of Pure and Applied Logic, Volume 165, Pages 963--995, 2014.
  9. The Frobenius anatomy of word meanings II: possessive relative pronouns, Mehrnoosh Sadrzadeh, Stephen Clark, Bob Coecke, Journal of Logic and Computation, Essays dedicated to Roy Dyckhoff on the occasion of his retirement, S. Graham-Lengrand and D. Galmiche (eds.), doi: 10.1093/logcom/exu027, 2014.
Ph.D. Students
  • Edward Grefenstette (graduated, Oxford,  with Profs Coecke and Pulman)
  • Pere Pardo (graduated, Barcelona, with Prof Godo)
  • Dimitri Kartsaklis (graduated, Oxford, with Profs Coecke and Pulman)
  • Dmitrijs Milajevs (QMUL, with Dr Purver)
  • Esma Balkir (QMUL)
Current Teaching (2013-2014 and 2014-2015)
  • Computability ECS 613 U, EECS, QMUL, joint with Greta Yorsh

Past Teaching and Organization

Return to top