Marzieh Razavi
About Me
I am currently a Post-doctoral researcher at Idiap research institute. My research interests include sign language recognition and assessment, automatic speech recognition (ASR) in particular for under-resourced languages, acoustic data-driven grapheme-to-phoneme conversion, and automatic subword unit derivation and pronunciation generation.
Education
March 2013 - August 2017
PhD in Electrical Engineering, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland. Supervised by Professor Hervé Bourlard and Dr. Mathew Magimai Doss.
September 2010 - August 2012:
M. Sc. in Computing Science, Simon Fraser University, Burnaby, British Columbia, Canada.
September 2005 - September 2009:
B. Sc. in Computer Engineering - Hardware Engineering, Sharif University of Technology, Tehran, Iran.
Professional Experience
August 2017 - Present:
Post-doctoral Researcher at Idiap Research Institute.
March 2013 - June 2017:
Doctoral Researcher at Idiap Research Institute and École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland.
September 2010 - December 2012:
Research Assistant at Natural Language Lab, Simon Fraser University, Burnaby, Canada.
March 2009 - January 2010:
Staff member, Asr-Gooyesh Corporation, Natural Language Understanding and Dialog Manage-ment Group, Developing spoken dialogue systems
Publications
You can visit my Google Scholar profile.
Journal Papers
Marzieh Razavi, Ramya Rasipuram and Mathew Magimai.-Doss:
Towards Weakly Supervised Acoustic Subword Unit Discovery and Lexicon Development Using Hidden Markov Models. Speech Communication, 2018.
Marzieh Razavi and Mathew Magimai.-Doss:
A Posterior-Based Multi-Stream Formulation for G2P Conversion. Signal Processing Letters , 2017.
Marzieh Razavi, Ramya Rasipuram and Mathew Magimai.-Doss:
Acoustic Data-Driven Grapheme-to-Phoneme Conversion in the Probabilistic Lexical Modeling Framework. Speech Communication, 2016.
Conference Papers
Marzieh Razavi, Ramya Rasipuram and Mathew Magimai.-Doss:
Pronunciation Lexicon Development for Under-Resourced Languages Using Automatically Derived Subword Units: A Case Study on Scottish Gaelic. 4th Biennial Workshop on Less-Resourced Languages, 2015, Poznan, Poland. pdf
Marzieh Razavi and Mathew Magimai.-Doss:
An HMM-Based Formalism for Automatic Subword Unit Derivation and Pronunciation Generation . ICASSP 2015, Brisbane, Australia. pdf
Ramya Rasipuram, Marzieh Razavi and Mathew Magimai.-Doss:
Integrated Pronunciation Learning for Automatic Speech Recognition Using Probabilistic Lexical Modeling. ICASSP 2015, Brisbane, Australia. pdf
Marzieh Razavi and Mathew Magimai.-Doss: On Recognition of Non-Native Speech Using Probabilistic Lexical Model. Interspeech 2014, Singapore city, Singapore. pdf
Marzieh Razavi, Ramya Rasipuram and Mathew Magimai.-Doss:
On Modeling Context-Dependent Clustered States: Comparing HMM/GMM, Hybrid HMM/ANN and KL-HMM Approaches. ICASSP 2014, Florence, Italy. pdf
Ramya Rasipuram,Marzieh Razavi and Mathew Magimai.-Doss:
Probabilistic Lexical Modeling and Unsupervised Training for Zero-Resourced ASR. ASRU 2013, Olomouc, Czech. pdf
Gholamreza Haffari, Marzieh Razavi and Anoop Sarkar: An ensemble model that combines syntactic and semantic clustering for discriminative dependency parsing. ACL 2011, Portland,US. pdf
Reports
Marzieh Razavi and Mathew Magimai.-Doss:
Posterior-Based Multi-Stream Formulation To Combine Multiple Grapheme-to-Phoneme Conversion Techniques, Idiap-RR-33-2015, Idiap, 2015. pdf
Marzieh Razavi, Ramya Rasipuram and Mathew Magimai.-Doss:
Towards Multiple Pronunciation Generation in Acoustic G2P Conversion Framework, Idiap-RR-34-2015, Idiap, 2015. pdf
Honors and Awards
Winter 2015:
Receipient of the Best Student Paper Award at Language and Technology Conference.
Summer 2014:
Awarded the ISCA Student Travel Grant to attend INTERSPEECH 2014.
Fall 2011:
Recipient of Simon Fraser University graduate Fellowship.
Summer 2005: Ranked 258th in the nationwide university entrance exam (among more than 300,000 students), Mathematics section.
Skills
Java: Proficient. Java has been my primary programming language since 2006. I have done various types of projects, from desktop GUIs (Swing) to to scientific computing (Machine learning, NLP).
Python: Proficient. I have done scripting for research in the areas of natural language processing and speech processing.
Matlab: Experienced. I have been using Matlab for signal and speech processing.
C/C++: Experienced. I have developed projects in C/C++ during under-gradute studies.
$\LaTeX$: Fluent and experienced. I have been actively using and developing $\LaTeX$ since 2010 for every type of document (Reports, posters, presentation slides,$\dots$).
Operating Systems: I have worked with various flavours of Linux, MacOS and Windows for both personal and professional use.
Version Control and Build Automation Systems: I have worked with Git and SVN for version controlling in the projects I've been involved.
Contact Information and Social Network Profiles
Emails:
myFirstName.myLastName --at-- idiap.ch,
myFirstName.myLastName --at-- epfl.ch
Duolingo profile
I learn new languages here.
Language Portfolio
Persian: Native.
English: Professional fluency. I use English intensively in the work and study environments. C1 level.
French: Limited fluency. I've been learning French since 2014. Currently in the A2 level.