About Me
Hi! I’m Liz Salesky (/lɪz səˈlɛski/), a PhD student at the Center for Language and Speech Processing at Johns Hopkins University, advised by Matt Post and Philipp Koehn.
I am very lucky to be supported by the Apple Scholars in AI/ML PhD fellowship.
My research primarily focuses on machine translation and language representations, including how to create models which are more data-efficient and robust to variation across languages and data sources. I co-organize NLP with Friends, an online student seminar, with Abhilasha Ravichander, Yanai Elazar, and Zeerak Waseem.
Previously, I was a Masters student at the Language Technologies Institute at Carnegie Mellon University, where I was advised by Alex Waibel and often collaborated with Alan W Black and the lab at KIT, where I worked in the summers. Before that, I worked at MIT Lincoln Laboratory in the Human Language Technology group from 2012-2017, focused primarily on machine translation and language learning applications. I graduated from Dartmouth College in 2012, where I studied Linguistics and Math. My undergraduate thesis with Ann Irvine compared the linguistic validity of unsupervised segmentation methods.
When not at my computer, I like to learn languages, run, and bike to ice cream!



Publications
2021 |
---|
Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLP
|
Assessing Evaluation Metrics for Speech-to-Speech Translation
|
Robust Open-Vocabulary Translation from Visual Text Representations
|
A surprisal—duration trade-off across and within the world's languages
|
The Multilingual TEDx Corpus for Speech Recognition and Translation
|
FINDINGS OF THE IWSLT 2021 EVALUATION CAMPAIGN
|
SIGTYP 2021 Shared Task: Robust Spoken Language Identification
|
2020 |
SIGTYP 2020 Shared Task: Prediction of Typological Features
|
Relative Positional Encoding for Speech Recognition and Direct Translation
|
A Corpus For Large-Scale Phonetic Typology
|
Generalized Entropy Regularization or: There's Nothing Special about Label Smoothing
|
Phone Features Improve Speech Translation
|
Findings of the 2020 IWSLT Evaluation Campaign
|
SIGMORPHON 2020 Shared Task 0: Typologically Diverse Morphological Inflection
|
Optimizing Segmentation Granularity for Neural Machine Translation
|
2019 |
The IWSLT 2019 Evaluation Campaign
|
Exploring Phoneme-Level Speech Representations for End-to-End Speech Translation
|
CMU-01 at the SIGMORPHON 2019 Shared Task on Crosslinguality and Context in Morphology
|
Fluent Translations from Disfluent Speech in End-to-End Speech Translation
|
2018 |
Towards Fluent Translations from Disfluent Speech
|
KIT Lecture Translator: Multilingual Speech Translation with One-Shot Learning
|
2017 |
KIT’s Multilingual Neural Machine Translation systems for IWSLT 2017
|
The AFRL-MITLL WMT17 Systems: Old, New, Borrowed, BLEU
|
2016 |
The MITLL-AFRL IWSLT 2016 Systems
|
The AFRL-MITLL WMT16 News-Translation Task Systems
|
Operational Assessment of Keyword Search on Oral History
|
2015 |
The MITLL-AFRL IWSLT 2015 MT System
|
The AFRL-MITLL WMT15 System: There’s More than One Way to Decode It!
|
2014 |
The MITLL-AFRL IWSLT 2014 MT System
|
Exploiting Morphological, Grammatical, and Semantic Correlates for Improved Text Difficulty Assessment
|
2013 |
The MIT-LL/AFRL IWSLT-2013 MT system
|
A Language-Independent Approach to Automatic Text Difficulty Assessment for Second-Language Learners
|