Alvin
Grissom II
**As of August, 2017, I am an assistant professor at Ursinus College.
Department of Computer Science
University of Colorado Boulder
UPDATE: I am now an assistant professor at Ursinus College in Collegeville, Pennsylvania.
Before coming here with my
advisor, Jordan Boyd-Graber,
I was a Ph.D. student at the University of Maryland, College
Park.
Prior to that, I received master's degree in computer science
from Emory University. I earned a bachelor of arts
degree
in computer science from Hendrix College.
During Fall 2016, I am teaching undergraduate machine learning.
Research:
I do research in computational
linguistics, and I'm generally
interested in collaborating with linguists, cognitive
scientists, and of course computer scientists.
Reinforcement Learning for Simultaneous
Machine Translation
[Project Page]
[Paper]
Simultaneous interpretation is the process of translating
in real-time sentences from one language to another,
before they're finished, rather than working with
complete sentences.
My colleagues and I are using machine learning to
teach computers to do simultaneous translation between
distant language pairs. Machine translation is hard;
simultaneous interpretation is harder. Simultaneous
interpretation
between distant languages combines the most difficult parts
of both tasks. If you'd like to do simultaneous
interpretation
between, say, Japanese and English, you have a problem.
In Japanese, verbs (and many other important things) come
at the end. So, you must either do some verbal gymnastics
to express your English sentences with sentence-final verbs,
or you must guess what the speaker will say before it's said.
But how do you decide when to trust these guesses? Rather
than tell the system, present the system with some things to
look for and allow it to learn from experience.
Publications:
Alvin C. Grissom II, Naho Orita, and Jordan Boyd-Graber. Incremental Prediction of Sentence-final Verbs: Human versus Machines. CoNLL 2016. (Acceptance rate: 20%)
[PDF] [BibTeX]
He He, Alvin C. Grissom II, Jordan
Boyd-Graber, and Hal Daumé III. Syntax-based
Reordering for Simultaneous Machine
Translation. EMNLP 2015. (Acceptance rate: 25%) [PDF] [BibTeX]
Alvin C. Grissom II, He He, Jordan
Boyd-Graber, John Morgan, and Hal Daumé III. Don’t
Until the Final Verb Wait: Reinforcement
Learning for Simultaneous Machine Translation. EMNLP 2014. (Acceptance rate: 30%) [PDF] [BibTeX][Video]
Alvin C. Grissom II. Sentiment in Japanese: A
Corpus-based Approach with Sociolinguistic Implications.
Master's Thesis. Emory University, 2009. [PDF]
My Non-research:
I also compose music and upload it to
SoundCloud. In high school, I wrote a novel. One
day, I'll write another.