Fwd: LTI Colloquium: Tomorrow, 9/28/18, Marine Carpuat

[Emily Ahn]

Hello everyone,

Tomorrow we have Marine Carpuat from the University of Maryland for our Colloquium!

---------- Forwarded message ---------
From: Tessa Samuelson <tes...@andrew.cmu.edu>
Date: Thu, Sep 27, 2018 at 9:18 AM
Subject: LTI Colloquium: Tomorrow, 9/28/18, Marine Carpuat
To: <lti-s...@cs.cmu.edu>

Please join us for LTI's colloquium tomorrow at 2:30.  

Who: Marine Carpuat

When: Tomorrow, 9/28/18, 2:30

Where: Porter Hall 100

Title:

Semantic and Style Divergences in Machine Translation  

Abstract:

Neural machine translation achieves impressive quality when trained on

millions of parallel sentences (translations produced by

humans). However, there is growing evidence that parallel sentences

are not equally useful. While noise and domain mismatch naturally

affect their usefulness, we hypothesize that more subtle semantic and

style divergences, reflecting choices made by human translators, also

matter for machine translation. We first show that mismatches in the

meaning of source and target are surprisingly frequent in parallel

corpora, and have a substantial impact on neural machine translation

at training and decoding time (Vyas, Niu and Carpuat NAACL 2018). We

then turn to the problem of producing machine translation for a

specific audience by controlling not only the content, but also the

style of the output (Niu, Rao and Carpuat COLING 2018).

Biography:

Marine Carpuat is an Assistant Professor in Computer Science at the

University of Maryland. Her research focuses on multilingual natural

language processing and machine translation. Marine is the recipient

of an NSF CAREER award, research awards from Google and Amazon, best paper awards at *SEM and TALN, and an Outstanding Teaching Award. She received her PhD in Computer Science from the Hong Kong University of Science & Technology, a MPhil in Electrical Engineering from the Hong Kong University of Science & Technology and a Diplome d'Ingenieur from the French Grande Ecole Supelec.

Refreshments will be served at 4:00 in the 5th floor kitchen. 
 
Instructor: Yulia Tsvetkov

Teaching Assistant: Emily Ahn