[FRIAM] Fwd: LTI Colloquium, September 25th

[George Duncan]

George Duncan
Emeritus Professor of Statistics, Carnegie Mellon University
georgeduncanart.com
See posts on Facebook, Twitter, and Instagram
Land: (505) 983-6895
Mobile: (505) 469-4671

My art theme: Dynamic exposition of the tension between matrix order and
luminous chaos.

"Attempt what is not certain. Certainty may or may not come later. It may
then be a valuable delusion."
>From "Notes to myself on beginning a painting" by Richard Diebenkorn.

"It's that knife-edge of uncertainty where we come alive to our truest
power." Joanna Macy.




---------- Forwarded message ---------
From: John Friday <jfriday at cs.cmu.edu>
Date: Mon, Sep 21, 2020 at 12:44 PM
Subject: LTI Colloquium, September 25th
To: <lti-seminar at cs.cmu.edu>


Hi Everyone,

This week we have a double feature at the LTI Colloquium. Both Shruti
Rijhwani and Zirui Wang will be presenting talks on Friday, September 25th
from 1:30 to 2:50 PM EST. The talks will be presented on Zoom
<https://cmu.zoom.us/j/96867532227?pwd=QWtBbTB2ZDFUMit3b1dHK1BnZEhnZz09>,
passcode 883155.

Here's the information on this week's speakers and their topics.

*Shruti Rijhwani* is a PhD student at the Languages Technologies Institute
at Carnegie Mellon University. Her primary research interest is
multilingual natural language processing, with a focus on low-resource and
endangered languages. Her research is supported by a Bloomberg Data Science
Ph.D. Fellowship. Much of her published work focuses on improving named
entity recognition and entity linking for low-resource languages and
domains.

Title: Zero-shot Neural Transfer for Cross-lingual Entity Linking


Abstract:  Cross-lingual entity linking maps a named entity in a source
language to its corresponding entry in a structured knowledge base that is
in a different (target) language. While previous work relies heavily on
bilingual lexical resources to bridge the gap between the source and the
target languages, these resources are scarce or unavailable for many
low-resource languages. To address this problem, we investigate zero-shot
cross-lingual entity linking, in which we assume no bilingual lexical
resources are available in the source low-resource language. Specifically,
we propose pivot-based entity linking, which leverages information from a
high-resource "pivot" language to train character-level neural entity
linking models that are transferred to the source low-resource language in
a zero-shot manner. With experiments on nine low-resource languages and
transfer through a total of 54 languages, we show that our proposed
pivot-based framework improves entity linking accuracy 17% (absolute) on
average over the baseline systems for the zero-shot scenario. Further, we
also investigate the use of language-universal phonological representations
which improves average accuracy (absolute) by 36% when transferring between
languages that use different scripts.


*Zirui Wang* is currently a PhD student at the Language Technologies
Institute (LTI). He works on transfer learning, meta learning, and
multilingual models. He is advised by Jaime Carbonell, Yulia Tsvetkov, and
Emma Strubell.


Title: Cross-lingual Alignment vs Joint Training: A Comparative Study and A
Simple Unified Framework


Abstract:  Learning multilingual representations of text has proven a
successful method for many cross-lingual transfer learning tasks. There are
two main paradigms for learning such representations: (1) alignment, which
maps different independently trained monolingual representations into a
shared space, and (2) joint training, which directly learns unified
multilingual representations using monolingual and cross-lingual objectives
jointly. In this work, we first conduct direct comparisons of
representations learned using both of these methods across diverse
cross-lingual tasks. Our empirical results reveal a set of pros and cons
for both methods, and show that the relative performance of alignment
versus joint training is task-dependent. Stemming from this analysis, we
propose a simple and novel framework that combines these two previously
mutually-exclusive approaches. We show that our proposed framework
alleviates limitations of both approaches and can generalize to
contextualized representations such as Multilingual BERT.


Please reach out to me if you have any questions.


Best wishes,

John Friday
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20200921/ef924f8f/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Zirui & Shruti_poster.pdf
Type: application/pdf
Size: 266220 bytes
Desc: not available
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20200921/ef924f8f/attachment-0001.pdf>