[Colloquium] REMINDER: 12/3 TTIC Colloquium: Douglas Downey, Northwestern University

[Mary Marre via Colloquium]

*When:    *  Monday, December 3rd at 11:00 am



*Where:     *TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526



*Who:       * Douglas Downey, Northwestern University


*Title:*        Inspecting Word Embeddings with Definition Modeling



*Abstract:* Word embeddings form a foundation for today's natural language
processing systems, and have been shown to encode the syntax and semantics
of terms.  But, exactly what lexical semantics is captured by a given
embedding is often unclear—we typically only observe the semantics
indirectly, through tasks like analogy or word similarity.  In this talk, I
will present definition modeling, which aims to make the semantics of a
given embedding explicit by generating a natural-language definition for
each word conditioned on its embedding.  In experiments, definition models
based on recurrent neural networks are shown to output fluent and accurate
definitions in many cases.  In the second half of the talk, I will
illustrate how the models can be improved using a dynamic regularization
technique that trains an RNN to match given n-gram statistics, which can
reduce for example the excessive repetition often exhibited by RNN text
generators.



Host: Kevin Gimpel <kgimpel at ttic.edu>



For more information on the colloquium series or to subscribe to the
mailing list,please see http://www.ttic.edu/colloquium.php



Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 517*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*


On Sun, Dec 2, 2018 at 5:41 PM Mary Marre <mmarre at ttic.edu> wrote:

> *When:    *  Monday, December 3rd at 11:00 am
>
>
>
> *Where:     *TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>
>
>
> *Who:       * Douglas Downey, Northwestern University
>
>
> *Title:*        Inspecting Word Embeddings with Definition Modeling
>
>
>
> *Abstract:* Word embeddings form a foundation for today's natural
> language processing systems, and have been shown to encode the syntax and
> semantics of terms.  But, exactly what lexical semantics is captured by a
> given embedding is often unclear—we typically only observe the semantics
> indirectly, through tasks like analogy or word similarity.  In this talk, I
> will present definition modeling, which aims to make the semantics of a
> given embedding explicit by generating a natural-language definition for
> each word conditioned on its embedding.  In experiments, definition models
> based on recurrent neural networks are shown to output fluent and accurate
> definitions in many cases.  In the second half of the talk, I will
> illustrate how the models can be improved using a dynamic regularization
> technique that trains an RNN to match given n-gram statistics, which can
> reduce for example the excessive repetition often exhibited by RNN text
> generators.
>
>
>
> Host: Kevin Gimpel <kgimpel at ttic.edu>
>
>
>
> For more information on the colloquium series or to subscribe to the
> mailing list,please see http://www.ttic.edu/colloquium.php
>
>
> Mary C. Marre
> Administrative Assistant
> *Toyota Technological Institute*
> *6045 S. Kenwood Avenue*
> *Room 517*
> *Chicago, IL  60637*
> *p:(773) 834-1757*
> *f: (773) 357-6970*
> *mmarre at ttic.edu <mmarre at ttic.edu>*
>
>
> On Tue, Nov 27, 2018 at 12:46 PM Mary Marre <mmarre at ttic.edu> wrote:
>
>> *When:    *  Monday, December 3rd at 11:00 am
>>
>>
>>
>> *Where:     *TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>>
>>
>>
>> *Who:       * Douglas Downey, Northwestern University
>>
>>
>> *Title:*        Inspecting Word Embeddings with Definition Modeling
>>
>>
>>
>> *Abstract:* Word embeddings form a foundation for today's natural
>> language processing systems, and have been shown to encode the syntax and
>> semantics of terms.  But, exactly what lexical semantics is captured by a
>> given embedding is often unclear—we typically only observe the semantics
>> indirectly, through tasks like analogy or word similarity.  In this talk, I
>> will present definition modeling, which aims to make the semantics of a
>> given embedding explicit by generating a natural-language definition for
>> each word conditioned on its embedding.  In experiments, definition models
>> based on recurrent neural networks are shown to output fluent and accurate
>> definitions in many cases.  In the second half of the talk, I will
>> illustrate how the models can be improved using a dynamic regularization
>> technique that trains an RNN to match given n-gram statistics, which can
>> reduce for example the excessive repetition often exhibited by RNN text
>> generators.
>>
>>
>>
>> Host: Kevin Gimpel <kgimpel at ttic.edu>
>>
>>
>>
>> For more information on the colloquium series or to subscribe to the
>> mailing list,please see http://www.ttic.edu/colloquium.php
>>
>>
>>
>>
>> Mary C. Marre
>> Administrative Assistant
>> *Toyota Technological Institute*
>> *6045 S. Kenwood Avenue*
>> *Room 517*
>> *Chicago, IL  60637*
>> *p:(773) 834-1757*
>> *f: (773) 357-6970*
>> *mmarre at ttic.edu <mmarre at ttic.edu>*
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20181203/1f34a823/attachment-0001.html>