, 13 tweets, 10 min read Read on Twitter
Marco Baroni is starting the first invited talk at #rep4nlp
"Language is representations by itself. " #ACL2019nlp @sigrep_acl
@sigrep_acl Marco is talking about his previous work about the emergence of language communication between agents.

References:

"MULTI-AGENT COOPERATION
AND THE EMERGENCE OF (NATURAL) LANGUAGE"
arxiv.org/pdf/1612.07182…

"How agents see things"
aclweb.org/anthology/D18-…

#ACL2019nlp
@sigrep_acl Efficient encoding of input information:

"We explore providing some information to the sender and receiver agents and look at the emerging language if it develops to ignore the redundant parts of the input"

Kharitonov et al. 2019
arxiv.org/abs/1905.13687

#ACLnlp2019 #rep4nlp
@sigrep_acl The Straight-through Gumbelsoftmax was used to backpropagate through the discrete sampling process of language generation between agents.
The discreteness of language (lower values of Gumbel softmax temp) is useful to overcome noise "maybe this is why language is discrete in the first place?"

Very interesting point of view to see language!
Human language doesn't only efficiently encode language but also in how it encodes it.

Most frequent words are short while less frequent ones are usually longer

#REP4nlp #ACL2019nlp
Agents have similar patterns as the human language on how to efficiently encode messages forms in between them (assigning short forms to frequent meanings)

@RahmaChaabouni et al.
arxiv.org/abs/1905.12561

#REP4nlp #ACL2019nlp
However, the longer messages are easier to discriminate for the receiver agent so the encoder is not pushed to stick to that

This yields to the "anti-efficient encoding".

Penalizing agents with brevity penalty fixes that but still doesn't reach the efficiency of human languages
@RahmaChaabouni However, the longer messages are easier to discriminate for the receiver agent so the encoder is not pushed to stick to that. This yields to the "anti-efficient encoding".

Penalizing agents with brevity penalty fixes that but still doesn't reach the efficiency of human languages
@RahmaChaabouni Conclusions see pic :)

End of talk.

#ACL2019nlp
#REP4nlp
@RahmaChaabouni Conclusions see pic :)

End of talk.

#ACL2019nlp
#REP4nlp
Full summary of Marco baroni's invited talk at #rep4nlp #acl2019nlp about "emerging (natural) language between agents" 👆👆👆

Thanks Marco for the great talk!!
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Hady Elsahar
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!