Talk by Svetlana Stechikova about dialogue system
Joseph Turian
From: Satoshi Sekine <sek...@cs.nyu.edu>
Date: Thu, Jun 3, 2010 at 5:34 PM
Subject: [NLP around NYC] Talk by Svetlana Stechikova about dialogue system
To: nyc...@cs.nyu.edu, Zofia Stankiewicz
<zosia.st...@gmail.com>, Ligon Liu <tank...@gmail.com>,
takeuchi kosuke <kt28...@hotmail.com>
Cc: Svetlana Stenchikova <svetaste...@gmail.com>, Leslie Cerve
<ce...@cs.nyu.edu>
Dear all,
We will have a talk below at 11:00am-noon on Thursday, June 10th.
If you plan to attend and non-NYU, please send me email,
(in order to know the number of participants).
Thanks,
Satoshi
---------
Speaker: Svetlana Stenchikova (The Open University, England)
http://users.mct.open.ac.uk/ss24382/
Time: 11am-noon on Thursday, June 10th
Location: 715 Broadway, 7th floor conference room
Title: Towards Automatic Generation of COherent Dialogues from Text (CODA)
Abstract:
The goal of the CODA project is to automatically generate
information-delivering dialogues from text. We constructed a resource
(CODA corpus) for extracting rules for converting text into dialogue.
CODA corpus is a parallel corpus of monologues and expository
dialogues. The dialogue part of the corpus consists of expository,
i.e., information-delivering rather than dramatic, dialogues written
by several acclaimed authors. The monologue part of the corpus is a
paraphrase in monologue form of these dialogues by a human annotator.
Using authored dialogues allows us to analyse the techniques used by
accomplished writers for presenting information in the form of
dialogue.
In this talk I will motivate the importance of dialogue generation
task, present the CODA corpus and the monologue-to-dialogue rules
that can be extracted from it.