*apologies for cross-postings*



CODI, 4th Workshop on Computational Approaches to Discourse



2023-07-13–14 - ACL 2023 - Toronto, Canada



** Submission deadline: April 24th, 2023 **



Aims and scope



The last ten years have seen a dramatic improvement in the ability of NLP systems to understand and produce words and sentences. This development has created a renewed interest in discourse phenomena as researchers move towards the processing of long-form text and conversations. There is a surge of activity in discourse parsing, coherence models, text summarization, corpora for discourse level reading comprehension, and discourse related/aided representation learning, to name a few, but the problems in computational approaches to discourse are still substantial. At this juncture, we have organized three Workshops on Computational Approaches to Discourse (CODI) at EMNLP 2020, EMNLP 2021 and COLING 2022 to bring together discourse experts and upcoming researchers. These workshops have catalyzed work to improve the speed and knowledge needed to solve such problems and have served as a forum for the discussion of suitable datasets and reliable evaluation methods.



The previous workshops on discourse in machine translation (DiscoMT), linking lexical, sentential and discourse semantics (LSDSem), discourse structure in natural language generation (DSNNLG), discourse relation parsing and treebanking (DISRPT) and coreference (CORBON/CRAC), have shown that there is considerable interest and success in bringing together the community working on specific problems in discourse. We believe that the discourse community will also benefit from a general forum where work ranging from corpus development/analysis to computational models, and evaluation is discussed, and desiderata can be drawn for future progress.



The 4th CODI workshop is planned as a 2 day event which brings together different subcommunities. It will feature invited talks and regular papers on the first day. The second day will be dedicated to shared tasks and special sessions which focus on the issues mentioned above. After a first successful iteration in 2019 and 2021 the shared task on Discourse Relation Parsing and Treebanking (DISRPT) will be held again in 2023, with three tasks: discourse segmentation, discourse connective identification and discourse relation classification, including new datasets and languages. For more information on the shared task see:



https://sites.google.com/view/disrpt2023/



Topics of interest



We welcome symbolic and probabilistic approaches, corpus development and analysis, as well as machine and deep learning approaches to discourse. We appreciate theoretical contributions as well as practical applications, including demos of systems and tools. The goal of the workshop is to provide a forum for the community of NLP researchers working on all aspects of discourse.

Topics of interest include, but are not limited to:

discourse structure
discourse connectives
discourse relations
annotation tools and schemes for discourse phenomena
corpora annotated with discourse phenomena
discourse parsing
cross-lingual discourse processing
cross-domain discourse processing
anaphora and coreference resolution
event coreference
argument mining
coherence modeling
discourse and semantics
discourse in applications such as machine translation, summarization, etc.
evaluation methodology for discourse processing



Submissions



We solicit four categories of papers: regular workshop papers, demos, shared task papers and extended abstracts. Only regular workshop papers, shared task papers and demos will be included in the proceedings as archival publications.



Regular papers must describe original unpublished research. Long papers may consist of up to 8 pages of content, plus unlimited pages for references.



Short papers can be up to 4 pages, plus unlimited pages for references.



Demo submissions may describe systems, tools, visualizations, etc., and may consist of up to 4 pages, plus unlimited pages for references.



Each submission can contain unlimited pages for Appendices but the paper submissions need to remain fully self-contained, as these supplementary materials are completely optional, and reviewers are not even asked to review them.

Accepted long, short, and demo papers will be presented orally.



Extended abstracts can describe work in progress or those already published elsewhere. These may be two pages long (without references). Extended abstracts are non-archival. They will be presented orally, and included in the workshop program and handbook, but will not appear in the workshop proceedings.



Double submission of papers is allowed but will need to be indicated at submission.



Submission website



All submissions must be anonymous and follow the ACL 2023 formatting instructions described here: https://2023.aclweb.org/calls/style_and_formatting/

Please submit your workshop papers at https://www.softconf.com/acl2023/CODI2023

Shared task papers should be submitted to the links specified on the shared task pages.



Important dates

2023-04-24: CODI papers due
2023-05-22: Notification of acceptance
2023-06-06: Camera ready deadline for main conference and CODI
2023-07-13 – 2022-07-14: CODI workshop



All deadlines are 11.59 pm UTC -12h ("anywhere on Earth").



Invited Speakers

Yufang Hou, IBM Research
Giuseppe Carenini, University of British Columbia



Organizers

Chloé Braud, CNRS-IRIT
Christian Hardmeier, IT University of Copenhagen and Uppsala University
Jessy Li, University of Texas, Austin
Sharid Loáiciga, University of Gothenburg
Michael Strube, Heidelberg Institute for Theoretical Studies
Amir Zeldes, Georgetown University

To contact the organizers, please send an email to: codi-workshop@googlegroups.com