Workshop website: https://nsnli.github.io/
Call For Papers We invite authors to submit short and positional papers in IJCAI double-column format (Author's Kit<https://www.ijcai.org/authors_kit>) discussing the reasoning aspects in natural language understanding tasks: * Short papers (maximum 4 pages, excluding references) should discuss a technical contribution or datasets creating/curation/modification related to reasoning aspects in NLU tasks. Short papers discussing technical contributions should include some validation of the idea, but we are not expecting an extensive and bullet-proof analysis; limited analysis is acceptable. Short papers discussing dataset aspects should provide some estimation of how existing methods perform on the new dataset. * Position papers (maximum 2 pages, excluding references) can discuss ideas without evaluation but have to make them concrete and actionable. The impact of the idea should be clear. We seek papers that explore NLU topics on the shared resources which, we believe, can kick-start the analysis of different models along specific reasoning dimensions. The shared resources include: * The TaxiNLI<https://nsnli.github.io/taxinli/> dataset is from a recent CoNLL 2020<https://www.aclweb.org/anthology/2020.conll-1.4.pdf> work, which introduces a taxonomy of 15 reasoning capabilities inspired from first principles of Language and Logic. The authors also released a large part of the MultiNLI dataset re-annotated with reasoning categories required for inferencing. * The CheckList<https://nsnli.github.io/checklist/> resource instructs how to create templated test-suites<https://homes.cs.washington.edu/~marcotcr/acl20_checklist.pdf> to specifically test different reasoning categories. The short papers may involve: * analysis or extensions of the shared datasets and taxonomy of reasoning, * extension of CheckList for NLI for testing different aspects of reasoning (multiple hops, inductive/deductive/abductive and so on), * Extended Abstract with interesting use-cases using a neuro-symbolic method, and * A demonstration involving finding bugs in reasoning capabilities of NLI/NLU models and enhance with human feedback. The papers do not need to introduce novel technical contributions. We are interested in insights that identify the limitation of neuro-symbolic methods and push the field forward. Applying existing techniques to shared resources would make an interesting submission to this workshop. We are explicitly interested in (1) works that are deemed too simple or too incremental to publish at a conference (we as the organisers don't agree with this view), and (2) work-in-progress papers that are not fully matured yet. The paper can approach the problem from any of the following areas: * Integration of reasoning and Learning * Natural Language Logical Reasoning Tasks and Datasets * Neuro-symbolic Reasoning for Natural Language * Neuro-symbolic Program Synthesis for Natural Language * Program Synthesis and Deep Learning * Integrating knowledge using Deep Learning (Knowledge Distillation, Relational Reasoning) * Commonsense knowledge and Reasoning * Applications of Probabilistic Logical Reasoning * Inference in Knowledge Bases * Reasoning in Natural Language and X (Computer Vision, Robotics) Important Dates Submission: May 1 (11:59pm AoE) Notification: May 25 (11:59pm AoE) Camera Ready: June 30 (11:59pm AoE) Workshop: August 21 Submission Instructions Submissions to the workshop are limited to 4 pages of content (excluding references and appendices). Please note, while making decisions, the reviewers may not consider the appendix content. The submission process is single-blind. All submissions must be in IJCAI double-column format (IJCAI 2021 Author's Kit<https://www.ijcai.org/authors_kit>). Submission link: OpenReview Site<https://openreview.net/group?id=ijcai.org/IJCAI/2021/Workshop/NSNLI>. Please note, accepted papers in this workshop will not be archival in nature (in the sense that we do not wish to publish an official proceedings).?
_______________________________________________ uai mailing list uai@engr.orst.edu https://it.engineering.oregonstate.edu/mailman/listinfo/uai
