Cross-linguistic Multilayer Semantic Annotation and Parsing with UCCA

These days, many NLP tasks involving meaning (such as Machine Translation and Question Answering) are trained end-to-end, under the assumption that the semantic categories and structures needed to perform the task are somehow encoded in the continuous representations of the learning architecture. 
While this trend has in fact produced significant performance jumps, the lack of interpretability and adequate evaluation metrics remains an issue. 
I want to make the case for symbolic structured meaning representations, particularly Universal Conceptual Cognitive Annotation (UCCA; Abend & Rappoport, 2013), as a modular, anchored, and language-independent intermediary between linguistic form and abstract meaning.

In the first part of my talk, I will motivate and introduce UCCA, its main philosophy, and proposed extensions for higher-level and more fine-grained semantic layers (Prange et al., 2019; ongoing work).
The second part will touch on several techniques for automatic parsing into UCCA (Hershcovich et al. 2017, 2018, 2019; Jiang et al., 2019; ongoing work), as well as its application to text simplification (Sulem et al., 2018b) and evaluation of text-to-text generation (Birch et al., 2016; Sulem et al., 2018a; Choshen and Abend, 2018).


References:

Omri Abend and Ari Rappoport. 2013. Universal Conceptual
Cognitive Annotation (UCCA). In Proc. of
ACL, pages 228–238, Sofia, Bulgaria.

Alexandra Birch, Omri Abend, Ondˇrej Bojar, and
Barry Haddow. 2016. HUME: Human UCCAbased
evaluation of machine translation. In Proc. of
EMNLP, pages 1264–1274, Austin, Texas.	

Leshem Choshen and Omri Abend. 2018. Referenceless
measure of faithfulness for grammatical error
correction. In Proc. of NAACL-HLT, pages 124–129,
New Orleans, Louisiana.

Daniel Hershcovich, Omri Abend, and Ari Rappoport.
2017. A transition-based directed acyclic graph
parser for UCCA. In Proc. of ACL, pages 1127–
1138, Vancouver, Canada.

Daniel Hershcovich, Omri Abend, and Ari Rappoport.
2018. Multitask parsing across semantic representations.
In Proc. of ACL, pages 373–385, Melbourne,
Australia.

Daniel Hershcovich, Omri Abend, and Ari Rappoport.
2019. Content differences in syntactic and semantic
representation. In Proc. of NAACL-HLT.

Wei Jiang, Yu Zhang, Zhenghua Li, and Min Zhang. 2019. 
HLT@ SUDA at SemEval 2019 Task 1: UCCA Graph Parsing as 
Constituent Tree Parsing. arXiv preprint arXiv:1903.04153.

Jakob Prange, Nathan Schneider, and Omri Abend. 
2019. Semantically Constrained Multilayer Annotation:
The Case of Coreference. In Proc. of DMR, pages 164–176,
Florence, Italy.

Elior Sulem, Omri Abend, and Ari Rappoport. 2018a.
Semantic structural evaluation for text simplification.
In Proc. of NAACL-HLT, pages 685–696, New Orleans,
Louisiana.

Elior Sulem, Omri Abend, and Ari Rappoport. 2018b.
Simple and effective text simplification using semantic
and neural methods. In Proc. of ACL, pages 162–
173, Melbourne, Australia.