WISeR: Semantic Representation

The Widely Interpretable Semantic Representation (WISeR) captures both abstract and concrete concepts as well as their relations in plain text. One major focus of this project is to design a meaning representation that can be easily converted into semantic graphs for probabilistic reasoning on dialogue context. WISeR is distinguished from general Abstract Meaning Representation (AMR) as:

  • It provides deeper and richer levels of abstraction.
  • It is Interpretable by non-experts in computational structures.
  • It covers relations across utterances (e.g., referents, external arguments).

Directors

Distributions

  • WISeR - main repository
  • StreamSide - annotation tool for meaning representations

Related Project

Publications

  1. Automatic Enrichment of Abstract Meaning Representations. Ji, Y.; Williamson, G.; Choi, J. D. Proceedings of the LREC Workshop on Linguistic Annotation (LAW), 2022.
  2. A Cognitive Approach to Annotating Causal Constructions in a Cross-Genre Corpus. Cao, A.; Williamson, G.; Choi, J. D. Proceedings of the LREC Workshop on Linguistic Annotation (LAW), 2022.
  3. StreamSide: A Fully-Customizable Open-Source Toolkit for Efficient Annotation of Meaning Representations. Choi, J. D. and Williamson, G. arXiv, 2109.09853, 2021.
  4. UMR-Writer: A Web Application for Annotating Uniform Meaning Representations. Zhao, J.; Xue, N.; Gysel, J. V.; and Choi, J. D. Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP): System Demonstrations, 2021.
  5. Intensionalizing Abstract Meaning Representations: Non-veridicality and Scope. Williamson, G.; Elliott, P.; Ji, Y. Proceedings of the EMNLP Workshop on Linguistic Annotation Workshop (LAW) and Designing Meaning Representations (DMR), 2021.