Universal Decompositional Semantic Types

PredPatt makes very coarse-grained typing distinctions—between predicate and argument nodes, on the one hand, and between dependency and head edges, on the other. UDS provides ultra fine-grained typing distinctions, represented as collections of real-valued attributes. The union of all node and edge attributes defined in UDS determines the UDS type space; any proper subset determines a UDS type subspace.

UDS attributes are derived from crowd-sourced annotations of the heads or spans corresponding to predicates and/or arguments and are represented in the dataset as node and/or edge attributes. It is important to note that, though all nodes and edges in the semantics domain have a type attribute, UDS does not afford any special status to these types. That is, the only thing that UDS “sees” are the nodes and edges in the semantics domain. The set of nodes and edges visible to UDS is a superset of those associated with PredPatt predicates and their arguments.

There are currently four node type subspaces annotated on nodes in sentence-level graphs.

There is currently one edge type subspace annotated on edges in sentence-level graphs.

There is currently (starting in UDS2.0) one edge type subspace annotated on edges in document-level graphs.

Each subspace key lies at the same level as the type attribute and maps to a dictionary value. This dictionary maps from attribute keys (see Attributes in each section below) to dictionaries that always have two keys value and confidence. See the below paper for information on how the these are derived from the underlying dataset.

Two versions of these annotations are currently available: one containing the raw annotator data ("raw") and the other containing normalized data ("normalized"). In the former case, both the value and confidence fields described above map to dictionaries keyed on (anonymized) annotator IDs, where the corresponding value contains that annotator’s response (for the value dictionary) or confidence (for the confidence dictionary). In the latter case, the value and confidence fields map to single, normalized value and confidence scores, respectively.

For more information on the normalization used to produce the normalized annotations, see:

White, Aaron Steven, Elias Stengel-Eskin, Siddharth Vashishtha, Venkata Subrahmanyan Govindarajan, Dee Ann Reisinger, Tim Vieira, Keisuke Sakaguchi, et al. 2020. The Universal Decompositional Semantics Dataset and Decomp Toolkit. Proceedings of The 12th Language Resources and Evaluation Conference, 5698–5707. Marseille, France: European Language Resources Association.

@inproceedings{white-etal-2020-universal,
    title = "The Universal Decompositional Semantics Dataset and Decomp Toolkit",
    author = "White, Aaron Steven  and
      Stengel-Eskin, Elias  and
      Vashishtha, Siddharth  and
      Govindarajan, Venkata Subrahmanyan  and
      Reisinger, Dee Ann  and
      Vieira, Tim  and
      Sakaguchi, Keisuke  and
      Zhang, Sheng  and
      Ferraro, Francis  and
      Rudinger, Rachel  and
      Rawlins, Kyle  and
      Van Durme, Benjamin",
    booktitle = "Proceedings of The 12th Language Resources and Evaluation Conference",
    month = may,
    year = "2020",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://www.aclweb.org/anthology/2020.lrec-1.699",
    pages = "5698--5707",
    ISBN = "979-10-95546-34-4",
}

Information about each subspace can be found below. Unless otherwise specified the properties in a particular subspace remain constant across the raw and normalized formats.

Factuality

Project page

http://decomp.io/projects/factuality/

Sentence-level attributes

factual

First UDS version

1.0

References

White, A.S., D. Reisinger, K. Sakaguchi, T. Vieira, S. Zhang, R. Rudinger, K. Rawlins, & B. Van Durme. 2016. Universal Decompositional Semantics on Universal Dependencies. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1713–1723, Austin, Texas, November 1-5, 2016.

Rudinger, R., White, A.S., & B. Van Durme. 2018. Neural models of factuality. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 731–744. New Orleans, Louisiana, June 1-6, 2018.

@inproceedings{white-etal-2016-universal,
    title = "Universal Decompositional Semantics on {U}niversal {D}ependencies",
    author = "White, Aaron Steven  and
      Reisinger, Dee Ann  and
      Sakaguchi, Keisuke  and
      Vieira, Tim  and
      Zhang, Sheng  and
      Rudinger, Rachel  and
      Rawlins, Kyle  and
      Van Durme, Benjamin",
    booktitle = "Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2016",
    address = "Austin, Texas",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D16-1177",
    doi = "10.18653/v1/D16-1177",
    pages = "1713--1723",
}

@inproceedings{rudinger-etal-2018-neural-models,
    title = "Neural Models of Factuality",
    author = "Rudinger, Rachel  and
      White, Aaron Steven  and
      Van Durme, Benjamin",
    booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers)",
    month = jun,
    year = "2018",
    address = "New Orleans, Louisiana",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/N18-1067",
    doi = "10.18653/v1/N18-1067",
    pages = "731--744",
}

Genericity

Project page

http://decomp.io/projects/genericity/

Sentence-level attributes

arg-particular, arg-kind, arg-abstract, pred-particular, pred-dynamic, pred-hypothetical

First UDS version

1.0

References

Govindarajan, V.S., B. Van Durme, & A.S. White. 2019. Decomposing Generalization: Models of Generic, Habitual, and Episodic Statements. Transactions of the Association for Computational Linguistics.

@article{govindarajan-etal-2019-decomposing,
    title = "Decomposing Generalization: Models of Generic, Habitual, and Episodic Statements",
    author = "Govindarajan, Venkata  and
      Van Durme, Benjamin  and
      White, Aaron Steven",
    journal = "Transactions of the Association for Computational Linguistics",
    volume = "7",
    month = mar,
    year = "2019",
    url = "https://www.aclweb.org/anthology/Q19-1035",
    doi = "10.1162/tacl_a_00285",
    pages = "501--517"
}

Time

Project page

http://decomp.io/projects/time/

Sentence-level attributes

normalized

dur-hours, dur-instant, dur-forever, dur-weeks, dur-days, dur-months, dur-years, dur-centuries, dur-seconds, dur-minutes, dur-decades

raw

duration

Document-level attributes

raw

rel-start1, rel-start2, rel-end1, rel-end2

First UDS version

1.0 (sentence-level), 2.0 (document-level)

References

Vashishtha, S., B. Van Durme, & A.S. White. 2019. Fine-Grained Temporal Relation Extraction. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019), 2906—2919. Florence, Italy, July 29-31, 2019.

@inproceedings{vashishtha-etal-2019-fine,
    title = "Fine-Grained Temporal Relation Extraction",
    author = "Vashishtha, Siddharth  and
      Van Durme, Benjamin  and
      White, Aaron Steven",
    booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/P19-1280",
    doi = "10.18653/v1/P19-1280",
    pages = "2906--2919"
}

Notes

  1. The Time dataset has different formats for raw and normalized annotations. The duration attributes from the normalized version are each assigned an ordinal value in the raw version (in ascending order of duration), which is assigned to the single attribute duration.

  2. The document-level relation annotations are only available in the raw format and only starting in UDS2.0.

Entity type

Project page

http://decomp.io/projects/word-sense/

Sentence-level attributes

supersense-noun.shape, supersense-noun.process, supersense-noun.relation, supersense-noun.communication, supersense-noun.time, supersense-noun.plant, supersense-noun.phenomenon, supersense-noun.animal, supersense-noun.state, supersense-noun.substance, supersense-noun.person, supersense-noun.possession, supersense-noun.Tops, supersense-noun.object, supersense-noun.event, supersense-noun.artifact, supersense-noun.act, supersense-noun.body, supersense-noun.attribute, supersense-noun.quantity, supersense-noun.motive, supersense-noun.location, supersense-noun.cognition, supersense-noun.group, supersense-noun.food, supersense-noun.feeling

First UDS version

1.0

Notes

  1. The key is called wordsense because the normalized annotations come from UDS-Word Sense (v1.0).

References

White, A.S., D. Reisinger, K. Sakaguchi, T. Vieira, S. Zhang, R. Rudinger, K. Rawlins, & B. Van Durme. 2016. Universal Decompositional Semantics on Universal Dependencies. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1713–1723, Austin, Texas, November 1-5, 2016.

@inproceedings{white-etal-2016-universal,
    title = "Universal Decompositional Semantics on {U}niversal {D}ependencies",
    author = "White, Aaron Steven  and
      Reisinger, Dee Ann  and
      Sakaguchi, Keisuke  and
      Vieira, Tim  and
      Zhang, Sheng  and
      Rudinger, Rachel  and
      Rawlins, Kyle  and
      Van Durme, Benjamin",
    booktitle = "Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2016",
    address = "Austin, Texas",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D16-1177",
    doi = "10.18653/v1/D16-1177",
    pages = "1713--1723",
}

Semantic Proto-Roles

Project page

http://decomp.io/projects/semantic-proto-roles/

Sentence-level attributes

was_used, purpose, partitive, location, instigation, existed_after, time, awareness, change_of_location, manner, sentient, was_for_benefit, change_of_state_continuous, existed_during, change_of_possession, existed_before, volition, change_of_state

References

Reisinger, D., R. Rudinger, F. Ferraro, C. Harman, K. Rawlins, & B. Van Durme. (2015). Semantic Proto-Roles. Transactions of the Association for Computational Linguistics 3:475–488.

White, A.S., D. Reisinger, K. Sakaguchi, T. Vieira, S. Zhang, R. Rudinger, K. Rawlins, & B. Van Durme. 2016. Universal Decompositional Semantics on Universal Dependencies. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1713–1723, Austin, Texas, November 1-5, 2016.

@article{reisinger-etal-2015-semantic,
    title = "Semantic Proto-Roles",
    author = "Reisinger, Dee Ann  and
      Rudinger, Rachel  and
      Ferraro, Francis  and
      Harman, Craig  and
      Rawlins, Kyle  and
      Van Durme, Benjamin",
    journal = "Transactions of the Association for Computational Linguistics",
    volume = "3",
    year = "2015",
    url = "https://www.aclweb.org/anthology/Q15-1034",
    doi = "10.1162/tacl_a_00152",
    pages = "475--488",
}

@inproceedings{white-etal-2016-universal,
    title = "Universal Decompositional Semantics on {U}niversal {D}ependencies",
    author = "White, Aaron Steven  and
      Reisinger, Dee Ann  and
      Sakaguchi, Keisuke  and
      Vieira, Tim  and
      Zhang, Sheng  and
      Rudinger, Rachel  and
      Rawlins, Kyle  and
      Van Durme, Benjamin",
    booktitle = "Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2016",
    address = "Austin, Texas",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D16-1177",
    doi = "10.18653/v1/D16-1177",
    pages = "1713--1723",
}

Event structure

Project page

http://decomp.io/projects/event-structure/

Sentence-level attributes

normalized

distributive, dynamic, natural_parts, part_similarity, telic, avg_part_duration_lbound-centuries, avg_part_duration_ubound-centuries, situation_duration_lbound-centuries, situation_duration_ubound-centuries, avg_part_duration_lbound-days, avg_part_duration_ubound-days, situation_duration_lbound-days, situation_duration_ubound-days, avg_part_duration_lbound-decades, avg_part_duration_ubound-decades, situation_duration_lbound-decades, situation_duration_ubound-decades, avg_part_duration_lbound-forever, avg_part_duration_ubound-forever, situation_duration_lbound-forever, situation_duration_ubound-forever, avg_part_duration_lbound-fractions_of_a_second, avg_part_duration_ubound-fractions_of_a_second, situation_duration_lbound-fractions_of_a_second, situation_duration_ubound-fractions_of_a_second, avg_part_duration_lbound-hours, avg_part_duration_ubound-hours, situation_duration_lbound-hours, situation_duration_ubound-hours, avg_part_duration_lbound-instant, avg_part_duration_ubound-instant, situation_duration_lbound-instant, situation_duration_ubound-instant, avg_part_duration_lbound-minutes, avg_part_duration_ubound-minutes, situation_duration_lbound-minutes, situation_duration_ubound-minutes, avg_part_duration_lbound-months, avg_part_duration_ubound-months, situation_duration_lbound-months, situation_duration_ubound-months, avg_part_duration_lbound-seconds, avg_part_duration_ubound-seconds, situation_duration_lbound-seconds, situation_duration_ubound-seconds, avg_part_duration_lbound-weeks, avg_part_duration_ubound-weeks, situation_duration_lbound-weeks, situation_duration_ubound-weeks, avg_part_duration_lbound-years, avg_part_duration_ubound-years, situation_duration_lbound-years, situation_duration_ubound-years

raw

dynamic, natural_parts, part_similarity, telic, avg_part_duration_lbound, avg_part_duration_ubound, situation_duration_lbound, situation_duration_ubound

Document-level attributes

pred1_contains_pred2, pred2_contains_pred1

First UDS version

2.0

Notes

  1. Whether dynamic, situation_duration_lbound, and situation_duration_ubound are answered or part_similarity, avg_part_duration_lbound, and avg_part_duration_ubound are answered is dependent on the answer an annotator gives to natural_parts. Thus, not all node attributes will necessarily be present on all nodes.

References

Gantt, W., L. Glass, & A.S. White. 2021. Decomposing and Recomposing Event Structure. arXiv:2103.10387 [cs.CL].

@misc{gantt2021decomposing,
    title={Decomposing and Recomposing Event Structure},
    author={William Gantt and Lelia Glass and Aaron Steven White},
    year={2021},
    eprint={2103.10387},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}