PredPatt: Predicate-Argument Extraction from Universal Dependencies

We present PredPatt, a framework of extensible, interpretable, language-neutral predicate-argument extraction patterns. PredPatt bridges the deep syntax of the Universal Dependency project to an initial shallow semantic layer: this can form the basis for future layering of semantic annotations atop Universal Dependency treebanks, and separately can be considered a linguistically well-founded component of a "Universal IE" mechanism.

PredPatt is part of a wider initiative on decompositional semantics at Johns Hopkins University. To that end, it has been used to bootstrap semantic annotations in our recent EMNLP 2016 paper (White et al., 2016).

PredPatt extracts predicates and arguments from text .

?a extracts ?b from ?c
    ?a: PredPatt
    ?b: predicates
    ?c: text
?a extracts ?b from ?c
    ?a: PredPatt
    ?b: arguments
    ?c: text

Table of contents

• Get started (See our new tutorial for programmatic usage.)
• Motivation
• Sample output:
  • Documentation tests: English
  • UD Bank: English, Portuguese, and Spanish.
  • Selected examples: Chinese, Portuguese, Spanish
• How PredPatt works: high-level overview, detailed list of rules
• Evaluation
• Related work
• References

Citation

If you use PredPatt please cite it as follows.

@InProceedings{white-EtAl:2016:EMNLP2016,
    author    = {White, Aaron Steven  and  Reisinger, Drew  and  Sakaguchi, Keisuke  and  Vieira, Tim  and  Zhang, Sheng  and  Rudinger, Rachel  and  Rawlins, Kyle  and  Van Durme, Benjamin},
    title     = {Universal Decompositional Semantics on Universal Dependencies},
    booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing},
    month     = {November},
    year      = {2016},
    address   = {Austin, Texas},
    publisher = {Association for Computational Linguistics},
    pages     = {1713--1723},
    url       = {https://aclweb.org/anthology/D16-1177}
}