Lexical Semantics Resources for English

Download

• STREUSLE is a corpus of online reviews hand-annotated for comprehensive multiword expressions and for noun, verb, and preposition supersenses.

  
  Not to be confused with streusel, the delicious dessert topping shown above.

  • A partial example:
    I  googledcommunication restaurantsGROUP inLocus the  areaLOCATION and  Fuji_SushiGROUP came_upcommunication and  reviewsCOMMUNICATION werestative great so I  made_communication a  carry_outpossession _order 
  • The annotations cover the reviews section of the English Web Treebank. The source text and gold part-of-speech tags are redistributed here with the permission of Google and LDC.
  • MWE Annotation Guidelines, new guidelines for prepositional verbs
  • Guidelines and examples for noun supersense annotation (from the AQMAR Arabic Wikipedia Supersense Corpus)
  • Verb Supersense Annotation Guidelines
• English Multiword Expression Lexicons (README.md, LICENSE)
  • This includes 9 lexicons mapped to a common JSON format, as well as a hierarchical clustering of words in a large corpus.
  • The SAID lexicon is distributed by LDC. We provide a script to extract its entries in our JSON format. (It has NLTK 2.0.4 as a dependency.)
• AMALGrAM 2.0 (README.md, LICENSE) is a statistical tagger for multiword expressions and supersenses that was trained on the STREUSLE 2.0 corpus.
  • Dependencies: Python 2.7, Cython, NLTK 3.0.2+ with WordNet downloaded
  • The simplest way to run AMALGrAM is with the pretrained model for English. You will also need to install the lexicons (including SAID for best results).
  • The evaluation script in this release contains a bug affecting Exact Match MWE scores only. The bug is fixed in this version of mweval.py.
• POS tagging/syntactic dependency parsing models for web text: These were trained on the non-reviews sections of the English Web Treebank, i.e., the weblogs, newsgroups, email, and question-answers sections.
  • POS tagging model (1.3MB, requires ARK TweetNLP POS Tagger)
  • Dependency parsing model for Stanford basic dependencies (508MB, requires TurboParser) (Thanks to Lingpeng Kong)

Further Reading

• Comprehensive Annotation of Multiword Expressions in a Social Web Corpus
  Nathan Schneider, Spencer Onuffer, Nora Kazour, Emily Danchik, Michael T. Mordowanec, Henrietta Conrad, and Noah A. Smith.
  In Proceedings of the 9th Linguistic Resources and Evaluation Conference, Reykjavík, Iceland, May 2014.

• Discriminative Lexical Semantic Segmentation with Gaps: Running the MWE Gamut [errata]
  Nathan Schneider, Emily Danchik, Chris Dyer, and Noah A. Smith.
  In Transactions of the Association for Computational Linguistics, 2(April):193−206, 2014.

• A Corpus and Model Integrating Multiword Expressions and Supersenses
  Nathan Schneider and Noah A. Smith.
  In Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Denver, Colorado, June 2015.

• A Hierarchy with, of, and for Preposition Supersenses
  Nathan Schneider, Vivek Srikumar, Jena D. Hwang, and Martha Palmer.
  In Proceedings of the 9th Linguistic Annotation Workshop, Denver, Colorado, June 5, 2015.
• A Corpus of Preposition Supersenses
  Nathan Schneider, Jena D. Hwang, Vivek Srikumar, Meredith Green, Abhijit Suresh, Kathryn Conger, Tim O’Gorman, and Martha Palmer.
  In Proceedings of the 10th Linguistic Annotation Workshop, Berlin, Germany, August 11, 2016.

• SemEval-2016 Task 10: Detecting Minimal Semantic Units and their Meanings (DiMSUM)
  Nathan Schneider, Dirk Hovy, Anders Johannsen, and Marine Carpuat.
  In Proceedings of the 10th International Workshop on Semantic Evaluation, San Diego, California, June 16–17, 2016.

Acknowledgments

This research was supported in part by NSF CAREER grant IIS-1054319, Google through the Reading is Believing project at CMU, and DARPA grant FA8750-12-2-0342 funded under the DEFT program. We are grateful to Google and LDC for permission to redistribute their data along with our annotations.

Contact

Please e-mail nathan.schneider [strudel] georgetown.edu with questions.