eXtended WordFrameNet
Summary
FrameNet is costly to build by hand and its lexicon is small. It has about 10,000 lexical units (LUs), while WordNet3.0 has over 200,000 word senses, and every new language needs the same manual effort again. The authors link FrameNet LUs to WordNet synsets automatically. They treat all LUs of one frame as a shared context and run a knowledge-based WSD algorithm, SSI-Dijkstra+. It finds shortest paths in a large graph built from WordNet 1.6 relations and eXtended WordNet gloss relations. The original SSI-Dijkstra fails on frames that have no monosemous LUs, so the new variant begins with a best guess for the least ambiguous word. On a gold standard of 372 LUs, SSI-Dijkstra+ beats both the most-frequent-sense baseline and the earlier algorithm. The resulting mappings let the authors:
- add WordNet synonyms to frames, roughly doubling the number of LUs;
- build frame lexicons for other languages from aligned wordnets in the MCR;
- attach SUMO and WordNet Domains labels to LUs.
The approach offers a cheap route to wider frame-semantic coverage, including for languages with little or no FrameNet.
Key points
- Knowledge graph: 99,635 synset nodes and 636,077 edges from WordNet 1.6 and eXtended WordNet. Version 1.7 relations were mapped to 1.6 with the automatic WordNet Mappings.
- Frames without anchors: about 26% of frames (190) contain only polysemous LUs. The original SSI-Dijkstra cannot resolve these cases, which is what SSI-Dijkstra+ fixes.
- Gold standard (Tonelli and Pianta, 2009; 372 LUs, one per frame): SSI-Dijkstra+ reaches F1 0.75 overall (P 0.76, R 0.74). Both wn-mfs and the original SSI-Dijkstra score 0.69.
- Verbal mapping of Shi and Mihalcea (2005): SSI-Dijkstra+ scores F1 0.79, compared with 0.76 for SSI-Dijkstra and 0.67 for wn-mfs.
- Resource size: eXtended WordFrameNet has about 20,600 English LUs. The text gives 20,587 and Table 3 gives 20,857, compared with 9,328 originally.
- Other languages: local WordFrameNets were built for Spanish (14,106 LUs), Italian (12,478), Basque (10,980) and Catalan (13,128). For comparison, Spanish FrameNet has 1,047 LUs.
Cited by
Cite
@inproceedings{laparra2010extended,
title = {{eXtended WordFrameNet}},
author = {Laparra, Egoitz and Rigau, German},
booktitle = {Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10)},
year = {2010},
url = {https://aclanthology.org/L10-1550/}
}References
- J. Atserias, L. Villarejo, G. Rigau, E. Agirre, J. Carroll, B. Magnini, and Piek Vossen. 2004. The meaning multilingual central repository. In Proceedings of GWC, Brno, Czech Republic. find
- C. Baker, C. Fillmore, and J. Lowe. 1998. The berkeley framenet project. In COLING/ACL’98, Montreal, Canada. find our card
- A. Burchardt, K. Erk, and A. Frank. 2005. A Word-Net Detour to FrameNet. In Proceedings of the GLDV 2005 GermaNet II Workshop, pages 408– 421, Bonn, Germany. find
- M. Cuadros and G. Rigau. 2008. Knownet: Building a large net of knowledge from the web. In 22nd International Conference on Computational Linguistics (COLING’08), Manchester, UK. find
- J. Daudé, L. Padró, and G. Rigau. 2003. Validation and Tuning of Wordnet Mapping Techniques. In Proceedings of RANLP, Borovets, Bulgaria. find
- C. Fellbaum, editor. 1998. WordNet. An Electronic Lexical Database. The MIT Press. find
- WordFrameNet lead take direct guide command control SUMO Guiding Synset 01364494-v FrameNet lead.v Managing 01662860-v command.v govern rule Guiding 01763262-v govern.v rule.v power powerfulness potency authority dominance say-so Subjective-Assesment-Attribute NormalAttribute 04041746-n power.n 0404… find
- K. Kipper. 2005. VerbNet: A broad-coverage, comprehensive verb lexicon. Ph.D. thesis, University of Pennsylvania. find
- E. Laparra and G. Rigau. 2009. Integrating wordnet and framenet using a knowledge-based word sense disambiguation algorithm. In Proceedings of CatalanWFN dirigir portar guiar conduir controlar BasqueWFN eroan gidatu eraman ItalianWFN portare condurre menare SpanishWFN llevar conducir murriztu mender… find our card
- E. Laparra, G. Rigau, and M. Cuadros. 2010. Exploring the integration of wordnet and framenet. In Proceedings of GWC, Mumbai, India. find
- B. Magnini and G. Cavaglià. 2000. Integrating subject field codes into wordnet. In Proceedings of LREC, Athens. Greece. find our card
- D. McCarthy, R. Koeling, J. Weeds, and J. Carroll. 2004. Finding predominant senses in untagged text. In Proceedings of ACL, pages 280–297. find our card
- R. Mihalcea and D. Moldovan. 2001. extended wordnet: Progress report. In Proceedings of NAACL Workshop on WordNet and Other Lexical Resources, Pittsburgh, PA. find
- R. Navigli and P. Velardi. 2005. Structural semantic interconnections: a knowledge-based approach to word sense disambiguation. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 27(7):1063–1074. find
- I. Niles and A. Pease. 2001. Towards a standard upper ontology. In Proceedings of the 2nd International Conference on Formal Ontology in Information Systems (FOIS-2001), pages 17–19. Chris Welty and Barry Smith, eds. find
- M. Palmer, D. Gildea, and P. Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Computational Linguistics, 31(1):71–106, March. find our card
- D. Shen and M. Lapata. 2007. Using semantic roles to improve question answering. In Proceedings of the Joint Conference on (EMNLP-CoNLL), pages 12– 21. find
- L. Shi and R. Mihalcea. 2005. Putting pieces together: Combining framenet, verbnet and wordnet for robust semantic parsing. In Proceedings of CI-CLing, Mexico. find
- C. Subirats and M. Petruck. 2003. Surprise: Spanish framenet! In Proceedings of the International Congress of Linguists, Praga. find
- S. Tonelli and E. Pianta. 2009. A novel approach to mapping framenet lexical units to wordnet synsets. In Proceedings of IWCS-8, Tilburg, The Netherlands. find
This page in Markdown → https://md.research.elamur.ai/free/papers/laparra-2010-extended-wordframenet/