How Bad is the Problem of PP-Attachment? A Comparison of English, German and Swedish

What it is about

The paper asks whether PP attachment results can be compared across languages at all. It queries treebanks with TIGER-Search for every noun followed by a PP, and counts how many of these PPs attach to the noun, the noun attachment rate (NAR). For English, sections 0 to 12 of the Wall Street Journal Penn Treebank give 20,858 noun-plus-PP sequences and a NAR of 76.4%. German treebanks give NARs from 59.1% to 63.0%, and Swedish gives 73% and 64% in its two treebanks. The English figure rests on the preposition "of", which covers 40% of the sequences with 99% noun attachment; without "of" the training set's NAR drops to 35%.

Ideas

• I1. Profile every data set Mova uses before any PP experiment: for each preposition, the number of noun-plus-PP sequences and the share attached to the noun. Store the profile next to each en parse-eval run. Measure on EWT first; GUM can be used for tests only.
• I2. Test the "of" effect: report PP attachment accuracy for of phrases and for the rest in en parse-eval. The paper says of is almost always a noun attachment, so a rule for of is the baseline the rest must beat. Measure on EWT test.
• I3. Floor before model: compute the majority-attachment rate per treebank and report it with every PP score, so that a parser's accuracy is read against it, not against 50%.
• I4. Keep the English figures apart from any German or Swedish figures in reports, as the paper argues.

For Mova

The chapters are 22. Phrases: noun, adjective, adverb and prepositional phrases and 23-complementation.md. The code is en/src/parse.rs, and the profile can sit beside the en folds output. The paper's data is the LDC-licensed Penn Treebank, so any measurement on it is test only. Its noun attachment is read from constituent trees, while Mova works with UD arcs, so the rates are not directly comparable.

Topics: Treebanks

Cite

The paper itself, not this card. BibTeX from the library's record of the work.

@inproceedings{volk2006how,
  title = {{How Bad is the Problem of PP-Attachment? A Comparison of English, German and Swedish}},
  author = {Volk, Martin},
  booktitle = {Proceedings of the Third ACL-SIGSEM Workshop on Prepositions},
  year = {2006},
  url = {https://aclanthology.org/W06-2112/}
}

References

19 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Jörgen Aasa. 2004. Unsupervised resolution of PP attachment ambiguities in Swedish. Master’s thesis, Stockholm University. Combined C/D level thesis. find
2. Michael Collins and James Brooks. 1995. Prepositional phrase attachment through a backed-off model. In Proc. of the Third Workshop on Very Large Corpora. find our card
3. Bo Hagström. 2004. A TIGER-XML version of Syn-Tag. Master’s thesis, Stockhom University. find
4. D. Hindle and M. Rooth. 1993. Structural ambiguity and lexical relations. Computational Linguistics, 19(1):103–120. find
5. Jerker Järborg. 1986. SynTag Dokumentation. Manual för SynTaggning. Technical report, Department of Swedish, Göteborg University. find
6. Esther König and Wolfgang Lezius. 2002. The TIGER language - a description language for syntax graphs. Part 1: User’s guidelines. Technical report. find
7. Mitchell P. Marcus, Beatrice Santorini, and Mary Ann Marcinkiewicz. 1993. Building a large annotated corpus of English: The Penn treebank. Computational Linguistics, 19(2):313–330. ↗ our card
8. P. Merlo, M.W. Crocker, and C. Berthouzoz. 1997. Attaching multiple prepositional phrases: generalized backed-off estimation. In Proceedings of the Second Conference on Empirical Methods in Natural Language Processing. Brown University, RI. find
9. A. Ratnaparkhi, J. Reynar, and S. Roukos. 1994. A maximum entropy model for prepositional phrase attachment. In Proceedings of the ARPA Workshop on Human Language Technology, Plainsboro, NJ, March. find our card
10. Adwait Ratnaparkhi. 1998. Statistical models for unsupervised prepositional phrase attachment. In Proceedings of COLING-ACL-98, Montreal. find
11. Douglas L. T. Rohde, 2005. TGrep2 User Manual. MIT. Available from http://tedlab.mit.edu/∼dr/Tgrep2/. ↗
12. Yvonne Samuelsson and Martin Volk. 2005. Presentation and representation of parallel treebanks. In Proc. of the Treebank-Workshop at Nodalida, Joensuu, May. find
13. J. Stetina and M. Nagao. 1997. Corpus-based PP attachment ambiguity resolution with a semantic dictionary. In J. Zhou and K. Church, editors, Proc. of the 5th Workshop on Very Large Corpora, pages 66–80, Beijing and Hongkong. find
14. Ulf Telemann. 1974. Manual För Grammatisk Beskrivning Av Talad Och Skriven Svenska. Inst. för nordiska språk, Lund. find
15. Vincent Vandeghinste. 2002. Resolving PP attachment ambiguities using the WWW (abstract). In Computational Linguistics in the Netherlands, Groningen. find
16. Martin Volk and Yvonne Samuelsson. 2004. Bootstrapping parallel treebanks. In Proc. of Workshop on Linguistically Interpreted Corpora (LINC) at COLING, Geneva. find
17. Martin Volk. 2001. The automatic resolution of prepositional phrase attachment ambiguities in German. Habilitationsschrift, University of Zurich. find
18. Martin Volk. 2002. Combining unsupervised and supervised methods for PP attachment disambiguation. In Proc. of COLING-2002, Taipeh. find our card
19. Martin Volk. 2003. German prepositions and their kin. a survey with respect to the resolution of PP attachment ambiguities. In Proc. of ACL-SIGSEM Workshop: The Linguistic Dimensions of Prepositions and their Use in Computational Linguistics Formalisms and Applications, pages 77–88, Toulouse, France, September. IRIT. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2006-volk-how-bad-is-problem-pp/