An Unsupervised Approach to Prepositional Phrase Attachment using Contextually Similar Words

What it is about

The paper decides whether a preposition attaches to the verb or to the noun before it, without hand-labelled training data. The training examples come from a 125-million-word newspaper corpus parsed with Minipar, and are refined in rounds: each ambiguous attachment adds 1/k to each of its k candidate heads. Triples that are too rare to count are scored through contextually similar words, taken from a thesaurus built on a collocation database of 11 million dependency relations. On 3,097 test examples from the Penn Treebank, the classifier reaches 84.31%. A baseline that puts "of" on the noun and every other preposition on the verb reaches 70.39%, and human agreement is 88.2%. The authors note that the gain from the extra rounds lies inside the confidence intervals.

Ideas

• I1. en/src/parse.rs: keep counts of (verb, prep, noun2) and (noun1, prep, noun2) from the parser's own output on a raw corpus, adding 1/k per ambiguous attachment as the paper does. Measure attachment accuracy on UD EWT test against the same "of to noun, else verb" rule on the same sentences, and the LAS change in en parse-eval.
• I2. Run one round, then two, and compare them. Keep the second round only if two identical runs show a gain that the paired test on EWT test supports.
• I3. Replace the thesaurus with a deterministic back-off: for a rare noun2, use its WordNet lexicographer file from onto/src/wsd.rs. Measure accuracy on the subset of PP cases whose noun2 has fewer than five counts.

For Mova

The chapter is 22. Phrases: noun, adjective, adverb and prepositional phrases, with 23-complementation.md for the verb-or-noun choice. The code is en/src/parse.rs. The paper's 3,097 test set is derived from the Penn Treebank, so it is test only and cannot enter what is built. Its numbers are from 2000 on newspaper text; the first measurement must be on UD EWT test.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3115/1075218.1075232

@inproceedings{pantel2000unsupervised,
  title = {{An Unsupervised Approach to Prepositional Phrase Attachment using Contextually Similar Words}},
  author = {Pantel, Patrick and Lin, Dekang},
  booktitle = {Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics},
  year = {2000},
  doi = {10.3115/1075218.1075232},
  url = {https://doi.org/10.3115/1075218.1075232}
}

References

14 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Altmann, G. and Steedman, M. 1988. Interaction with Context During Human Sentence Processing. Cognition, 30:191-238. find
2. Brill, E. 1995. Transformation-based Error-driven Learning and Natural Language Processing: A case study in part of speech tagging. Computational Linguistics, December. find
3. Brill, E. and Resnik. P. 1994. A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation. In Proceedings of COLING-94. Kyoto, Japan. find
4. Collins, M. and Brooks, J. 1995. Prepositional Phrase Attachment through a Backed-off Model. In Proceedings of the Third Workshop on Very Large Corpora, pp. 27-38. Cambridge, Massachusetts. find our card
5. Hindle, D. and Rooth, M. 1993. Structural Ambiguity and Lexical Relations. Computational Linguistics, 19(1):103-120. find
6. Lin, D. 1999. Automatic Identification of Non-Compositional Phrases. In Proceedings of ACL-99, pp. 317-324. College Park, Maryland. find
7. Lin, D. 1998a. Extracting Collocations from Text Corpora. Workshop on Computational Terminology. Montreal, Canada. find
8. Lin, D. 1998b. Automatic Retrieval and Clustering of Similar Words. In Proceedings of COLING-ACL98. Montreal, Canada. find
9. Lin, D. (1994). Principar - an Efficient, Broad-Coverage, Principle-Based Parser. In Proceedings of COLING-94. Kyoto, Japan. find
10. Miller, G. 1990. Wordnet: an On-Line Lexical Database. International Journal of Lexicography, 1990. find
11. Ratnaparkhi, A. 1998. Unsupervised Statistical Models for Prepositional Phrase Attachment. In Proceedings of COLING-ACL98. Montreal, Canada. find
12. Ratnaparkhi, A., Reynar, J., and Roukos, S. 1994. A Maximum Entropy Model for Prepositional Phrase Attachment. In Proceedings of the ARPA Human Language Technology Workshop, pp. 250-255. Plainsboro, N.J. find our card
13. Roth, D. 1998. Learning to Resolve Natural Language Ambiguities: A Unified Approach. In Proceedings of AAAI-98, pp. 806-813. Madison, Wisconsin. find
14. Stetina, J. and Nagao, M. 1997. Corpus Based PP Attachment Ambiguity Resolution with a Semantic Dictionary. In Proceedings of the Fifth Workshop on Very Large Corpora, pp. 66-80. Beijing and Hong Kong. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2000-pantel-unsupervised-approach-prepositional-phrase-attachment/