Elamur Research⌕
☰
home / papers / 2010 / Using High-Quality Resources in NLP: The Valency Dictionary of English as a Resource for Left-Associative Grammars

Using High-Quality Resources in NLP: The Valency Dictionary of English as a Resource for Left-Associative Grammars

Paper · 2010 · Thomas Proisl and Besim Kabashi

Summary

The paper argues that rule-based parsing becomes more reliable when it draws on a carefully curated lexical resource rather than hand-built or noisy valency information. Valency, meaning which complements a word demands, is largely idiosyncratic: even near-synonyms can select different prepositions or clause types. The authors take the Valency Dictionary of English (VDE), a corpus-based specialist dictionary of verbs, nouns and adjectives available electronically through the Erlangen Valency Pattern Bank. They convert its patterns into lexical categories for a Left-Associative Grammar (LAG), a formalism that processes a sentence strictly left to right and treats each word's open valency slots as possible continuations. A worked BNC example shows how slots are filled word by word and how competing readings are discarded until one complete dependency analysis remains. The paper reports no quantitative parsing evaluation. Its argument rests on the example, on the dictionary's high corpus coverage, and on a qualitative comparison with COMLEX, VALEX and FrameNet. The claim is that a trustworthy resource lets grammar writers stop worrying about lexical correctness.

Key points

  • The VDE describes 511 verbs, 274 nouns and 544 adjectives (1,329 entries). Its patterns are illustrated with Bank of English examples and were checked by native speakers.
  • Despite its small size, the VDE covers 77.36% of verb tokens in the BNC. Excluding be, do, have and go, it still covers 65.32%. Coverage is 19.30% for adjective tokens and 12.36% for noun tokens.
  • Patterns carry extra marks, such as P(it) for passivizable complements that also allow it-extraposition. Complements are grouped into semantically similar sets (I, II, III) with informal role glosses.
  • In the LAG lexicon, each word has a set of category tuples made of valency slots plus a formal-property segment. Prepositions are given compound slots such as in_v-ing. The authors added subject slots for finite verbs and explicit case marking.
  • The authors report three open problems:
    • The underspecified ADV slot, which can be filled by several phrase types.
    • Collocational hints such as "often how or what", which are hard to encode.
    • Limited entry count.

Interesting ideas

  • I1. Keep every valency reading of a word as a parallel candidate, and drop readings as soon as the next word fails to fit any open slot, so ambiguity is resolved incrementally rather than by global scoring.
  • I2. Treat the end of a sentence as a completeness check: any analysis that still has an unfilled obligatory slot is rejected as ungrammatical.
  • I3. Give valency to nouns and adjectives with preposition-specific slots (master at vs amateur in, helpful in V-ing), not only to verbs.
  • I4. Measure a small curated lexicon by the share of corpus tokens it covers, not by its entry count, to show where high-precision coverage is already strong.
  • I5. Annotate each argument pattern with its allowed transformations (passive subject, extraposition with dummy it), so one lexical entry licenses several surface forms.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (tree): Mova's parser could use lexicon-driven slot pruning as an explainable filter on perceptron or gparse candidates, logging which reading was eliminated and by which word.
  • I2 (learn): Unfilled required slots in a silver tree could act as an automatic negative signal when choosing among witness parsers' outputs.
  • I3 (frames): Frame building could attach roles to noun- and adjective-headed predicates using such slots, filling gaps that VerbNet leaves.
  • I4 (world): Measuring token coverage of compiled verb classes and ontology entries over Mova's test texts would show where hand-curated knowledge pays off and where expansion matters most.
  • I5 (back): The generator could consult these marks to decide when a frame may be realized as a passive or an it-extraposed sentence without producing ungrammatical output.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Friedrich-Alexander-Universität Erlangen-NürnbergErlangen · Germany8 works · 48 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,263 works

Cite

Cite as Thomas Proisl and Besim Kabashi. 2010. Using High-Quality Resources in NLP: The Valency Dictionary of English as a Resource for Left-Associative Grammars. Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10).

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

@inproceedings{proisl2010using,
  title = {{Using High-Quality Resources in NLP: The Valency Dictionary of English as a Resource for Left-Associative Grammars}},
  author = {Thomas Proisl and Besim Kabashi},
  booktitle = {Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10)},
  year = {2010},
  doi = {10.63317/2hy2qvcsdwtc},
  url = {https://doi.org/10.63317/2hy2qvcsdwtc}
}

References

15 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. The British National Corpus. 2007. Version 3 (BNC XML edition). Distributed by Oxford University Computing Services on behalf of the BNC Consortium. http://www.natcorp.ox.ac.uk. ↗
  2. Charles J. Fillmore, Christopher R. Johnson, and Miriam R. L. Petruck. 2003. Background to framenet. International Journal of Lexicography, 16(3):235–250. ↗
  3. Ralph Grishman, Catherine Macleod, and Adam Meyers. 1994. Comlex syntax: Building a computational lexicon. In Proceedings of the 15th International Conference on Computational Linguistics (COLING), volume 1, pages 268–272. ↗
  4. Roland Hausser. 2001. Foundations of Computational Linguistics: Human-Computer Communication in Natural Language. Springer, Berlin, New York, 2nd edition. find
  5. Roland Hausser. 2006. A Computational Model of Natural Language Communication: Interpretation, Inference, and Production in Database Semantics. Springer, Berlin, Heidelberg, New York. ↗
  6. Ulrich Heid. 2007. Valency data for natural language processing: What can the Valency Dictionary of English provide? In Thomas Herbst and Katrin Götz-Votteler, editors, Valency. Theoretical, Descriptive and Cognitive Issues, pages 365–382. Mouton de Gruyter, Berlin, New York. ↗
  7. Thomas Herbst and Peter Uhrig. 2009. Erlangen Valency Pattern Bank – a corpus-based research tool for work on valency and argument structure constructions. Website. http://www.patternbank.uni-erlangen.de. ↗
  8. Thomas Herbst, David Heath, Ian F. Roe, and Dieter Götz. 2004. A Valency Dictionary of English: A Corpus-Based Analysis of the Complementation Patterns of English Verbs, Nouns and Adjectives. Mouton de Gruyter, Berlin, New York. find
  9. Besim Kabashi. 2007. Pronominal clitics and valency in Albanian: A computational linguistics perspective and modelling within the LAG-framework. In Thomas Herbst and Katrin Götz-Votteler, editors, Valency. Theoretical, Descriptive and Cognitive Issues, pages 339–352. Mouton de Gruyter, Berlin, New York. ↗
  10. Anna Korhonen, Yuval Krymolowski, and Ted Briscoe. 2006. A large subcategorization lexicon for natural language processing applications. In Proceedings of the 5th International Conference on Language Resources and Evaluation (LREC), pages 1015–1020, Genua. find
  11. Martin Pickering and Guy Barry. 1993. Dependency categorial grammar and coordination. Linguistics, 31:855–902. ↗
  12. Carl Pollard and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar. University of Chicago Press, Chicago. find
  13. Thomas Proisl. 2008. Integration von Valenzdaten in die grammatische Analyse unter Verwendung des Valency Dictionary of English. Master’s thesis, Philosophische Fakultät und Fachbereich Theologie, Friedrich-Alexander-Universität Erlangen-Nürnberg. find
  14. Dennis Spohr. 2004. Using A Valency Dictionary of English to enhance the lexicon of an English LFG grammar. M. A. study. Institut für maschinelle Sprachverarbeitung, Universität Stuttgart. find
  15. Lucien Tesnière. 1959. Éléments de syntaxe structurale. Klincksieck, Paris. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/proisl-2010-high-quality-resources-nlp-valency/