Elamur Research⌕
☰
home / papers / 2006 / The SALSA Corpus: a German Corpus Resource for Lexical Semantics

The SALSA Corpus: a German Corpus Resource for Lexical Semantics

Summary

Lexical-semantic resources, both annotated corpora and lexicons, are scarce for nearly every language other than English. This paper presents SALSA, a German resource that adds word senses and semantic roles to the syntactically annotated TIGER newspaper treebank. The annotation follows Frame Semantics and reuses the frames and frame elements of the English Berkeley FrameNet. Each frame is stored as a flat tree whose role edges point to TIGER constituents. The work proceeds predicate by predicate, and every occurrence of a predicate is labelled. Two annotators work independently, and their conflicts are settled in an adjudication step.

The authors report that most FrameNet frames carry over to German. Some cases still required changes:

  • merging roles that FrameNet separates on ontological grounds,
  • adding roles for German dative arguments,
  • handling verbs whose senses cut across English frame boundaries,
  • creating new frames.

The paper also covers quality control: how underspecification is handled, why agreement is reported without kappa, and why systematic errors need separate checks. The first release, planned for mid-2006, contains roughly 20,000 annotated instances. It is intended to support semantic parsing, lexicon building and cross-lingual research.

Key points

  • Scope of the first release: about 20,000 annotated predicate instances, covering roughly half of TIGER. These come from around 500 German verbs across all frequency bands, plus some deverbal nouns.
  • Format, tools and access: annotation was done in the SALTO graphical tool and is stored in SALSA/TIGER XML. The corpus is free for academic research and can be converted to FrameNet XML for querying with FrameSQL.
  • Underlying inventory: the FrameNet release used at the time had about 600 frames, 8,700 lexical units and 133,000 annotated BNC sentences.
  • Example adaptations for German: the SOURCE role of TAKING was split into SOURCE and a new POSSESSOR role to capture dative possessors (as with nehmen). For fahren, which spans drive and ride, the more general frame USE_VEHICLE is the right level.
  • Underspecification statistics:
    • Frame-level underspecification occurred about three times as often as role-level underspecification.
    • Frame underspecification rose slightly during adjudication.
    • About half of the role underspecifications were rejected.

Interesting ideas

  • I1. Stand-off flat trees over a treebank. Each frame is a shallow tree whose role edges point to existing syntactic nodes. This lets a semantic layer be added to any treebank without changing its syntax.
  • I2. Underspecification as a monitored annotation option. Annotators may leave a choice open between frames or roles. The project then checks whether adjudication keeps or removes these choices. This separates genuine vagueness from annotator hesitation and shows which subtask (here, frame choice) is truly hard.
  • I3. Skipping kappa when one label dominates. In role labelling, most constituents get "no role", which inflates chance-corrected agreement figures. Strict labelled exact-match percentages give a more honest picture for many structured annotation tasks.
  • I4. Random audits for shared errors. Double annotation cannot catch mistakes that both annotators make. Sampling and inspecting finished predicate-frame pairs adds a safeguard. The authors also tried intra-annotator outlier detection, but it produced false positives because the data are highly lexicalised.
  • I5. Adapting an English inventory instead of building from scratch. Language-specific needs are handled by merging roles that differ only ontologically, splitting roles where the target language encodes more (such as datives), and moving to more general frames. This is a practical recipe for porting FrameNet-style resources to other languages.

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

Fields: Linguistics & languages

Topics: Treebanks · Semantic role labeling

Cited by

87 papers of our library cite it (of the 140673 whose bibliographies we read); 7 of them have a page here, 80 not yet.

Cite

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

@inproceedings{burchardt2006salsa,
  title = {{The SALSA Corpus: a German Corpus Resource for Lexical Semantics}},
  author = {Burchardt, Aljoscha and Erk, Katrin and Frank, Anette and Kowalski, Andrea and Padó, Sebastian and Pinkal, Manfred},
  booktitle = {Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)},
  year = {2006},
  doi = {10.63317/47fwsi8neeex},
  url = {https://doi.org/10.63317/47fwsi8neeex}
}

References

27 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. C. F. Baker, C. J. Fillmore, and J. B. Lowe. 1998. The Berkeley FrameNet project. In Proceedings of COLING-ACL 1998, Montréal, Canada. find our card
  2. H. C. Boas. 2005. Semantic frames as interlingual representations for multilingual lexical databases. International Journal of Lexicography, 18(4):445–478. ↗
  3. S. Brants, S. Dipper, S. Hansen, W. Lezius, and G. Smith. 2002. The TIGER treebank. In Proceedings of the Workshop on Treebanks and Linguistic Theories, Sozopol. find
  4. A. Burchardt and A. Frank. 2006. Approaching textual entailment with LFG and FrameNet frames. In Proceedings of RTE-2 Workshop, Venice, Italy. find
  5. A. Burchardt, K. Erk, and A. Frank. 2005a. A WordNet Detour to FrameNet. In Sprachtechnologie, mobile Kommunikation und linguistische Resourcen, volume 8 of Computer Studies in Language and Speech. Peter Lang, Frankfurt. find
  6. A. Burchardt, A. Frank, and M. Pinkal. 2005b. Building text meaning representations from contextually related frames – a case study. In Proceedings of IWCS 2005, Tilburg. find
  7. A. Burchardt, A. Frank, K. Erk, A. Kowalski, and S. Padó. 2006. SALTO – a versatile multi-level annotation tool. In Proceedings of LREC 2006, Genoa, Italy. find
  8. M. Ellsworth, K. Erk, P. Kingsbury, and S. Padó. 2004. Prop-Bank, SALSA and FrameNet: How design determines product. In Proceedings of the Workshop on Building Lexical Resources From Semantically Annotated Corpora at LREC 2004. find
  9. K. Erk and S. Padó. 2004. A powerful and versatile XML format for representing role-semantic annotation. In Proceedings of LREC 2004, Lisbon, Portugal. find
  10. K. Erk and S. Padó. 2006. Shalmaneser – a toolchain for shallow semantic parsing. In Proceedings of LREC 2006, Genoa, Italy. find
  11. C. J. Fillmore, C. R. Johnson, and M. Petruck. 2003. Background to FrameNet. International Journal of Lexicography, 16:235– 250. ↗
  12. C. J. Fillmore. 1985. Frames and the semantics of understanding. Quaderni di Semantica, IV(2). find
  13. G. Fliedner. 2006. Towards natural interactive question answering. In Proceedings of LREC 2006, Genoa, Italy. find
  14. A. Frank and K. Erk. 2004. Towards an LFG syntax-semantics interface for frame semantics annotation. In Proceedings of CICLing 2004. find
  15. A. Frank and J. Semecky. 2004. Corpus-based Induction of an LFG Syntax-Semantics Interface for F rame Semantic Processing. In Proceedings of LINC 2004, Geneva, Switzerland. find
  16. A. Frank, H.-U. Krieger, F. Xu, H. Uszkoreit, B. Crysmann, B. Jörg, and U. Schäfer. 2006. Question answering from structured knowledge sources. Journal of Applied Logic, Special Issue on Questions and Answers: Theoretical and Applied Perspectives. to appear. ↗
  17. D. Gildea and D. Jurafsky. 2002. Automatic labeling of semantic roles. Computational Linguistics, 28(3):245–288. ↗ our card
  18. A. Kilgarriff and J. Rosenzweig. 2000. Framework and results for English Senseval. Computers and the Humanities, 34(1-2). ↗
  19. G. Lakoff and M. Johnson. 1980. Metaphors we live by. University of Chicago Press, Chicago, IL. find
  20. E. Miltsakaki, R. Prasad, A. Joshi, and B. Webber. 2004. Annotating discouse connectives and their arguments. In Proceedings of the Workshop on frontiers in corpus annotation at HLT/NAACL. find
  21. K. Ohara, S. Fujii, T. Ohori, R. Suzuki, H. Saito, and S. Ishizaki. 2004. The Japanese FrameNet project: An introduction. In Proceedings of the Workshop on Building Lexical Resources from Semantically Annotated Corpora at LREC 2004. find
  22. S. Padó and K. Erk. 2005. To cause or not to cause: Cross-lingual semantic matching for paraphrase modelling. In Proceedings of the Workshop on Cross-Linguistic Knowledge Induction at EUROLAN 2005. find
  23. M. Palmer, D. Gildea, and P. Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Computational Linguistics, 31(1). ↗ our card
  24. H. Sato. 2003. FrameSQL: A software tool for the FrameNet database. In Proceedings of the 3rd Conference of the Asian Association for Lexicography, Tokyo, Japan. find
  25. S. Siegel and N.J. Castellan. 1988. Nonparametric statistics for the Behavorial Sciences. McGraw-Hill, 2nd edition. find
  26. C. Subirats and M. Petruck. 2003. Surprise! Spanish FrameNet! In Proceedings of the Workshop on Frame Semantics at the XVII. International Congress of Linguists, Prague. find
  27. M. Tatu and D. Moldovan. 2005. A semantic approach to recognizing textual entailment. In Proceedings of HLT/EMNLP 2005, pages 371–378, Vancouver, BC. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/burchardt-2006-salsa-corpus-german-corpus-resource/