Enriching lexical transfer with cross-linguistic semantic features or how to do interlingua without interlingua

What it is about

The paper proposes lexico-structural transfer as an alternative to an interlingua. The source and target are dependency trees in Meaning-Text Theory's deep syntactic level, and transfer rules relate lexemes together with their syntax. The test case is motion verbs: English "swim across the lake" becomes French "traverser a la nage", where the direction goes into the verb and the manner becomes an adjunct. The authors first put the interlingua's components into cross-linguistic semantic features (manner, path, event) on the nodes. That works, but it makes the monolingual lexicons carry features that point only at the interlingua. Their second step uses a lexical function, Adv0, which links a manner adjunct to the head verb. Since the number of paths is fixed by the prepositions of the language, the transfer dictionary stays small, while manners are open. The paper gives examples and no counts; the system, TransLex, is described as implemented.

Ideas

• I1. Treat a lexical function as a meta relation. Where a meta token is a manner adjunct of a head verb, store that as a relation from a closed set rather than as a new concept. Count how many meta tokens in the Bible pairs are manner adjuncts before and after, run twice.
• I2. Carry the distinction in the dependency label instead of the concept. An adjunct (ATTR) and an argument (II) already differ in the UD tree; make the reader keep that label on the meta token as the relation. This is the coarse-relation step of open problem 1.
• I3. Keep the set of meta path features closed, like the prepositions of a language, and list its size after each lexicon round. Open classes (manners) should stay out of the concept table.
• I4. Check the claim that feature-based transfer needs no interlingua by measuring the lexicon size with and without the features on the same data.

For Mova

The places are tools/translate/meta.rs (the lexicon merges), tools/roundtrip (the tree and the frames) and en/src/lin.rs (the linearizer, which must place the adjunct). The closed-feature idea touches open problem 1 and open problem 2, since a merge that changes a relation is not the same as a merge of two words. The paper has no numbers, so every measure above is new.

Topics: Machine translation

Cite

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

@inproceedings{nasr1997enriching,
  title = {{Enriching lexical transfer with cross-linguistic semantic features or how to do interlingua without interlingua}},
  author = {Nasr, Alexis and Rambow, Owen and Palmer, Martha and Rosenzweig, Joseph},
  booktitle = {AMTA/SIG-IL First Workshop on Interlinguas},
  year = {1997},
  url = {https://aclanthology.org/1997.mtsummit-workshop.12/}
}

References

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

1. Dorr, B. J. (1993). Machine Translation: A View from the Lexicon. MIT Press, Boston, Mass. find
2. Dorr, B. J. (1994). Machine translation divergences: A formal description and proposed solution. Computational Linguistics, 20(4):597-635. find our card
3. Hoa Trang Dang, Joseph Rosenzweig, M. P. (1997). Verb classifications as a basis for machine translation. In submitted to the Interlingua Workshop at the MT Summit, San Diego, California. find
4. Joshi, A. K. (1987). An introduction to Tree Adjoining Grammars. In Manaster-Ramer, A., editor, Mathematics of Language, pages 87-115. John Benjamins, Amsterdam. find
5. Kaplan, R. M. and Bresnan, J. W. (1982). Lexical-functional grammar: A formal system for grammatical representation. In Bresnan, J. W., editor, The Mental Representation of Grammatical Relations. MIT Press, Cambridge, Mass. find
6. Levin, B. and Hovav., M. R. (1995). Unaccusativity. At the Syntax-Semantics Interface. MIT Press. find
7. Mel’čuk, I. A. (1988). Dependency Syntax: Theory and Practice. State University of New York Press, New York. find
8. Mel’čuk, I. A. and Polguère, A. (1987). A formal lexicon. Computational Linguistics, 13(3-4):13-54. find
9. Palmer, M. and Rosenzweig, J. (1996). Capturing motion verb generalizations with synchronous TAGs. In Proceedings of AMTA-96, Montreal, Quebec. find
10. Palmer, M., Rosenzweig, J., Dang, H., and Xia, F. (1997). Capturing syntactic/semantic generalizations in a lexicalized grammar, presentation in working session of Semantic Tagging Workshop, ANLP-97. find
11. Rambow, O. and Joshi, A. (1996). A formal look at dependency grammars and phrase-structure grammars, with special consideration of word-order phenomena. In Wanner, L., editor, Current Issues in Meaning-Text Theory. Pinter, London. find
12. Rambow, O., Nasr, A., Palmer, M., Bleam, T., Collins, M., Kipper, K., Melamed, D., Park, J., Rosenzweig, J., Schuler, W., and Srinivas, B. (1997). Machine translation of battlefield messages using lexico-structural transfer. Technical report, CoGenTex, Inc. find
13. Talmy, L. (1991). Path to realization: A typology of event conflation. In Proceedings of the Berkeley Linguistics Society Parasession, Berkeley, CA. find
14. Wanner, L., editor (1996). Lexical Functions in Lexicography and Natural Language Processing. Benjamins, Amsterdam and Philadelphia. We observe that, formally, a LF is really a relation, which is a subset of the cross-product of the lexicon with itself, while a verb class is simply a subset of the lexicon. John find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/nasr-1997-enriching-lexical-transfer-cross-linguistic/