The KANT perspective: a critique of pure transfer (and pure interlingua, pure statistics, .. )

What it is about

The authors argue that the main MT paradigms (transfer, interlingua, statistics, case-based) are each judged by their weakest early form, and that the right question is which combination of methods suits a given task. They separate assimilation (translation for relevance scanning, lower quality, many source languages) from dissemination (high quality, few domains, several target languages). They then describe KANT, an interlingua system for controlled English manuals. Its design keeps the knowledge base minimal: only the domain concepts, and only the role fillers needed to rule out wrong attachments. Interpretation and mapping are separate, so adding a language means writing one set of rules, not one per pair. The paper reports practical figures: 1 to 2 seconds per sentence on old hardware and about ten man-years of effort, with no measured quality score.

Ideas

• I1. Keep a minimal concept set. A meta-world concept is kept only if it lowers the residual on the target texts it is used for. Count the concepts that never help the residual and remove them.
• I2. Use role-filler restrictions to block bad merges. Each meta concept lists which concepts may fill each of its roles, learned from the data. A merge that would put a filler of another type into a role is rejected, and the count of rejected merges is reported.
• I3. Assign each task its own measure. For a relevance-scanning use, score coverage of content words; for a dissemination use, score exact match and the residual. Report both for Mova's output on the same verses.

For Mova

The paper's split between interpretation and mapping matches the reader and the encoder of tools/translate. The concrete places are meta.rs (merges and role fillers) and profile.rs (the encoder). The paper gives no quality numbers to copy, so the first step is a count of concepts that do not change the residual on the dev verses, run twice to confirm identical output.

Topics: Machine translation

Cite

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

@inproceedings{carbonell1992kant,
  title = {{The KANT perspective: a critique of pure transfer (and pure interlingua, pure statistics, .. )}},
  author = {Carbonell, Jaime G. and Mitamura, Teruko},
  booktitle = {Proceedings of the Fourth Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages},
  year = {1992},
  url = {https://aclanthology.org/1992.tmi-1.20/}
}

References

11 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Brown, P., J. Cocke, S. Della Pietra, V. Della Pietra, F. Jelinek, J. Lafferty, R. Mercer and P. Roossin (1990). "A Statistical Approach to Machine Translation," Computational Linguistics 16:79-85. find
2. Brown, R. (1991). "Automatic and Interactive Augmentation," in Goodman and Nirenburg, eds., The KBMT Project: A Case Study in Knowledge-Based Machine Translation, San Mateo, CA: Morgan Kaufmann. find
3. Carbonell, J., R. Cullingford and A. Gershman (1981). "Steps Towards Knowledge-Based Machine Translation, IEEE Transactions on Pattern Analysis and Machine Intelligence, 3:376-392. find
5. Jain, A., A. McNair, A. Waibel, H. Saito, A. Hauptmann and J. Tebelskis (1991). "Connectionist and Symbolic Processing in Speech-to-Speech Translation: The JANUS System," Proceedings of Machine Translation Summit III, Washington, DC, July 2-4. find
6. Lenat, D. and R. V. Guha (1988). "The World According to CYC," MCC Technical Report (ACA-AI-300-88), Austin, TX. find
7. Mitamura, T, E. Nyberg and J. Carbonell (1991). "An Efficient Interlingua Translation System for Multilingual Document Production," Proceedings of Machine Translation Summit III, Washington, DC, July 2-4. find
8. Nirenburg, S. and K. Goodman (1990). "Treatment of Meaning in MT Systems," Proceedings of the Third International Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages, Austin, TX., pp. 171-188. find
9. Nirenburg, S., J. Carbonell, M. Tomita and K. Goodman (1991). Machine Translation: A Knowledge-Based Approach, San Mateo, CA: Morgan Kaufmann. find
10. Sato, S. (1991). "Example-Based Translation Approach to Machine Translation," Proceedings of the International Workshop on Fundamental Research for the Future Generation of Natural Language Processing, ATR Interpreting Telephony Research Laboratories. find
11. Sumita, E. and H. lida (1991). "Experiments and Prospects of Example-Based Machine Translation," Proceedings of the 29th Annual Meeting of the Association for Computational Linguistics, University of California, Berkeley, June 18-21. find
12. Weaver, W. (1949). Translation, Reproduced in Locke and Booth, eds.: Machine Translation of Languages, Cambridge, MA: MIT Press, 1955. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/carbonell-1992-kant-perspective-critique-pure-transfer/