On Knowledge-Based Machine Translation

What it is about

The paper describes the interlingua of the TRANSLATOR project, which is meant to represent the meaning of texts in a limited subworld. The interlingua has two parts. DIL, the dictionary of the interlingua, describes the concepts and their semantics for the subworld. TIL, the text of the interlingua, holds the instances that a given text produces. Because the interlingua is an artificial language, its syntax and semantics can be controlled. The authors answer four common objections to knowledge-based MT: that full meaning is unnecessary, that translation can start before parsing ends, that syntax can be ignored, and that generation is simple. They reject each, with arguments but no measurements. The work is a design specification; the implementation is future work.

Ideas

• I1. Keep the concept dictionary and the instances in two tables. The meta-world concepts (in meta.rs) should not change with each verse, while the instances for one verse are rebuilt. Count how many concept entries change between runs; that count should be near zero.
• I2. Make the generation choice a counted decision. The paper says generation needs real choices. In the encoder, measure how often the chosen target chunk matches the author's own word against the most frequent chunk, which is the control.
• I3. Define a subworld and count it. Measure how many distinct concepts a book or a genre needs, and how the count grows with each added text. This gives a concrete measure of the subworld the translator covers.

For Mova

The distinction between dictionary and text fits the meta-world, where meta.rs holds the concepts and the reader builds the instances. The objections are useful as a checklist for the design, though they are argued, not tested. The first concrete step is the count of concept changes between two identical runs, which tests the stability claim.

Topics: Machine translation

Cite

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

@inproceedings{nirenburg1986knowledge,
  title = {{On Knowledge-Based Machine Translation}},
  author = {Nirenburg, Sergei and Raskin, Victor and Tucker, Allen},
  booktitle = {Coling 1986 Volume 1: The 11th International Conference on Computational Linguistics},
  year = {1986},
  url = {https://aclanthology.org/C86-1148/}
}

References

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

1. Brachman, R. and J. Schmoltze 1985. An overview on the KL-ONE knowledge representation system. Cognitive Science, vol. 9, issue 2. find
2. Carbonell, J. and M. Tomita 1985. New approaches to machine translation. In: S. Nirenburg (ed.). find
3. Johnson, R. and P. Whitelock 1985. Machine translation as an expert task. In: S. Nirenburg (ed.). Nirenburg, S. (ed.), Proceedings of the Conference on Theoretical and Methodological Issues in Machine Translation. find
4. Hamilton, NY, August 1985. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/nirenburg-1986-knowledge-based-machine-translation/