KBMT-89 - A knowledge-based MT project at Carnegie Mellon University

What it is about

KBMT-89 is a prototype interlingua system for English and Japanese, in a subworld of personal-computer installation and maintenance manuals. Its size is listed exactly: an ontology of about 1,500 concepts, roughly 800 to 900 lexical units per language for analysis and for generation, and grammars for both languages. Analysis parses to an LFG-style f-structure, and a mapping-rule interpreter turns it into interlingua frames, using semantic constraints to remove ambiguous readings. A later augmentor handles the readings that remain by asking the user. Generation is in two steps, a semantic step that chooses lexemes and constructions, then a syntactic step that produces the string. The paper is a project overview; it gives sizes but no translation scores.

Ideas

I1. Replace the user dialogue in the augmentor with a count. When the reader leaves several candidate meta-readings for a verse, keep all of them, score them by the residual, and log how many verses have more than one candidate. This is the deterministic version of the paper's residual ambiguity step.
I2. Report the size of the concept set per author and per subworld in the same units the paper uses (concepts, lexical units per language). That makes the translator's size comparable to a known system.
I3. Check that every open-class word points to exactly one concept. The paper's lexicon links each unit to an ontology term. In meta.rs, count the chunks mapped to two concepts and the concepts with no chunk; both counts should be reported.

For Mova

The augmentor is the closest precedent for the ambiguity step of the reader, and the size report is a cheap habit to adopt. The code is tools/translate (meta.rs, hebrew.rs). The paper's ontology is hand-built; in Mova the concept set is learned from the verse pairs, so the size and the one-to-one checks are measures of that learning rather than design choices.

Topics: Machine translation

Cite

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

@inproceedings{nirenburg1989kbmt,
  title = {{KBMT-89 - A knowledge-based MT project at Carnegie Mellon University}},
  author = {Nirenburg, Sergei},
  booktitle = {Proceedings of Machine Translation Summit II},
  year = {1989},
  url = {https://aclanthology.org/1989.mtsummit-1.27/}
}

References

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

[1 ] Nyberg, E. 1988. F RAME K LT User's Guide. Technical Memo, Center for Machine Translation, Carnegie Mellon University. find
[2 ] Tomita, M, and E. Nyberg. 1988. Generation Kit and Transformation Kit Version 3.2 User's Manual. Technical Memo, Center for Machine Translation, Carnegie Mellon University. find
[3 ] Nirenburg, S., R. McCardell, E. Nyberg, P. Werner, S. Huffman, E. Kenschaft, and I. Nirenburg. 1988. D LOGENES -88. Technical Report, Center for Machine Translation, Carnegie Mellon University. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/nirenburg-1989-kbmt-89-knowledge-based-mt/