Controlled English for Knowledge-Based MT: Experience with the KANT System

What it is about

The paper describes KANT Controlled English (KCE), the source language of the KANT system, and how it is built for a heavy-equipment manual. Control has three parts. The lexicon: each domain term gets one meaning per word and part-of-speech pair, split into synonyms where possible, and genuinely ambiguous words are marked for interactive disambiguation with SGML tags. The vocabulary was taken from about 50 MB of existing documentation, through deformatting, word and sentence corpora, and automatic first-pass lexicons, then refined by a lexicographer who browses each word in context, and the domain holds about 50,000 phrases as single entries. The sentences: no coordination of verb phrases, no WH-questions where a direct question will do, relative clauses only with that or which and only on the subject, and rules for participles and punctuation. Authoring is supported by a checker that offers alternative words and asks the author to rewrite a sentence that breaks the grammar, and by SGML tags that fix an attachment or a meaning.

Ideas

• I1. Build the concept lexicon from counts, then let a linguist review each concept in its contexts. A concept view that lists the verses where each merge occurs, with the residual before and after, is the KWIC browser for the meta-world.
• I2. Mark ambiguous source words explicitly in the meta-world. A word with two concept candidates keeps both, and the encoder chooses by the residual, not by a fixed rule. Count how many lemmas of each author are ambiguous in this way.
• I3. Write the phrase-level rule as a test. A multi-word concept (such as abide by) should contain no function-word-only member, the same check that would catch unto_thou in the current lexicon. The test is a count of phrase concepts that fail it.
• I4. Measure the reader's accuracy separately on sentences that satisfy a controlled-grammar test (simple clauses, no coordinated verb phrases) and on the rest. This shows whether the reader's errors come from sentence complexity or from the lexicon.

For Mova

The code is tools/translate (meta.rs for the lexicon and the phrase rule, and measure.rs for the controlled-sentence split) and the review view can be a plain report. The controlled-grammar idea is useful as a test set, not as a constraint on the Bible text, which Mova must read as it is. The three constraints of KCE are a design for authors, which Mova does not have.

Topics: Machine translation

Cite

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

@inproceedings{mitamura1995controlled,
  title = {{Controlled English for Knowledge-Based MT: Experience with the KANT System}},
  author = {Mitamura, Teruko},
  booktitle = {Proceedings of the Sixth Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages},
  year = {1995},
  url = {https://aclanthology.org/1995.tmi-1.12/}
}

References

9 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Adriaens, G. and D. Schreurs (1992). “From COGRAM to ALCOGRAM: Toward a Controlled English Grammar Checker,” Proceedings of COLING-92. find
2. Baker, K., A. Franz, P. Jordan, T. Mitamura and E. Nyberg (1994). “Coping With Ambiguity in a Large-Scale Machine Translation System,” Proceedings of CO LING-94. find
3. DeMauro, P. and M. J. Russo (1984). “Computer Assisted Translation at XE-ROX Corporation,” Proceedings of the 25th Annual Conference of the American Translators Association, New York, NY, September 19-23. find
4. Francis, W. and H. Kuĉera (1982). Frequency Analysis of English Usage, Boston, MA: Houghton Mifflin. find
5. Goodman, K. and S. Nirenburg (eds.) (1991). A Case Study in Knowledge-Based Machine Translation, San Mateo, CA: Morgan Kaufmann. find
6. Grishman. R. and R. Kittredge (eds.) (1986). Analyzing Language in Restricted Domains: Sublanguage Description and Processing, Hillsdale, NJ: Lawrence Erlbaum. find
7. Mitamura, T., E. Nyberg and J. Carbonell (1991). “An Efficient Interlingua Translation System for Multi-lingual Document Production,” Proceedings of Machine Translation Summit III, Washington, DC, July 2-4. find our card
8. Mitamura, T, E. Nyberg and J. Carbonell (1993). “Automated Corpus Analysis and the Acquisition of Large, Multi-Lingual Knowledge Bases for MT,” Proceedings of TMI-93. find
9. Nyberg, E., T. Mitamura and J. Carbonell (1994). “Evaluation Metrics for Knowledge-Based Machine Translation,” Proceedings of COLING-94, Kyoto, Japan, August 5-9. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/mitamura-1995-controlled-english-knowledge-based-mt/