Mova0.5 pre-release⌕
☰
home / docs / papers / Machine translation / An Efficient Interlingua Translation System for Multi-lingual Document Production

An Efficient Interlingua Translation System for Multi-lingual Document Production

What it is about

The KANT prototype aims at knowledge-based translation without post-editing and without human help during disambiguation. The price is a controlled source language: a basic vocabulary of about 14,000 word senses plus a fixed set of technical terms, and a list of banned constructions that cause ambiguity in parsing. The source is parsed with an LFG grammar, and an interpreter maps the parse into an interlingua. A domain model of over 500 concept frames narrows the possible phrase attachments, and a separate mapper and generator produce each target language. Knowledge sources are compiled into fast decision structures for run-time speed. The paper reports 100 percent accuracy on over 200 pre-authored sentences, and 1 to 3 seconds per sentence on old workstations. The test is small and is on pre-authored text.

Ideas

For Mova

The attachment problem is the same one the Hebrew reader and the verse-pair readers meet, so the attachment table belongs in tools/translate (meta.rs) next to the reader. The controlled-language idea is a useful analogy for the test sets: a list of constructions the reader covers, kept with its own error count. The paper gives no quality measure beyond its small test, so each idea needs a count first.

Topics: Machine translation

Cite

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

@inproceedings{mitamura1991efficient,
  title = {{An Efficient Interlingua Translation System for Multi-lingual Document Production}},
  author = {Mitamura, Teruko and Nyberg, Eric H. and Carbonell, Jaime G.},
  booktitle = {Proceedings of Machine Translation Summit III: Papers},
  year = {1991},
  url = {https://aclanthology.org/1991.mtsummit-papers.9/}
}

References

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

  1. Carbonell, J. G, and M. Tomita (1987). “Knowledge-based Machine Translation: The CMU Approach,” in Nirenburg, S. (ed.), Machine Translation: Theoretical and Methodological Issues, New York: Cambridge University Press. find
  2. Carlson, L. and S. Nirenburg (1990). World Modelling for NLP, Technical Report CMU-CMT-90-121, Center for Machine Translation, Carnegie Mellon University. find
  3. DeMauro, P. and M. J. Russo (1984). “Computer Assisted Translation at XEROX Corporation,” Proceedings of the 25th Annual Conference of the American Translators Association, New York, NY, September 19-23. find
  4. Goodman and Nirenburg, eds. (1991). A Case Study in Knowledge-Based Machine Translation, San Mateo, CA: Morgan Kaufmann. find
  5. Kaufmann, T. (1991). The ONTOS User’s Guide. Technical Memo, Center for Machine Translation, Carnegie Mellon University. find
  6. Mitamura, T. (1989). The Hierarchical Organization of Predicate Frames for Interpretive Mapping in Natural Language Processing, PhD thesis, University of Pittsburgh. find
  7. Mitamura, T and E. Nyberg (1990). “Multiple Inheritance and Interpretive Mapping in Machine Translation,” unpublished manuscript. find
  8. Nyberg, E., R. McCardell, D. Gates and S. Nirenburg (1991). “Target Text Generation,” in Goodman and Nirenburg (eds), A Case Study in Knowledge-Based Machine Translation, San Mateo, CA: Morgan Kaufmann. find
  9. Tomita, M., T. Mitamura, H. Musha, and M. Kee (1988). The LR Parser-Compiler User's Guide, Version 8.1, CMU-CMT-88-MEMO. Center for Machine Translation, Carnegie Mellon University, Safety Warnings Read the “General Installation Information” section of this manual. Then, follow the instructions in the “Safety Warnings” section, In order to prevent a fire hazard, do not overload AC outlets. In the following cases, TV sets can overheat: 1. The ventilation slots are blocked. 2. The TV set is placed in a built-in enclosure. Periodically clean the ventilation slots with your vacuum cleaner. If the TV set has been dropped, a shock hazard may exist. In this case, unplug the TV set. Then call your dealer. Figure 3: Sample English Source Text Input to KANT Conseils de sécurité Consulter la section de ce manuel intitulée “Renseignements pour installation”. Ensuite, se conformer aux instructions figurant à la section intitulée “Conseils de sécurité”. Afin d’éviter tout risque d’incendie, ne jamais surcharger les prises CA. Dans les cas suivants, un téléviseur peut surchauffer: 1. La grille de ventilation est bloquée. 2, Le téléviseur est placé dans un coin renfoncé. Dépoussiérer périodiquement la grille de ventilation à l’aide d’un aspirateur. La chute du téléviseur peut provoquer un risque de choc électrique. En ce cas, débrancher le té1éviseur. Ensuite faire appel au détaillant. Figure 4: French Target Text Produced by KANT Sicherheitsbestimmungen Lesen Sie den Abschnitt “Allgemeine Informationen zur Installation” in diesem Handbuch. Folgen Sie dann den Anweisungen in dem Abschnitt “Sicherheitsbestimmungen”. Vermeiden Sie Feuergefahren, indem Sie die Netzanschlüsse nicht überlasten. Fernsehgeräte können in den folgenden Fällen überhitzen: 1. Die Kühlschlitze sind blockiert. 2. Das Fernsehgerät steht in einem Einbauschrank. Reinigen Sie regelmäßig die Kühlschlitze mit dem Staubsauger. Wenn Sie das Fernsehgerät fallenlassen, kann die Gefahr eines Elektroschocks bestehen. Ziehen Sie in diesem Fall den Netz-stecker. Verständigen Sie dann Ihren Kundendienst. Figure 5: German Target Text Produced by KANT Figure 6: Japanese Target Text Produced by KANT (*ALL) (*OBJECT (is-a *ALL)) (*EVENT (is-a *ALL)) ( *PHYSICAL-OBJECT (is-a *OBJECT)) (*E-CLEAN (is-a (agent (theme *EVENT) *USER) *PHYSICAL-LOCATION *PHYSICAL-OBJECT) (instrument *O-CLEANING-INSTRUMENT)) (*PHYSICAL-LOCATION (is-a *OBJECT)) (*O-APERTURE (is-a *PHYSICAL-LOCATION)) (*O-CLEANING-INSTRUMENT (is-a *PHYSICAL-OBJECT)) (*O-VENTILATION-SLOT (is-a *O-APERTURE)) (*O-VACUUM-CLEANER (is-a *O-CLEANING-INSTRUMENT)} (*O-MANUAL (is-a *PHYSICAL-OBJECT)) Figure 7: Partial Domain Knowledge (:ROOT "clean" :CAT V :CLASS verbs-of-cleaning :HEAD *E-CLEAN) (:ROOT "periodically" :CAT ADV :HEAD NIL :SEM ((EVENT-FREQUENCY &PERIODICALLY))) (:ROOT "vacuum cleaner" :CAT N :HEAD *O-VACUUM-CLEANER) (:ROOT "ventilation slot." :CAT N :HEAD *O-VENTILATION-SLOT) (:ROOT "manual" :CAT N :HEAD *O-MANUAL) Figure 8: Example Lexical Mappings * (translate sent8) ; *Periodically, clean the ventilation slots with your vacuum cleaner." 1 source f-structure(s) found in 0.89 seconds of real time ((MOOD DP) (FORM ROOTFORM) (GAP -) (VALENCY TRANS) (CAT V) (ROOT "clean") (PRE-MOD-ADV ((CAT ADV) (ROOT "periodically"))) (OBJ (COUNT *) (CAT N) (SEM *O-VENTILATION-SLOT) (NUMBER PL) (ROOT "slot") (DET ((CAT DET (ROOT "the"))))) (PP ((GAP -) (CAT P) (ROOT "with") (SEMSLOT INSTRUMENT) (OBJ ((COUNT *) (CAT N) (SEM *O-VACUUM-CLEANER) (ROOT "cleaner" ) (DET ((CAT DET) (ROOT "your")))))))) Interlingua representation(s) found: (*E-CLEAN (MOOD IMP) (EVENT-FREQUENCY &PERIODICALLY) (THEME (* O -vENTIiLATION-SLOT (NUMBER PL) (REFERENCE DEFINITE))) (INSTRUMENT (*O-VACUUM-CLEANER (PERSON SECOND)) (POSSESSIVE *))) 1 target f-structure(s) found ((TIME ((ROOT PRESENT))) (FORMAL *) (CAUSATIVE -) (PASSIVE -) (MOOD ((ROOT IMP))) (ROOT SOLUISURU) (CAT V) (SUBCAT TRANS) (VTYPE V-SAHEN) (SUBJ-CASE GA) (OBJ-CASE 0) (OBJ ((CASE O) (ROOT TUUKIKOU) (CAT H WH -))) (ADVADJUNCT ((ROOT TEIKITEKINI) (CAT ADV))) (PPADJUNCT ((ROOT SQLUIKI) (CAT N) (WH -) (PART OF) (COMPNOUN CN)))) 1 output string(s) found: Figure 9: Sample Translation to Japanese of One Selected Sentence, Showing Intermediate F-Structures and Interlingua Representation find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/mitamura-1991-efficient-interlingua-translation-system-multi/