An Experiment in English—Spanish Automated Translation of Medical Language Data

Abstract

An English to Spanish translation procedure and its associated dictionaries were developed and implemented for the 1,426 terms of the morphology section of the International Classification of Diseases for Oncology. Morphological substitutions and respefling rules permit translation of most of the ICD—O vocabulary composed of Latin and Greek derived terms, which are cognate in the source and target languages, without the construction of a large word lexicon.A fairly simple classification of words which could be implemented by recognition of terminal morphemes, and which classifies them both syntactically and semantically served as an adequate basis for translation, and sheds some light on the linguistic structure of this type of complex noun phrase seen universally in medical writings and communications. A set of 17 reduction-transformation rules based on the word classification provide syntactic control of the translation process.

Dutch Respelling Rules for English and German Medical Word Lists

“…These lists of terms can be thought of as raw material for translator lexicons. Medical words and idioms in one language can be respelled III a cognate language, using methods introduced by investigators at the U. S. National Institutes of Health for English-to-French and English-to-Spanish [3,14]. This is possible because medical terminology in Western European languages has a tendency to form compounds and shares a common origin in Latin and Greek, properties which are exploited by all respelling algorithms [13,24].…”

“…Accordingly, it would be desirable to have methods for the rapidintroduction of a medical vocabulary into a computer translator as well as for updating this vocabulary. One potential source of new words and phrases is to respell comparable words which occur in another source language [3,9,14]. This might be particularly useful for new drug names or new technological fields which have been described primarily in another language.…”

Group Theory Approach to Computer Translation of Medical German

“…The sentence (initial pal'sandum) is designated by an upper case letter at the beginning of the alphabet, such as A = (A I ,A 2 , ... ,Ag)=(1,2, ... , 9). Each offset which may be applied to A (or other pal'sanda) is represented by a lower case letter in the middle of the alphabet, such as p. An offset is an nvector which may contain integers between -n and +n.…”

“…Potential idioms were obtained from a list of all word pairs or word triples occurring in the source document, listed in descending order of frequency. Words were accepted as final lexicon entries by a bilingual speaker who assigned a default translation and a syntactic-semantic class designator [9] to each entry, using the 21 classes listed below. Each word in the lexicon was also assigned any number of alternate translations which are dependent upon the classes of neighboring words.…”

Three Solutions to Problems of Categorized Medical Nomenclatures

“…Study has been made of linguistic rules for noun phrase generation which would explain these inconsistencies within existing categorized nomenclatures [7]. Linguists have given the problem attention in a more general setting, sometimes regarding the complex noun phrases and com- pound word forms as simple sentences, transformed by verb ellipsis or other processes [5,14,18,19,23].…”