Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Inverting Noun-de-Noun Constructions in French-to-English Translation

IP.com Disclosure Number: IPCOM000113852D
Original Publication Date: 1994-Oct-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 103K

Publishing Venue

IBM

Related People

Brown, PF: AUTHOR [+5]

Abstract

A common problem in automatically translating French text into English is encountered when translating French phrases of the form f sub 1 "de" f sub 2 , where f sub 1 and f sub 2 are French nouns. Suppose that e sub 1 and e sub 2 are translations of f sub 1 and f sub 2 respectively. Sometimes these French phrases are translated into English as e sub 1 "of" e sub 2 , and sometimes they are translated as e sub 2 e sub 1 . Examples of the former case include: o "somme d'argent" -> "sum of money" o "ann\'ees de service" -> "years of service" o "pays d'orgin" -> "country of origin" o "question de privl\'ege" -> "question of privledge" o "conflit d'int\'er\^et" -> "conflict of interest"

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 50% of the total text.

Inverting Noun-de-Noun Constructions in French-to-English Translation

      A common problem in automatically translating French text into
English is encountered when translating French phrases of the form  f
sub 1  "de"   f sub 2 , where  f sub 1  and  f sub 2  are French
nouns.  Suppose that  e sub 1  and  e sub 2  are translations of
 f sub 1  and  f sub 2  respectively.  Sometimes these French phrases
are translated into English as  e sub 1  "of"  e sub 2 , and
sometimes they are translated as  e sub 2   e sub 1 .  Examples of
the former case include:
  o  "somme d'argent" -> "sum of money"
  o  "ann\'ees de service" -> "years of service"
  o  "pays d'orgin" -> "country of origin"
  o  "question de privl\'ege" -> "question of privledge"
  o  "conflit d'int\'er\^et" -> "conflict of interest"

Examples of the latter case include:
  o  "bureau de poste" -> "post office"
  o  "compagnie d'assurance" -> "insurance company"
  o  "gardien de prison" -> "prison guard"
  o  "communique\'e de presse" -> "press release "
  o  "taux d'int\'er\^et" -> "interest rate"

      The difficulty in translating such a phrase is that there is no
simple rule that can decide whether it should translate as
 e sub 1  "of"  e sub 2  or as  e sub 2   e sub 1 .  There are many
thousands, if not millions of such phrases and it would be a
Herculean task to attempt to decide by hand which way each such
phrase should translate.

      To some extent, how a particular phase translates is determined
by the first noun in the phrase; to some extent by the second noun;
and often both nouns matter.  To determine to what extent the answer
is determined by each of these factors, one must look at many many
examples of such noun-de-noun phrases and their translations.  To do
a good job, the number of examples required would be so large as to
make the number of man hours required a serious obstacle.

      The invention described herein consists of a procedure for
constructing a probabilistic model that can decide with high accuracy
which way to translate French noun-de-noun phrases.  This procedure
constructs such a model automatically from a large corpus of
bilingual text.

      It is known to describe a method of automatically aligning
words in one language with their translations in another language To
construct a probabilistic model that can be used for noun-de-noun
translation, it is first necessary to construct a database of
noun-de-noun examples.  Such a database is constructed from a large
French-Engish blingual corpus as follows:
  1.  Align the French sentences in the corpus with their English
      translations.
  2.  Align the words in the French sentences with the words in
      the English sentences which are their English translations.
  3.  Assign parts of speech to the words in all French sentences.
  4.  Examine each French-English sentence pair to find all
     ...