System and method for productive generation of compound words in statistical machine translation
First Claim
1. A method for making merging decisions for a translation comprising:
- providing a translated text string in a target language of a source text string in a source language;
with a merging system, wherein the merging system is implemented with a computer processor, outputting decisions on merging of pairs of words in the translated text string, the merging system comprising at least one of;
a set of stored heuristics comprising at least a first heuristic by which two consecutive words in the string are considered for merging if an observed frequency f1 of the two consecutive words as a closed compound word is larger than an observed frequency f2 of the two consecutive words as a bigram, anda merging model trained on features associated with pairs of consecutive tokens of text strings in a training set and predetermined merging decisions for the pairs to predict merging decisions for a new translated text string; and
outputting a translation in the target language based on the merging decisions for the translated text string.
1 Assignment
0 Petitions
Accused Products
Abstract
A method and a system for making merging decisions for a translation are disclosed which are suited to use where the target language is a productive compounding one. The method includes outputting decisions on merging of pairs of words in a translated text string with a merging system. The merging system can include a set of stored heuristics and/or a merging model. In the case of heuristics, these can include a heuristic by which two consecutive words in the string are considered for merging if the first word of the two consecutive words is recognized as a compound modifier and their observed frequency f1 as a closed compound word is larger than an observed frequency f2 of the two consecutive words as a bigram. In the case of a merging model, it can be one that is trained on features associated with pairs of consecutive tokens of text strings in a training set and predetermined merging decisions for the pairs. A translation in the target language is output, based on the merging decisions for the translated text string.
35 Citations
22 Claims
-
1. A method for making merging decisions for a translation comprising:
-
providing a translated text string in a target language of a source text string in a source language; with a merging system, wherein the merging system is implemented with a computer processor, outputting decisions on merging of pairs of words in the translated text string, the merging system comprising at least one of; a set of stored heuristics comprising at least a first heuristic by which two consecutive words in the string are considered for merging if an observed frequency f1 of the two consecutive words as a closed compound word is larger than an observed frequency f2 of the two consecutive words as a bigram, and a merging model trained on features associated with pairs of consecutive tokens of text strings in a training set and predetermined merging decisions for the pairs to predict merging decisions for a new translated text string; and outputting a translation in the target language based on the merging decisions for the translated text string. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18)
-
-
19. A translation system comprising:
-
a decoder which receives as input a source text string in a source language and outputs a translated text string in a target language in a target language, based on the source text string; a merging system, which receives the translated text string and outputs a translation in the target language based on translated text string, the merging system being configured for outputting decisions on merging of pairs of words in the translated text string, the merging system comprising at least one of; a set of stored heuristics comprising at least a first heuristic by which two consecutive words in the string are considered for merging if the observed frequency f1 of the two consecutive words as a closed compound word is larger than an observed frequency f2 of the two consecutive words as a bigram, and, optionally, if the first word of the two consecutive words is also recognized as a compound modifier, and a merging model trained on features associated with pairs of consecutive tokens of text strings in a training corpus and predetermined merging decisions for the pairs to predict merging decisions for a new translated text string. - View Dependent Claims (20, 21, 22)
-
Specification