Dramatically Reducing Training Data Size Through Vocabulary Saturation

William Lewis, Sauleh Eetemadi
2013 Conference on Machine Translation  
Our field has seen significant improvements in the quality of machine translation systems over the past several years. The single biggest factor in this improvement has been the accumulation of ever larger stores of data. However, we now find ourselves the victims of our own success, in that it has become increasingly difficult to train on such large sets of data, due to limitations in memory, processing power, and ultimately, speed (i.e., data to models takes an inordinate amount of time).
more » ... teams have dealt with this by focusing on data cleaning to arrive at smaller data sets (Denkowski et al., 2012a; Rarrick et al., 2011) , "domain adaptation" to arrive at data more suited to the task at hand (Moore and Lewis, 2010; Axelrod et al., 2011) , or by specifically focusing on data reduction by keeping only as much data as is needed for building models e.g., (Eck et al., 2005) . This paper focuses on techniques related to the latter efforts. We have developed a very simple n-gram counting method that reduces the size of data sets dramatically, as much as 90%, and is applicable independent of specific dev and test data. At the same time it reduces model sizes, improves training times, and, because it attempts to preserve contexts for all n-grams in a corpus, the cost in quality is minimal (as measured by BLEU ). Further, unlike other methods created specifically for data reduction that have similar effects on the data, our method scales to very large data, up to tens to hundreds of millions of parallel sentences.
dblp:conf/wmt/LewisE13 fatcat:oiopc5tslrelron4hydj2bfnjy