System Combination for Machine Translation of Spoken and Written Language

Abstract

Abstract-This paper describes an approach for computing a consensus translation from the outputs of multiple machine translation (MT) systems. The consensus translation is computed by weighted majority voting on a confusion network, similarly to the well-established ROVER approach of Fiscus for combining speech recognition hypotheses. To create the confusion network, pairwise word alignments of the original MT hypotheses are learned using an enhanced statistical alignment algorithm that explicitly models word reordering. The context of a whole corpus of automatic translations rather than a single sentence is taken into account in order to achieve high alignment quality. The confusion network is rescored with a special language model, and the consensus translation is extracted as the best path. The proposed system combination approach was evaluated in the framework of the TC-STAR speech translation project. Up to six state-of-the-art statistical phrase-based translation systems from different project partners were combined in the experiments. Significant improvements in translation quality from Spanish to English and from English to Spanish in comparison with the best of the individual MT systems were achieved under official evaluation conditions.

Automatic Translation Error Analysis

“…Meteor-Xray uses the alignment algorithm underlying the Meteor metric but the aligning component could be shared with other MT applications, e.g. system combination [11], where fully unsupervised GIZA++ has been successfully used [12].…”

The RWTH system combination system for WMT 2009

“…We follow the approach of Bangalore et al (2001) with some extensions. Multiple insertions with regard to the primary hypothesis are sub-aligned to each other, as described by Matusov et al (2008). Figure 2 gives an example for the alignment.…”

“…With our approach, we could also extract N -best hypotheses. In a subsequent step, these Nbest lists could be rescored with additional statistical models (Matusov et al, 2008). But as we did not have the resources in the WMT 2009 evaluation, this step was dropped for our submission.…”

Persian–Spanish Low-Resource Statistical Machine Translation Through English as Pivot Language

“…However, there are sufficient resources between them and some other languages. Though it is claimed that, the intermediary languages do not lead to an improvement in general case, this idea can be employed as a simple method to enrich the translation performance even for existing systems (Matusov et al, 2008).…”