Despite the success of multilingual sequence-to-sequence pre-training, most existing approaches rely on document-level monolingual corpora in many different languages, sentence-level bilingual corpora, and sometimes synthetic document-level bilingual corpora. This hampers the performance with cross-lingual document-level tasks such as document-level translation. Hence, we propose to mine and leverage document-level trilingual parallel corpora to improve sequence-to-sequence multilingual pre-training. We present Triangular Document-level Pre-training (TRIP) as the first in the field to accelerate the conventional monolingual and bilingual objectives into a trilingual objective with a novel method called Grafting. Experiments show that TRIP achieves several strong state-of-the-art (SOTA) scores on three multilingual document-level machine translation benchmarks and one cross-lingual abstractive summarization benchmark, including consistent improvements by up to 3.11 d-BLEU points and 8.9 ROUGE-L points.

This paper describes the submissions of the “Marian” team to the WNMT 2018 shared task. We investigate combinations of teacher-student training, low-precision matrix products, auto-tuning and other methods to optimize the Transformer model on GPU and CPU. By further integrating these methods with the new averaging attention networks, a recently introduced faster Transformer variant, we create a number of high-quality, high-performance models on the GPU and CPU, dominating the Pareto frontier for this shared task.

This paper describes the systems used for the MSR+FBK submission for the SLT track of IWSLT 2013. Starting from a baseline system we made a series of iterative and additive improvements, including a novel method for processing bilingual data used to train MT systems for use on ASR output. Our primary submission is a system combination of five individual systems, combining the output of multiple ASR engines with multiple MT techniques. There are two contrastive submissions to help place the combined system in context. We describe the systems used and present results on the test sets.

This poster is a preliminary report of our experiments for detecting semantically shifted terms between different domains for the purposes of new concept extraction. A given term in one domain may represent a different concept in another domain. In our approach, we quantify the degree of similarity of words between different domains by measuring the degree of overlap in their domain-specific semantic spaces. The domain-specific semantic spaces are defined by extracting families of syntactically similar words, i.e. words that occur in the same syntactic context. Our method does not rely on any external resources other than a syntactic parser. Yet it has the potential to extract semantically shifted terms between two different domains automatically while paying close attention to contextual information. The organization of the poster is as follows: Section 1 provides our motivation. Section 2 provides an overview of our NLP technology and explains how we extract syntactically similar words. Section 3 describes the design of our experiments and our method. Section 4 provides our observations and preliminary results. Section 5 presents some work to be done in the future and concluding remarks.