Questions for Self-monitoring: Machine Translation (2)

• How can backtranslations help to alleviate data sparsity?
• Is there an alternative to the use of backtranslations?
• How do multilingual models work? How does this help to deal with data sparsity?

• Which factors of data quality have to be considered in corpus preparation?

• How can the quality of MT output be assessed by human judges?
• What are the drawbacks of human evaluation?
• How can the quality of MT output be measured automatically?
• What are the limitations of automatic evaluation measures?