KantanMT Launches New Training Data Analysis Feature
KantanMT has recently launched a new training data analysis feature which automatically reports on the suitability of training data during the customisation of KantanMT engines.
KantanMT maintains exacting quality standards during the selection of training data - data is often rejected if these standards are not met. This new analysis feature will help members understand the reason why data is rejected from KantanMT engine customisations.
KantanMT members who customise engines with higher quality training data will notice significant increases in quality scores, such as BLEU, TER and F-Measure. In addition, KantanAnalytics™ - KantanMT’s new Quality Estimation technology – will report higher scores which will lead to reduction in post-editing effort.
What is KantanMT™?
KantanMT is a cloud-based implementation of Moses Statistical Machine Translation technology. Leveraging the power and flexibility of the cloud, KantanMT effortlessly scales to generate a high-quality, low-cost Machine Translation platform for small-to-medium sized Localization Service Providers. According to the research firm, Markets and Markets, the natural language processing sector will grow from €2.8 billion in 2013 to €7.4 billion in 2018, representing a compound annual growth rate of 21.1 percent.