What is Kantan BuildAnalytics?
Kantan BuildAnalytics shows you how to rapidly improve an engine after its initial training with minimum cost and effort. Kantan BuildAnalytics provides various tools such as Fluency Analysis, Recall and Precision Analysis, Post-Editing Estimation, Neural Network Perplexity Analysis, Gap Analysis, Training Reject Reports and Timeline.
All these functionalities of Kantan BuildAnalytics provide a detailed report of the performance of your KantanMT engine. Hence, providing insights into the training and development of KantanMT engines along with useful information that helps improve KantanMT engine performance and quality. BuildAnalytics makes it easier for KantanMT customers to improve their engines with minor cost and work.
How to use KantanBuildAnalytics
You will be directed to KantanMT Dashboard. You will see a list of all your engines and on the right hand side you will find the following orange database icon that directs you to Kantan BuildAnalytics.
To download the summary report click on the ‘Download summary report’ button on right hand side.
To view the amount of words in the training data that were rejected click on the ‘Rejects’ tab. You will now be directed to the ‘Rejects’ page.
The F-Measure, BLEU and TER tabs give greater insight into the engine's scores and provides the individual score for each segment in the test set.
Click here for a detailed description of F-Measure in BuildAnalytics
Click here for a detailed description of BLEU
Click here for a detailed description of TER
GAP Analytics is a useful tool for adding terminology to your engine. It provides a list of words that the engine does not have a translation for. This list can be downloaded, translated and added to the engine as terminology.
Timeline is a complete history of your engine. It keeps a record of every time files are uploaded, translated, deleted and every time an engine is changed or built.
The following video explains some of the features of Kantan BuildAnalytics.