This document discusses quality estimation of machine translation using the QuEst++ framework. It summarizes that QuEst++ can predict the quality of unseen machine translated text using only the source and target texts without references, extracting features to build models that estimate metrics like post-editing effort and time from limited labeled training data. The framework extracts features at the word, sentence and document level from the source and target texts and information from the machine translation system, then trains models using those features to predict quality scores for new translations.