Aliaksei Ivanou

Learn More
We present a technique to jointly learn the high level abstractions of sequential features (such as, pitch and MFCC's) and combine time-aggregated features (such as, mean length of pauses, recognizer confidence scores, etc.) to optimize the automated scoring of non-native spoken responses. We use a bidirectional long short term memory (BLSTM) network, a(More)
Recent advances in immersive computing technology have the potential to accelerate development of engaging intelligent agents that can guide one or multiple phases of learner instruction, learning, and assessment (both formative and summative). Such technologies are also important since they offer opportunities for personalizing the learning environment to(More)
  • 1