A data-driven framework for understanding human route instructions
I will discuss the performance of a data-driven approach for understanding human route instructions in three different setups. In the first setup, I evaluate the performance of the framework on understanding the transcriptions of human route instructions (in the IBL corpora). In the next setup, I evaluate the performance of this approach on the audio recording of these route instructions (using Recorded-Utterance Recognition). In the third setup, I evaluate the performance of the approach for incremental processing of the recorded route instructions.