Deep Learning for Document Summarisation and Question Answering

Computers are excellent at working with crisp or tabular information, but the ambiguity that exists in human language makes it much harder for them to understand textual data and to interact in natural language.

We exploit deep learning, a framework in machine learning that is now receiving significant attention in research and industry, to improve the quality of natural language processing applications such as document summarisation and question answering. Our main goal is to equip neural networks with more complex mechanisms to estimate their parameters and to decode ("do inference") with them.