Deep learning for semantic parsing

Abstract

This is the memory of an exploratory research project on techniques for reasoning on text with Deep Learning (DL). To study reasoning we focus on the problem of Natural Language Question-Understanding (NLQU), and in particular in the task of Semantic Parsing, a challenging Natural Language Processing (NLP) task that requires NLQU and even puts todays Deep Learning machinery to the test.

More specifically we provide a discussion about semantic parsing, and in concrete, deep learning techniques for semantic parsing. In our study of semantic parsing, we focus on two central topics: annotation and (deep learning) systems. At a more practical level, we run experiments of a state-of-the-art semantic parsing system a new and innovative semantic parsing dataset called OTTA \cite{OTTA}. Finally, we take the opportunity to learn the details of the system implementation, and we refactor the system to make it suitable (in terms of speed and integration) for future work.

Language: English