Starting in February 2020 I have a VENI, a grant from the Dutch government, for research on neural minimalist parsing. This means I'm working to build fully neural parsers for Minimalist Grammars, a formalisation of modern mainstream syntax. Parses from these parsers can then be used as an intermediate structure for neural semantic parsers.
Since Minimalism is implicitly semantically motivated, as well as syntactically, we expect minimalist parses to be particularly good inputs to semantic parsers. Moreover, since MGs are Mildly Context Sensitive, they have the right amount of power to handle the relationship between syntax and semantics in human language.
This work is part of the research programme VENI with project number VI.Veni.194.057, which is (partly) financed by the Dutch Research Council (NWO).
Learning meaning from structure
Dr. M. (Meaghan) Fowlie, UU
Personal digital assistants like Siri highlight the possibilities and challenges of building a computer that interacts with people using natural language. This research will create computer algorithms that automatically find sentence meanings by combining state-of-the-art ”deep learning” techniques from computer science with linguistic work on structure and meaning of sentences.
Keywords: syntax, parsing, semantics, neural networks, deep learning, reinforcement learning, minimalist grammars, minimalism