AI Seminar: Structured (Syntactic) Prediction without Structured Representations
David Vilares Calvo, Research Associate at the University of A Coruña.
The application of neural architectures to syntactic parsing to obtain context-aware feature representations has made it possible to parse natural language into phrase structure or dependency grammars with conceptually simpler models than before. I will first talk about such 'deep' conversion of natural language parsers to then discuss the work that we have been doing at the FASTPARSE ERC Starting Grant project to achieve simpler and faster models; by casting different parsing paradigms as sequence labeling. This strategy allows to establish a one-to-one mapping from words to their full syntactic structure, and it can be used together with any generic sequence tagging framework used in computer science, in order to create robust parsers.
David is a Research Associate at the University of A Coruña (Spain), where he develops techniques and algorithms to improve the speed and simplicity of natural language parsers. Before that, he did his Ph.D. on compositional language processing for multilingual sentiment analysis, and his thesis was awarded one of the six National Computer Science Awards by the BBVA Foundation and The Spanish Scientific Society for Computer Science. David is also interested in research topics that connect natural language processing with social media analytics to extract useful information from unstructured data.
The seminar is free and open for everyone.
This seminar is a part of the AI Seminar Series organised by SCIENCE AI Centre. The series highlights advances and challenges in research within Machine Learning, Data Science, and AI. Like the AI Centre itself, the seminar series has a broad scope, covering both new methodological contributions, ground-breaking applications, and impacts on society.