Deep Learning (DL) is a kind of Machine Learning (ML) that uses multiple layers of algorithms to tackle complex problems stepwise. DL has been applied successfully to many tasks in natural language processing (NLP). This class introduces basic concepts and linguistic applications of DL in the field of semantics (no previous knowledge of Machine Learning or programming skills is required). We envisage a syllabus with three main parts:
1. Structured knowledge in deep learning. Here we will look at neural networks and logic rules for semantic compositionality, learning semantic similarity and encoding distances as knowledge graphs, ontology-based text classification, and multilingual resources for neural representations of semantic role labelling.
2. Learning knowledge representations, including deep learning methods for knowledge-based completion, deep learning models for learning knowledge representations from text, and deep learning ontological annotations.
3. Applications like information retrieval and extraction with knowledge graphs and deep learning models, knowledge-based deep word sense disambiguation and entity linking, and investigation of compatibilities and incompatibilities between deep learning and Semantic Web. |
