Ana Marasovic Title: SRL4ORL: Semantic Role Labelling for Opinion Role Labelling Abstract: Fine-grained opinion analysis involves several subtasks: it aims to detect opinion expressions in a text (e.g. “hate”), measure their intensity (e.g. strong), classify their sentiment (e.g. negative), and identify their targets (entities or propositions at which the sentiment is directed) and holders (entities that express an opinion). We propose a neural pipeline sequence tagging model for labelling of opinion entities, i.e. opinion expressions and their holders and targets. To combat the problem of scarcity of labelled data, we exploit transfer learning from a related NLP task which has substantially more data, that is, semantic role labelling (SRL).