Party-based Sentiment Analysis Pipeline for the Legal Domain 
2021 21st International Conference on Advances in ICT for Emerging Regions (ICter)
Since the advent of research to automate existing manual language workflows, legal domain has risen as a key area. When it comes to making the process of court cases more efficient with the said automation, legal information extraction is vital for legal professionals. In this study, we propose a unified pipeline to annotate party-based sentiment for court cases which reduces manual work. To achieve this end, we combined two state-of-the-art models into a single workflow. The first model extracts the entities which represent each party in a court case, while the second model analyzes the sentiment with respect to each party in a given sentence of a court case. In this study we propose two approaches for defining the pipeline which maps the output of the party extraction system into the party-based sentiment analysis system. Further, we propose improvements to the existing party extraction system.
Keywords: Natural Language Processing | Machine Learning / Deep Learning | Law | Aspect-based Sentiment Analysis | Legal Party Identification | Legal Domain | Information Extraction |