LegoNet - classification and extractive summarization of Indian legal judgments with Capsule Networks and Sentence Embeddings

In this paper, we propose the LegoNet - a system to classify and summarize legal judgments using Sentence Embedding, Capsule Networks and Unsupervised Extractive Summarization. To train and test the system, we have created a mini-corpus of Indian legal judgments which have been annotated according to the classes: Facts, Arguments, Evidences and Judgments. The proposed framework uses Sentence Embedding and Capsule Networks to classify parts of legal judgments into the classes mentioned above. This is then used by the extractive summarizer to generate a concise and succinct summary of the document grouped according to the above mentioned classes. Such a system could be used to help enable the Legal Community by speeding up the processes involving reading and summarizing legal documents which a Law professional would undertake in preparing for a case. The performance of the Machine Learning Model in this architecture can improve over time as more annotated training data is added to the corpus.