Argumentation Mining in the field of Law
Case law plays an important role in legal argumentation and decision-making, especially in countries with legal systems using common law [1]. Lawyers need to dig into documents of past judicial decisions (precedents) and identify previous decisions supporting their side in the legal dispute and undermining the other. The analysis of these documents requires skill and a lot of time: Cases are expressed in natural language, consider highly complex matters under dispute, and have complex inter-relationships while the number of documents is constantly growing. Moreover, navigating the documents, interpreting, and applying the results successfully needs extensive training. These factors can furthermore lead to extensive costs of legal proceedings because of additional working hours and high fees of lawyers (skill premium).
This is where the LAWrgMiner comes into play. LAWrgMiner, a neural network trained on the ECHR dataset [2], aims to extract natural language arguments from legal documents. It allows lawyers to extract arguments from an unstructured legal proceeding and presents them in an easily understandable and structured way. While they normally must invest hours into manually searching through and annotating case law, LAWrgMiner reduces the time invested from searching for multiple hours to presenting claims and premises within the uploaded source in a couple of minutes.
- Simple upload
- Extraction of claims and premises with a built-in keyword search
- Supporting legal professionals in their daily activities
- Can save a lot of working hours and, therefore, costs of legal proceedings
[1]: Milward, D., Mochales, R., Moens, M. & Wyner, A. (2010). Approaches to Text Mining Arguments from Legal Cases. Semantic Processing of Legal Texts.
[2]: A legal corpus made for the purpose of argument mining. See: Poudyal, P., Savelka, J., Ieven, A., Moens, M., Gonçalves, T., & Quaresma, P. (2020).
ECHR: Legal Corpus for Argument Mining. Proceedings of the 7th Workshop on Argument Mining, 67–75.