Toward an NLP Approach for Transforming Paper Contracts into Smart Contracts

Identifying and extracting information from contracts is an important task of contract analysis, which is mostly performed manually by lawyers and legal specialists. This manual analysis is a time-consuming, error-prone task. We can overcome this by automating the task of legal entity extraction using the Natural Language Processing (NLP) techniques. For extracting information from the natural language text, we can use popular NLP methods Named Entity Recognition (NER) and relation extraction (RE). Most NER and RE methods rely on machine learning and deep learning to identify relevant entities in natural language text. The main concern in adapting the AI methods for contract element extraction is the scarcity of annotated datasets in the legal field. Aiming at tackling this challenge, we decided to prepare the contract datasets for NER and RE tasks by manually annotating publicly available English contracts. This work is a part of the research aimed at automating the conversion of natural language contracts into Smart Contracts in the blockchain-based Supply Chain context. This paper explains the implementation and comparison of NER models using the deep learning methods BiLSTM and transformer-based BERT for evaluating the dataset.