Information Extraction From Court Room Records

Abstract

The legal domain remains among various areas that have many opportuni ties when it comes to improvement and innovation through computational advancements. In Pakistan, in the recent past, the courts have made reported judgments available to the public. As this data continues to grow at a rapid rate, it has become essential to process this massive chunk of data to better meet the requirements of the respective stakeholders. However, extracting the required information from this unstructured legal text is the main issue. Therefore, our goal is to have a machine learning system that can automat ically extract information out of these publicly available judgments of the Supreme Court. Once this information has been extracted, it can then be used by the lawyers, judges as well as civilians and also for policy making in Pakistan. For the purpose of our work, a total of thirteen entities are being extracted including dates, case-numbers, respondent names, reference cases, FIR no., person names, references etc. A labeled dataset is created using the publicly available legal judgments from the Supreme Court of Pakistan by using annotation guidelines. A pre-trained BERT model is then further trained and fine-tuned on the created dataset for Named Entity Recognition to extract the desired information. Our model also improved the results of the similar dataset available consisting of judgments from Lahore High Court which has smaller number of labels.