This talk will be preceded by a short annual general meeting of the section.
The field and practice of law hold a vast array of data amenable to processing by computational approaches to support practising lawyers in complex legal tasks such as classification, the assessment of potential legal outcomes and the selection of trial strategy. In litigation, for example, attorneys may seek to assess the relative merit of brining suit, weigh in various factors, such as legal expenses and penalties, and deal with uncertainties. Over the past few years, computer scientists, statisticians and lawyers have intensified research efforts in computational methods for what is broadly termed “decision support” by adopting methodology and techniques from fields such as AI. However, there are often gaps in the broader understanding of the actual decision environment, the needs and specific challenges within legal practice. Clarification on these topics is widely considered as an important preliminary to purposeful research in this field. This seminar brings together researchers and practitioners working at the forefront of developments in this area with the aim or sharing insights, perspectives and research results. The talks are followed by a panel discussion.
Programme
2:00-2:05 Introduction
Speakers: Anjali Mazumder and Alex Biedermann Moderator: Jane Hutton
2:05-2:55 Legal Decision Support in Use – What are Lawyers and their Clients Doing Today?
Speakers: Alexander Oddy and Andrew Dunkley (Herbert Smith Freehills LLP)
Format: 45min + 5min Q/A
2:55-3:35 Approaching Legal Reasoning with Deep Learning: A Common Law Perspective
Speakers: John Armour and Alina Petrova (University of Oxford)
Format: 35min + 5min Q/A 3:35-3:45 Break
3:45-4:25 Natural Language Processing on Legal Text
Speakers: Nikos Aletras
Format : 35min + 5min Q/A
4:25-4:55 Panel discussion
Speakers: All
Format: 30 min Q/A
4:55-5:00 Wrap-up
Speakers: Jane Hutton
– Alexander Oddy and Andrew Dunkley (Herbert Smith Freehills LLP)
– Professor John Armour, University of Oxford, Faculty of Law
– Dr. Nikos Aletras, Department of Computer Science, The University of Sheffield
Legal Decision Support in Use – What are Lawyers and their Clients Doing Today? -Alexander Oddy and Andrew Dunkley (Herbert Smith Freehills LLP)
Driving real world benefits from advances in decision support methodologies requires that they are translated into a context that lawyers and their clients can actually use. Alexander Oddy (Partner, Herbert Smith Freehills LLP) and Andrew Dunkley (Technology Solutions Lead, Herbert Smith Freehills LLP) will discuss the extent to which decision support is applied in today’s legal market, including the use of statistical and computational methods. They will also address some of the products and techniques that are available to practitioners and clients today, as well as offering thoughts on current and potential trends for the evolution and development of this space.
______________
Approaching Legal Reasoning with Deep Learning: A Common Law Perspective - Professor John Armour (Faculty of Law) and Dr Alina Petrova (Department of Computer Science), University of Oxford
Modern advances with the application of deep learning and NLP methods have made possible striking results regarding the prediction of outcomes in cases. In this talk, we will draw on aspects of a current research agenda encompassed by the UKRI-funded project, Unlocking the Potential of AI for English Law to consider four distinct topics (i) actual and potential use cases; (ii) data sources and limitations; (iii) a review of the technical state of the art; (iv) approaching the challenge of legal reasoning.
______________
Natural Language Processing on Legal Text - Dr. Nikos Aletras, Department of Computer Science, The University of Sheffield
Legal judgment prediction is the task of automatically predicting the outcome of a court case, given a text describing the case’s facts. In this talk, I will describe work on legal judgement prediction on the European Court of Human Rights using various neural network architectures in a variety of tasks: (1) binary violation classification; (2) multi-label classification; (3) case importance prediction. I will also talk about developing large pre-trained language models on legal text.