Artificial Intelligence for Legal Assistance
(AILA 2020)
Artificial Intelligence for Legal Assistance (AILA) is a series of shared tasks aimed at developing datasets and methods for solving variety of legal informatics problems. The first edition of AILA was held in conjunction with FIRE 2019 and focused on precedent and statute retrievals for a given legal scenario. This year AILA will consist of two different tasks. While we retain last year's precedent and statute retrieval task, we add a new task for semantic segmentation of legal documents.
Task 1: Precedent & Statute retrieval
Given a query (description of a situation), identify relevant statutes and prior-cases
Task 2 : Rhetorical Role Labelling for Legal Judgements
The task is to semantically segment a legal case document. More formally, it is a sentence classification task, where each sentence has to be assigned one of the 7 predefined labels or "rhetorical roles".
A more detailed description of the tasks can be found on the description pages of task 1 and task 2. Participants can submit runs for either or both tasks.
Related literature
More details about the datasets used in both the tasks can be found in the works by Bhattacharya et al. highlighted below.
Task 1
Bhattacharya, et al. "Overview of the FIRE 2019 AILA Track: Artificial Intelligence for Legal Assistance." FIRE 2019
Wang et.al., "Hierarchical Matching Network for Crime Classification", SIGIR 2019
Wang et.al., "Modeling Dynamic Pairwise Attention for Crime Classification over Legal Articles", SIGIR 2018
Task 2
Bhattacharya, et al. "Identification of Rhetorical Roles of Sentences in Indian Legal Judgments" JURIX 2019
Savelka et.al. "Segmenting U.S. court decisions into functional and issue specific parts" JURIX 2018
Nejadghoii et.al., "A semi-supervised training method for semantic search of legal facts in Canadian immigration cases" JURIX 2017
NOTE:
Due to the ongoing covid-19 crisis AILA 2020 and FIRE 2020 will be completely virtual events
A participant team may participate in either or both the tasks and further in either or both subtasks within task 1.
Each team can have at most 4 participants
A team can submit up to 3 different runs for each task
While sending their runs, each team must also submit a detailed description of their algorithm(s). The format in which runs are to be submitted will be announced later
Participants are NOT allowed to use any external resources. For instance, it is not allowed to use Web search or existing legal search systems to identify relevant cases / statutes
Participants are allowed to use embeddings pre-trained on publicly available data subject to certain constraints. For task 1a & 1b, pre-trained embeddings are allowed ONLY if it is NOT trained on judgements of Indian supreme court. For task2 there are no such restrictions.