Page 447 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 11: Virtual Legal Assistant of the Rights Mapping and

Analysis Platform (RiMAP) to Improve Legal Analysis and Access
to Rights and Solutions for Forcibly Displaced and Stateless 4.4-Productivity

Persons

Organization: United Nations High Commissioner for Refugees (UNHCR)

Country: Switzerland (scope is multi countries)

Contact Person(s):

Esther Benizri: benizri@ unhcr .org ; Moises Maldonado: maldonam@ unhcr .org.

1 Use Case Summary Table

Item Details
Category Intelligence system / Humanitarian

Problem Addressed Difficulty collecting, analyzing, and maintaining up-to-date legal
information affecting forcibly displaced and stateless persons across
numerous countries, hindering effective legal support and advocacy.
Key aspects of Tech- 1. Automated legal data collection & updating from diverse sources.
nology 2. AI-powered analysis (summarization, extraction). Solution
3. Human-in-the-loop review for quality assurance.
4. Internal VLA interface & future external chatbot.
Data Keywords Large Language Models (LLMs), Natural Language Processing (NLP),
Technology, AI Chatbot, Web Scraping/Crawling, Data Extraction,
Summarization, Text Generation

Data Availability First private, later public (https:// rimap .unhcr .org/ ), composed of
domestic laws and policies regulating access to rights for forcibly
displaced and stateless persons.
Metadata Legal texts (laws, regulations, policies), case law documents, (Type
narrative legal and factual reports, government publications, and of
Data)
UNHCR guidance documents.

Model Training and Utilizing localized pre-trained LLMs fine-tuned on relevant legal corpora-
Fine Tuning tion and UNHCR-specific documents for summarization, question and
answering, chat with documents, and data extraction.
Deployments Internal Prototype evaluated by UNHCR staff.

Code To be shared publicly on the UN AI Toolbox repositories.

411