NLP - Segment Relevant Legislative Content or Updates

Context

Information about the Org: HaAguda Le’maan Ha’Lahatab Be Israel - known commonly as the Aguda, is an Israeli non-profit LGBTQ rights organization; They provide a range of professional services such as mental support, legal counseling, family mediation, assistance to refugees and immigrants, and assistance to any victim of discrimination or hatred on the basis of sexual orientation or gender identity. The Aguda focuses on three main sectors:

Social and legislative impact

Individual’s rights and services

LGBTQ Localization

As part of the first sector, the state authorities still discriminate against the LGBTQ community in legislation, policies and resources and many basic rights are still being denied. As every year, there is a significant increase in the number of reports of discrimination, hate crimes and inequality towards all the populations that make up the LGBTQ community. The report reveals the fact that LGBTQ-phobia has not bypassed any circle and forced us to deal with its consequences in every field: the workplace, social networks, government institutions, the public and private spheres, and more. This sector focuses on the work plan of the LGBTQ community’s rights upfront the Knesset and the government monitoring discussions and updates related to LGBTQ-phobia, Family life, relationship and parenting, Welfare, health and employment benefits.

What is the problem and its effects?

Today, the team responsible for this sector needs to do a lot of manual work of visiting different governmental resources in which discussions and updates are documented and manually filter which updates are relevant and potentially impactful to the LGBTQ community. Additionally they need to manually prioritize the pain score of each update the decide which action plan to take regarding each threat. Most of the time is spent on manual filtering and research and the hope is to use the Aguda’a power by increasing active motion and decreasing manual, time consuming segmenting process;

Current state

There is a weekly Agenda in which the team needs to access each day the resources below and locate relevant content before planning action plans regarding the threatening updates.

https://www.gov.il/he/Departments/policies?OfficeId=e744bba9-d17e-429f-abc3-50f7a8a55667&blockCollector=true&limit=10&PmoMinistersComittee=a0d9709c-f07d-4b0e-8c48-0939643eb020&skip=0

https://main.knesset.gov.il/about/departments/pages/sg/sgpresidium.aspx

https://main.knesset.gov.il/activity/committees/pages/allcommitteesagenda.aspx?tab=1

https://www.gov.il/he/Departments/publications/?OfficeId=e744bba9-d17e-429f-abc3-50f7a8a55667&blockCollector=true&limit=10&publicationType=06311039-a4dc-4457-af46-a8e7dbfbe5a0&skip=0

After the relevant info is found the process of creating action plans is (TBD);

Assumptions

There exists a pre-trained NLP model that can accurately classify updates as relevant or irrelevant to the LGBTQ community.

The relevant updates can be extracted and summarized automatically from the resources mentioned in the current state section.

Constraints

The solution must be developed using existing resources and technologies.

The solution must be compatible with the existing workflow and tools used by the team responsible for this sector.

The solution must ensure the privacy and security of the information being processed.

OUT OF SCOPE

Training a new NLP model for this task.

Developing a new workflow or tool for the team.

Proposed Solution

The proposed solution is to use an existing NLP model to segment relevant legislative updates automatically. The team responsible for this sector can continue to access the same resources mentioned in the current state section, and the updates can be automatically classified as relevant or irrelevant using the pre-trained model.

For the relevant updates, an extractive summarization technique can be used to create a summary of the update, including key points and potential impacts on the LGBTQ community. This summary can be displayed in a dashboard or email to the team, along with a pain score assigned by the model to prioritize the most urgent updates.