Llm features

[pranay8297]

Massachusetts Legislative Bill Analysis

This repository contains code for summarizing and tagging Massachusetts Legislative bills using LLMs.

Overview

The main functionality is implemented in demo_app_functions.py, which provides methods for:

1. Summarizing bills
2. Categorizing bills
3. Tagging bills with relevant topics

Files

• demo_app_functions.py: Main file containing the core functionality
• prompts.py: Contains predefined prompt templates for various NLP tasks
• tagging.py: Includes category and tag definitions for bill classification

Key Features

• Bill summarization using GPT models
• Categorization of bills into predefined categories
• Tagging bills with relevant topics from a curated list
• Handling of both small and large documents using different approaches (standard and RAG)
• Integration with Massachusetts General Laws (MGL) for context-aware analysis

Requirements

• Python 3.x
• Dependencies listed in requirements.txt

Usage

To use the main functions:

from demo_app_functions import get_summary_api_function, get_tags_api_function

# For summarization
summary = get_summary_api_function(bill_id, bill_title, bill_text)

# For tagging
tags = get_tags_api_function(bill_id, bill_title, bill_text)

Note

This project is designed to work with Massachusetts legislative bills and may require specific data sources or API access for full functionality.

For more detailed information on each module and function, please refer to the individual file docstrings.

[vercel[bot]]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
maple-dev	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Oct 30, 2024 0:10am

[nesanders]

We agreed with @mvictor55 @pranay8297 @AnnaKSteele tonight that we want to implement the following configuration by default,

• Use this updated 10/29 version of tags
• Assign tags based on the summary, not full bill and section text -- this requires that getting summaries is run first, and then fed into the tagging function.
• Implement no minimum number of tags, but a maximum number of tags (max=6)
• No hierarchical categorization (assign tags from any category, not just a single category)
• Instruct the model to require a high threshold of relevancy

[pranay8297]

Instead of 2 API functions, create one API function to get summaries and tags

Internal Details Chronological order of things:

1. Generate Summary
2. Use the generated summaries and all the tags to prompt the model to choose tags with updated instructions
3. return tags and summaries