v4.0.0

This is the final release for the SWENG Industry Awards Showcase. It includes a reduced chaining approach to summarization using LangChain; the persisting of summaries and citations in an SQL database; additional pytests; a finalized UI with panes for search and filtering, summarization and citations; and a data ingestion pipeline for an NLP model hosted on Hugging Face. There is an updated README.md file with comprehensive instructions on how to run the application and additional information and promotional material.

v3.0.0

This release comprises a data ingestion pipeline to train an NLP model. The model is trained on tokenized data, focusing on legal citations, which is stored in an SQL database and fed to a Huggingface NLP model. On the frontend, this release fixes some bugs that were discovered relating to special characters, as well as new buttons and functionality. Additionally, the UI has been polished to create a clean and intuitive experience.

v2.0.0

The second release uses Docker for containerization and deployment. Instructions to run the application can be found in the README.md file. New for this release is a React frontend, which has helped create a polished look for the Microsoft Word add-on side panel. New pages for citations, summaries, and searches have been added or refined since the last release of the frontend. On the backend, an SQL database has been configured to persist legislative data to be fed to the NLP model. Still in development is the tokenization of the data in order to train the NLP model to recognize legal citations, along with a reduce chaining approach to summarization.

v1.0.0

This first release introduces an integrated frontend and backend application in the form of a Microsoft Word add-on, offering initial functionality for legislative drafting and analysis. The backend is equipped with API endpoints for accessing Legiscan's legal document API and incorporating NLP features through OpenAI's API, enabling text summarization and citation detection within legislative texts. The frontend features include a custom ribbon tab and taskpane with buttons designed to facilitate interactions with Legiscan and NLP functions, such as summarizing bills or selected sections.