Question Answering for the Legal Domain

September 2019 - May 2020

For my Third Year Project at The University of Manchester, I developed a Question Answering system for querying legal documents in natural language. The goal of this project was to develop a system which could be utilised by legal professionals during legal discovery, to quickly find the pertinent information in case files.

The system works by first having the user pose a query, then filtering out which paragraphs in the documents are likely to be irrelevant to the query, based on soft cosine similarity between the query and the paragraph. Finally, the paragraphs which are similar enough to the query to be deemed relevant, are passed to the machine comprehension code which utilises Bi-Directional Attention Flow (BiDAF) and Embeddings from Language Models (ELMo) to determine which span(s) of text in the document(s) best answer the user's query.
View Code

Languages & Technologies Used: Python, AllenNLP, Gensim, NLTK

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Travelly

October 2019

Travelly is a project I worked on for Facebook's Hack-a-Project event in Manchester. It is meant to be a web application for finding or forming travel groups so that users can find like-minded persons to travel with and visit new countries. I worked mainly on the backend implementation of this project. In the end, my team earned second place following a tie-break decision by the judges.
View Code

Languages & Technologies Used: Java, JavaScript, HTML, CSS, Thymeleaf, Spring, Git

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Eventlite

February 2019 - May 2019

Eventlite is a website built as part of the Software Engineering 2 course unit at the University of Manchester. Eventlite is a web application for displaying events and their information and includes a differentiation between event attendee views and functionality, and event organizer views and functionality. The goal of this piece of coursework was to use the ‘Model View Controller’ (MVC) method of building web applications. In carrying out the project we acted as a team of software developers who were tasked weekly with implementing new feature specifications proposed by the would-be clients.

Languages & Technologies Used: Java, JavaScript, HTML, CSS, Thymeleaf, Spring, Eclipse, Git, Mapbox API, Twitter API

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

ChatApp!

April 2019

‘ChatApp!’ is a messaging application developed for Android devices, which was built in order to explore and experiment with data transmissions, by sending text and multimedia messages through a proxy server to a bot. The development of this application involved the implementation of bit error correction and packet replacement, to combat issues which can occur in mobile systems and data transmissions. This project also included working with sound and image processing in order to transmit the multimedia messages and correct discrepancies in the transmitted data.

Languages & Technologies Used: Java, Android Studio

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Marketshare

November 2018

Marketshare is a hackathon project built at the Durhack 2018 hackathon. It is a web application built with the intention of allowing beginner investors to share ideas with friends, track the performance of their stock portfolios and follow people who they view as mentors. For this project, the team of 4 divided into two sub teams; one working exclusively on backend functionality and the other working on the frontend design. I worked in the backend sub team and was responsible for writing code which got the stock data, ran calculations on it and presented the user with their portfolio performance statistics.
View Code

Languages & Technologies Used: JavaScript, Node.js, HTML, CSS, Alpha Vantage API, Git

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

OnWord

October 2017 - April 2018

OnWord is a web application built to allow users to create crossword puzzles and solve them in a competitive and collaborative environment. The target market for this website was senior citizens, as we hoped it would help in including them in the digital age, as technology advances. We aimed to provide them with an online community in order to combat the isolation of senior citizens. In consideration of this, the website was built with a simple user interface, utilising large icons and text. Furthermore, the colour decisions were made in consideration of making the website accessible to persons afflicted with colour blindness. I was primarily responsible for developing the user profile pages and calculating user statistics, while also assisting wherever necessary.
View Code

Languages & Technologies Used: HTML, CSS, JavaScript, PHP, SQL, Git