View on GitHub

Leonora-Schreurs-Portfolio

Leonora Schreurs’ Portfolio

LATEST PROJECT: CHATBOT IN TF2 WITH TRANSFORMER ARCHITECTURE

Movie Chatbot

Chatbot

This project implements a chatbot using a Transformer architecture with attention that can generate natural sounding responses that simulate a conversation between humans. The Transformer chatbot is built from scratch and implemented with Tensorflow 2. The model is trained on the famous Cornell Movie Dialogues Corpus. Evaluation makes use of greedy search- and top-k sampling during text generation to evaluate the model.



Classification with BERT

NLP Classifier with BERT-tokeniser

This NLP (natural language processing) project builds on part 1, It’s out there ! to improve the Stock Market Sentiment text classifier by using Google’s powerful open sourced BERT algorithm. The pre-training as well as the bidirectionality of BERT will hopefully improve results, while we keep the constraint of the limited number (+/- 6,000) of tweets in the dataset.



Unemployment in South Africa

Exploratory Data Analysis & Data Vizualisation

Unemployment in South Africa has reached crisis levels and current projections show no signs of improvement. The South African Government cites 6 reasons for the high unemployment. Analysing data from the World Bank Group databank, this notebook sets out to determines if there is data to support these theories.



Know Thy Customer

Customer Segmentation using Clustering Methods

This project explores how machine learning, and specifically clustering algorithms can assist the marketing department in a bank to perform customer segmentation on their credit card clients. The project looks at three different methods to achieve this.



It’s Out There !

Sentiment Analysis using Natural Language Processing

The GameStop incident early in 2021 took the importance of stock market sentiment and the role it plays in determining the price of a stock to a next level. This project analyses tweets about stock prices and attempts to predict the sentiment of these tweets by employing natural language processing (NLP) techniques. __ __

Probabilities - Monty Hall Game

Brain Teaser and Probability Puzzle

The Monty Hall problem is a brain teaser, in the form of a probability puzzle, loosely based on the American television game show Let’s Make a Deal. As a contender in a game show, should you stick to your original choice or change your door number based on the game host’s suggestion? This project delves into the statistics to determine which is your better option.

__ __