For my Final Year Project i decided to attempt a soloution for abuse on twitter. I decided to opt for a machine learning method that would use existing datasets to try and create an automated approach. I grasped the use of python techniques like Natural Language Processing, Machine learning pipelines, and sentiment analysis. As i progressed through my research i also attempted to evaluate a completely new proposed method that would utilise text classification and combine with a lexicon. This lead to my project evaluating 3 separate methods. I was able to achieve a first class honours through my effort achieving a 78. Below you can find my supervisors assessment of the overall project.
- Natural Language Processing : Studied natual language processing, to see how text is converted in to vector format and then utilised by python/machine learning. Utilising data cleaning.
- Machine Learning pipelines : Experimented with SVM, Logistic regression and Naive Bayes. Was able to learn how to create pipelines and also how to fine tune parameters
- Twitter API : Gained experaince in utilising API's. Used Twitter API to fetch tweets to test soloutions on
- Data Handling : Had to manage data via sampling to make training and test sets of data more usable, helped to give better accuracy for SVM pipeline
- Data collection : Collected data among participants to generate human results to compare with performance of projects soloutions
