Title : A comparison of Transformers with traditional and modern ML
algorithms (Support Vector Machines, Naive Bayes, KNN) in sentiment
analysis. The objectives of this assessment are: 1. To Critically
Analyse and Synthesise Key Literature, critically evaluating primary
sources in the literature to refine and define a research topic,
culminating in the formulation of a challenging yet feasible project
aim. 2. To Implement and Evaluate Research Methodologies, encompassing
the application of suitable research methods for the investigation, and
the synthesis of a solution that aligns with the project aims. 3. To
Construct and Justify a Comprehensive Dissertation, creating a well
structured, coherent, and cogent dissertati
on that effectively
communicates the findings of the research. The submitted dissertation
must include: –Python code that demonstrates the comparison in accuracy
between Transformer models and other ML models applied on a dataset (eg
Amazon product reviews, IMDB movies reviews). This can be achieved by
either using Python ML libraries (Tensorflow,Keras,Pytorch, Scikit etc)
or implementing the models from scratch. –A document of approximately
8000 words that describes the development process, presents background
literature etc. The document should also include an overview of topics
related to the subject of the dissertation such as the theoretical
background of neural networks, a detailed presentation of Transformers,
methods of tokenization and embedding of text, methods for positional
encoding etc.