This coursework is an individual practical assessment for the Advanced Artificial Intelligence module, focusing on the application of advanced artificial intelligence, natural language processing (NLP), natural language understanding (NLU), and neural network techniques to analyse text documents. The coursework uses a provided social-media dataset containing more than 89,000 posts associated with 947 news headlines. Each social media post is treated as an individual text document and is labelled as either real or fake according to the relationship between the linked news headline, its ground-truth status, and majority annotator agreement. The assessment consists of two main tasks. The first task focuses on the identification of fake text documents. Students are required to perform descriptive data analysis and NLP preprocessing before transforming text into numerical feature representations suitable for neural network classification. Students must design and implement both a Multi-Layer Perceptron (MLP) network and a deep learning neural network to classify documents as real or fake. The assessment requires justification of the selected architectures, including inputs, layers, neurons, activation functions and outputs. Students must also experiment with hyperparameters, evaluate model performance using appropriate metrics, compare different models, select suitable models, and save trained models for later demonstration. The second task focuses on topic discovery using natural language understanding techniques. Students are required to apply text-processing methods such as tokenisation, stop-word removal, lemmatisation or stemming and experiment with at least two different text representation strategies. Possible approaches include Bag of Words, TF-IDF, Latent Dirichlet Allocation, word vectors and word embeddings. The discovered topics must be analysed and interpreted in relation to the document content, associated news headlines, and class labels. The assessment is supported by a practical bench demonstration and a maximum of seven PowerPoint slides covering the design, model improvement process, performance evaluation and discussion of results. Students must submit their own Python code and presentation through Blackboard and demonstrate their saved models without retraining them during the demonstration. The marking scheme allocates 45% to fake text document identification, 35% to topic discovery using NLU, and 20% to the structure, organisation, professionalism and question-and-answer performance of the demonstration.
Advanced Artificial Intelligence · Natural Language Processing · NLP · Natural Language Understanding · NLU · Text Classification · Fake News Detection · Neural Networks · Multi-Layer Perceptron · MLP · Deep Learning · Topic Discovery
Megaminds has supported academic requirements in advanced artificial intelligence, advanced artificial intelligence and related disciplines.