This module introduces the fundamental concepts of Natural Language Processing (NLP). It begins with the definition of NLP and explores a variety of real-world applications. You will gain an understanding of Natural Language Understanding (NLU) and Natural Language Generation (NLG). The module also covers key evaluation metrics used to assess NLP systems. Additionally, a hands-on lab session will guide you through the implementation of basic NLP preprocessing techniques.

This module introduces essential NLP preprocessing techniques. It begins with regular expressions for text pattern matching, followed by an overview of words and corpora as foundational data sources. Sentence segmentation and tokenization are then covered through practical demonstrations. Finally, the module explores normalization, lemmatization, and stemming as methods to standardise text, with a demo highlighting their differences and effects.

This module explores lexical and vector semantics, focusing on computational representations of word meaning. It covers word vectors, Bag of Words, and co-occurrence matrices to capture contextual relationships. Techniques such as TF-IDF are introduced to measure word importance, along with methods for computing word similarity. Practical examples and mathematical exercises on TF-IDF help reinforce these core NLP concepts.

This module introduces Word Embeddings, focusing on the transition from sparse to dense vector representations of words. It covers Word2Vec models, including Skip-gram and CBOW, explained with simple, intuitive examples. The module also explores GloVe embeddings, which capture global word co-occurrence statistics for improved semantic understanding. Learners will visualise word embeddings to gain insights into how words relate in vector space. Finally, the module highlights real-world applications of word embeddings in NLP tasks like sentiment analysis, machine translation, and question answering.

This module introduces Language Modeling (LM) and its role in predicting word sequences in natural language. It explores practical applications of LMs and explains N-gram models, including challenges like generalization and handling zero probabilities. Techniques such as smoothing and stupid backoff are covered to improve model robustness. The module concludes with methods for evaluating language models using standard metrics.

This module explores the use of Neural Networks in Language Modelling, starting with the fundamentals of Feed-Forward Neural Networks and their training process for language tasks. It introduces Neural Language Models, which capture complex patterns in text beyond traditional statistical methods. The module also provides a foundational understanding of Large Language Models (LLMs) and their capabilities. Finally, it introduces Prompt Engineering as a technique to effectively interact with and guide LLMs for various NLP applications.

This module provides an introduction to Part-of-Speech (POS) Tagging, techniques to perform POS Tagging and their applications in NLP. POS tagging is a fundamental task in Natural Language Processing (NLP) that involves assigning grammatical categories (like noun, verb, adjective) to words in text. Starting from basic linguistic foundations and real-world applications, the module dives into the evolution of POS tagging techniques—from statistical models like Hidden Markov Models (HMMs) and Maximum Entropy classifiers, to modern deep learning approaches using Recurrent Neural Networks (RNNs). Learners will gain a strong theoretical understanding and insight into how POS tagging supports downstream tasks like parsing, named entity recognition, and machine translation. The module includes a hands-on coding demonstration for POS tagging.

This module introduces students to the syntactic structure of natural language and its critical role in Natural Language Processing (NLP) applications. Parsing is the task of assigning a structured representation—typically a tree—to a sentence, revealing the grammatical relationships between its components. The module begins by revisiting Context-Free Grammars (CFGs) and how they form the foundation for syntactic parsing. We explore Constituent Parsing, introducing classical parsing techniques such as the CKY (Cocke-Kasami-Younger) algorithm. The module then transitions to modern span-based neural parsing approaches that use neural networks to score and predict parse trees. A significant portion of the module is dedicated to Dependency Parsing, where syntactic structure is represented through direct relationships between words rather than phrases. Students will study both transition-based and graph-based dependency parsers, gaining insight into their strengths, algorithmic designs, and practical performance. Throughout the module, we emphasise real-world NLP applications.

This module explores the semantic dimension of natural language by covering both lexical semantics—including word senses, ambiguity, and disambiguation techniques—and the semantic web—a framework for enabling machine-readable, structured understanding of web data. The module starts with foundational concepts in lexical semantics and WordNet, then proceeds to classical and modern word sense disambiguation (WSD) methods. The second part focuses on Semantic Web technologies, covering ontologies, knowledge graphs, RDF/OWL, and their role in enabling intelligent systems and knowledge-driven NLP applications.

This module introduces students to the evolution of neural network architectures in NLP, beginning with recurrent models (RNNs), progressing through attention mechanisms, and culminating in Transformer-based models that have revolutionised natural language processing. Through hands-on coding and application-driven lessons, students will explore how Transformers power state-of-the-art systems in sentiment analysis (text classification), machine translation, and question answering. The module emphasises both theoretical foundations and practical implementation using modern deep learning frameworks.

End Term Examination