Introduction to Natural Language Processing

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Explain natural language processing and how text is represented as data
• Clean and normalise text while avoiding the removal of useful meaning
• Tokenise text and prepare reproducible text-processing workflows
• Create and interpret bag-of-words, n-gram and TF-IDF features
• Build and evaluate an introductory sentiment-analysis model
• Train text classifiers for spam detection and complaint categorisation
• Explain word embeddings and use basic similarity measures
• Identify practical challenges in African-language NLP, including limited datasets, spelling variation, morphology and code-switching
• Apply privacy, anonymisation and human-review principles when working with customer messages
• Build, evaluate and present a WhatsApp or SMS complaint-categorisation prototype

Modules

1. 1

   NLP Foundations — Text as Data

   Understand what NLP is, how computers represent language and where text analytics is used in Uganda, East Africa and Africa

   Natural language processing conceptsText as structured and unstructured dataCommon NLP tasksDocuments, sentences and tokensUnicode textRegional use casesPrototype versus production systems
2. 2

   Text Cleaning and Tokenisation

   Prepare text consistently while preserving information that may matter for the task

   Lowercasing and case preservationWhitespace and punctuationURLs, phone numbers, emojis and hashtagsStop wordsStemming and lemmatisationWord and sentence tokenisationHandling spelling variationAvoiding over-cleaning
3. 3

   Bag-of-Words and TF-IDF

   Convert text into numerical features for classical machine learning

   Vocabulary buildingCountVectorizerBag-of-words matricesUnigrams and n-gramsSparse matricesTF-IDF weightingFeature inspectionTrain-test separation
4. 4

   Sentiment Analysis Basics

   Build and assess a basic model for positive, negative or neutral text

   Defining sentiment labelsLexicon and machine-learning approachesPreparing labelled examplesBaseline classifiersNegation and contextAccuracy, precision, recall and F1-scoreBias and error analysis
5. 5

   Text Classification — Spam and Complaints

   Create reusable pipelines that assign messages to practical categories

   Problem and label definitionSpam detectionComplaint categoriesNaive BayesLogistic regressionLinear support vector machineScikit-learn pipelinesClass imbalanceConfusion matrixHuman review of uncertain cases
6. 6

   Introduction to Word Embeddings

   Develop intuition for representing words as dense vectors and measuring similarity

   From sparse to dense representationsDistributional meaningPre-trained embeddingsCosine similarityNearest wordsVisualising vectorsOut-of-vocabulary wordsBias and limitations
7. 7

   African-Language Text Challenges

   Evaluate the practical limits of NLP when data, tools and language coverage are uneven

   Limited labelled datasetsEnglish and African-language code-switchingDialect and spelling variationRich morphologyTokenisation challengesDataset documentationWorking with language speakersTransfer limitsResponsible claims
8. 8

   Capstone — Customer Complaint Classification Tool

   Build and present a prototype that categorises anonymised WhatsApp or SMS complaints for a small business

   Define useful categoriesCreate or select anonymised sample messagesClean and label dataBuild a TF-IDF classification pipelineEvaluate resultsReview misclassificationsDocument privacy and model limitationsCreate a simple notebook or interfacePresent the prototype