Audience
- Analysts working with feedback, enquiries, reports or survey responses.
- Developers beginning to add language features to business applications.
- Project and operations professionals who can work with simple Python examples.
AI & technology
Natural Language Processing
Understand how AI works with language and explore practical text-analysis applications.
Learning with a practical purpose.
Turn unstructured text into information that a business can use. This module follows a realistic customer-feedback problem from an untidy collection of messages to a small, evaluated language-analysis application. You will explore why the wording, context and quality of the source material matter as much as the model.
Practical exercises connect language concepts with decisions: which enquiries need attention, which subjects recur, and which facts can be extracted reliably. You will compare simple rules, conventional text models and language-model approaches, then explain your choice using evidence from a held-out evaluation set. The emphasis is an understandable workflow that a colleague can inspect and improve.
Live onlineTeaching through Kent Business College, UK
2 hoursOne live session each teaching week
14 sessionsA semester for each module
28 live hoursPer module, plus independent study
The outline below is a 14-session online module, with one two-hour live session each teaching week. Reading weeks, breaks and additional activities follow the agreed cohort calendar. The cohort syllabus, full route, dates, fees, independent study and assessment requirements are confirmed before enrolment.
Who this module is designed for.
Starting knowledge and preparation
- Basic Python variables, lists and functions; introductory practice is recommended before joining.
- Confidence working with CSV files and spreadsheet-style data.
- Use anonymised, public or synthetic text for classroom exercises.
What you will work towards.
- Frame a language problem with a clear output and measurable acceptance criteria.
- Prepare and label text while preserving information relevant to the task.
- Build and compare a baseline classifier and a more advanced alternative.
- Evaluate precision, recall and error patterns on previously unseen examples.
- Present findings through a small application with clear limits and review points.
Explore the 14-session learning plan.
Explore the focus and practical output of each teaching session. Open individual sessions below, or expand the complete plan.
Session 01 From text to a business question
Translate customer-feedback needs into classification, extraction and search tasks with distinct success measures.
Practical output: A problem brief and initial evaluation questions.
Session 02 Collecting a useful text corpus
Inspect sources, document permissions and sample messages that represent the intended operating context.
Practical output: A corpus inventory and sampling rationale.
Session 03 Text preparation without losing meaning
Compare tokenisation, normalisation and cleaning choices using abbreviations, negation and mixed-format messages.
Practical output: A reproducible text-preparation notebook.
Session 04 Labels and annotation quality
Create a category guide and resolve disagreement between two readers of the same message.
Practical output: An annotation guide with reviewed examples.
Session 05 Building a transparent baseline
Use keywords and simple rules to establish a reference point before introducing a learned model.
Practical output: A baseline classifier with documented shortcomings.
Session 06 Representing words and documents
Explore word counts, TF–IDF and vector similarity through a small search and classification exercise.
Practical output: A comparison of document representations.
Session 07 Training a text classifier
Separate training and evaluation data, fit a supervised model and inspect its predictions.
Practical output: A first model and reproducible data split.
Session 08 Measuring the right performance
Read a confusion matrix and discuss how false positives and missed cases affect different users.
Practical output: A precision, recall and error-analysis report.
Session 09 Extracting entities and facts
Identify named entities and structured fields, then validate extracted values against explicit rules.
Practical output: A structured extraction workflow with checks.
Session 10 Meaning and semantic retrieval
Use embeddings to compare meaning and inspect cases where semantic similarity gives misleading matches.
Practical output: A small semantic-search experiment.
Session 11 Language models for text work
Design a constrained classification or extraction prompt and compare its results with your baseline.
Practical output: A prompt specification and comparison table.
Session 12 Handling unfamiliar and difficult inputs
Test ambiguous, multilingual and out-of-scope messages and design an appropriate review or abstention route.
Practical output: An edge-case suite and review policy.
Session 13 Building the feedback workbench
Connect preparation, prediction and reporting into a usable prototype, with a visible audit trail.
Practical output: A working feedback-analysis workbench.
Session 14 Demonstration and analytical defence
Demonstrate the workflow, explain remaining errors and justify which improvements deserve the next investment.
Practical output: A final demonstration and improvement backlog.
Apply the learning
Customer-feedback intelligence workbench
Build a prototype that organises incoming feedback into useful categories and extracts selected facts. Choose a manageable question, such as identifying delivery concerns or recurring training needs. Your final recommendation must show how the evidence supports the proposed use, including cases that require a person to review the result.
Your project evidence can include
- A prepared dataset and annotation guide.
- A runnable analysis notebook or small application.
- An evaluation report and operational handover note.
Use anonymised, public or suitable synthetic material where workplace information is confidential. The issued assessment brief confirms the required evidence.
Assessment, feedback and certification.
- Short practical exercises demonstrate text preparation, labelling and evaluation decisions.
- The project is reviewed for reproducibility, analytical reasoning and evidence of improvement.
- A brief presentation explains model choice, limitations and the proposed human-review process.
Teaching and the relevant course certificate are provided through Kent Business College in the UK. Completion requirements are set out in the module information; attendance alone should not be assumed to meet every assessment requirement.
Questions worth asking.
Is this simply a prompting course?
No. Prompting is one method within a wider process covering data preparation, representations, classification, extraction and evaluation.
Will I train a large language model?
The module focuses on applying and evaluating existing methods. Training a large language model from scratch is outside its scope.
Can I use workplace feedback?
Yes, where you have permission and can remove confidential or personal information. A synthetic or public dataset can be used instead.