Developing an Intelligent Chatbot and Expert System for UK Train Services
This postgraduate Advanced Artificial Intelligence group project requires students to design, implement, evaluate and demonstrate an intelligent conversational system for a UK train operating company. The chatbot combines conversational AI, expert-system concepts, predictive modelling and knowledge-based reasoning to support both railway passengers and operational staff. The coursework is worth 70% of the module and is designed to develop practical experience in applying modern AI techniques to realistic service and operational problems. The first task requires the chatbot to interact with passengers, collect journey requirements such as origin, destination, date and travel time, and identify the cheapest available train ticket. Appropriate railway ticket data sources or APIs may be used, with the selected ticket presented together with access to the relevant booking service. The second task extends the system to improve customer service through train-delay prediction. The chatbot gathers information about a passenger's current train, location, delay and destination before passing these data to one or more predictive models. Students process historical railway-performance data, train and compare suitable machine-learning models, evaluate their accuracy and integrate an appropriate model into the chatbot. The third task introduces an expert system for railway contingencies. Students extract rules and operational knowledge from provided contingency and station-disruption documents and construct a knowledge base capable of advising railway staff during events such as partial or complete line blockages. The system should gather details such as event type, location, time and severity, then provide relevant operational guidance, diversion information, alternative services and passenger advice. The overall architecture may include a user interface, NLP/NLU component, knowledge base, reasoning engine, predictive model, database and optional knowledge-acquisition component. Particular emphasis is placed on context-aware dialogue, reliable reasoning, appropriate fallback responses and effective user experience. Assessment outputs include the working chatbot, source code, a live presentation and demonstration, a detailed group technical report, and an individual contribution report. Overview word count: approximately 375 words. AI-use note: pre-trained LLMs may be used as an engine within the system, but they must not be used to generate coursework code. Any use of an LLM within the solution must be clearly justified and explained in the group report.
Read Model Answer →