Large Language Models, AI for Text Data 2400-MIKRO-AIBF-LLM
The seminar is dedicated to the analysis of text data and the application of artificial intelligence and large language models in natural language processing. The goal of the course is to introduce students to modern methods for transforming unstructured text data, such as customer reviews, surveys, news, social media content, and business documents, into information that supports analysis, decision-making, and process automation.
The course covers the basics of text mining and NLP, preparing text data for analysis, and classical and contextual text representation methods. Topics covered include tokenization, stemming, lemmatization, frequency matrices, TF-IDF, n-grams, and the differences between the bag-of-words approach and sequential methods that consider word order and context. Knowledge representation techniques, part-of-speech identification, NER, and sentiment and emotion analysis methods are also discussed.
The next area of the course focuses on embeddings of words, sentences, and documents, and their applications in semantic similarity analysis, classification, document clustering, semantic search, recommender systems, and chatbots. Models such as Word2Vec, GloVe, FastText, ELMo, and BERT are presented. The seminar also covers the basics of neural networks, their architectures, activation functions, and optimization algorithms.
Particular attention is paid to the fundamentals of modern language models, including token prediction, learning on large datasets, and applications in classification, information extraction, summarization, translation, anomaly detection, and text generation. The course addresses the limitations and risks associated with using language models, including hallucinations, knowledge obsolescence, bias, data privacy, and regulatory compliance. Business chatbot architectures and RAG systems using vector search and verified knowledge sources are also discussed.
Theoretical content is combined with examples and practical exercises implemented in Python. Students analyze texts, compare methods, evaluate model performance, and select NLP solutions for specific business problems.
The project ‘Application of Artificial Intelligence in Business and Finance’ (No. BPI/SPI/2024/1/00078) is implemented by the University of Warsaw as part of the Spinaker programme organised by the National Agency for Academic Exchange (within the project ‘Wsparcie tworzenia i realizacji międzynarodowych programów kształcenia’, funded by the European Funds for Social Development 2021–2027).
Course coordinators
Micro-credential certificate
Learning outcomes
Upon completing the course, the participant
in terms of knowledge:
● knows and understands the fundamental concepts and methods of text mining, natural language processing (NLP), and text data analysis;
● knows classical and contemporary methods of text representation, including bag of words, TF-IDF, n-grams, and embeddings;
● knows the basic neural network architectures used in text analysis, as well as the principles underlying transformer models and large language models;
● knows the capabilities, limitations, and application areas of NLP methods and large language models in data analysis and business processes;
● knows the risks associated with the use of large language models, particularly those related to the reliability of results, data protection, and ethical considerations.
in terms of skills:
● is able to prepare text data for analysis using fundamental natural language processing techniques;
● is able to select appropriate methods of text representation and analysis for a given research or business problem;
● is able to apply selected text analysis techniques, including sentiment analysis, named entity recognition (NER), embeddings, and semantic similarity analysis;
● is able to interpret the results of text data analysis and use them to support decision-making processes.
in terms of social competences:
● is prepared to critically evaluate the results obtained using artificial intelligence methods and large language models;
● is prepared to use artificial intelligence-based tools responsibly and ethically, taking into account data protection and information confidentiality;
● is prepared to independently develop knowledge and skills in modern text data analysis methods and to keep up with advances in NLP and artificial intelligence technologies.
Assessment criteria
Learning outcomes are assessed based on practical tasks and class participation. Assessments include knowledge of theoretical concepts, the correct selection and application of methods to a given business problem, interpretation of results, and critical evaluation of the solutions used.
Bibliography
Jurafsky D., Martin J.H. (2026), Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition with Language Models, Stanford University.
Vaswani A. et al. (2017), Attention Is All You Need, „Advances in Neural Information Processing Systems”.
Teaching materials provided by the instructors during the course.