BIM 432 - Natural Language Processing

The aim of this course is to teach the basic concepts and techniques in the field of Natural Language Processing (NLP), to develop skills in working with text data, and to enable students to implement various NLP applications with Python and related libraries. In addition, the course aims to enable students to practice language models, sentiment analysis, text summarization, machine translation and chatbot development, and to develop solutions to real-world NLP problems.

The definition of NLP, its uses and the structure of the language (morphology, syntax, semantics) will be explained. Basic usage of Python and popular NLP libraries (NLTK, spaCy, Hugging Face) will be demonstrated. Students will learn how to process text data using tokenization, stopwords extraction, lemmatization and stemming. Students should practice text preprocessing steps with Python. Bag of Words (BoW) and TF-IDF methods are used to convert text to numeric data. Discover relationships between words with Word2Vec, GloVe and FastText. Logistic regression and Naive Bayes classifiers are used for sentiment analysis. RNN, LSTM, GRU and Transformer architectures are introduced. The basics of language models such as BERT, GPT, T5 and their use with Hugging Face are explained. Turkish NLP challenges and preprocessing operations on Turkish texts using Zemberek library. Students practice on Turkish texts. Noun entity recognition (NER) and word sense disambiguation (WSD) methods are taught. NER model is developed with spaCy to recognize entities in texts. Machine translation methods (rule-based, statistical, neural) and Transformer-based models are taught. English-Turkish translation is practiced with MarianMT. Abstractive and extractive summarization methods and TextRank algorithm are explained. Students summarize Turkish news texts. Rule-based and learning-based chatbots are introduced. University information chatbot is developed using Rasa and DialoGPT. Knowledge extraction and Transformer based question-answer systems are taught. Question and answer applications with BERT. Big language models (GPT-4, LLaMA) and ethical issues in NLP (bias, misinformation) are discussed. Text generation is done using GPT-4 API. Challenges in NLP projects and model evaluation metrics (Perplexity, BLEU, ROUGE, F1-score) are taught. Students perform model optimization on the projects. Students present their projects and demonstrate what they have learned.

Daniel Jurafsky, and James H. Martin, "Speech and Language Processing", Third Edition, Prentice Hall, 2020.

1- Indurkhya, N., & Damerau, F. J. (2010). Handbook of natural language processing. Chapman and Hall/CRC.

2- Manning, C., & Schutze, H. (1999). Foundations of statistical natural language processing. MIT press.

1. Quiz and/or homework will be related to the topics of the week. 1 homework/quiz will be given.

2. For the application part, students are expected to develop projects in accordance with the content of the course and the projects developed by the students are expected to be presented by the students in the 14th and 15th weeks.

1. Attendance: According to the regulations, if a student fails to attend 30% of the total class time, he/she will be marked absent (DZ) and will fail the course.

2. Academic honesty: In cases of unethical behavior such as plagiarism, copying, etc., you will be subject to disciplinary proceedings and penal sanctions in accordance with IZU's relevant regulations.