Natural Language Processing (CSCI 5541)

Course Information

Summary: This course provides an overview of computational techniques enabling machines to interpret and respond to human language. Topics include text classification, distributional semantics, language models, and generative tools like ChatGPT. Prerequisites include knowledge of linear algebra, calculus, probability, and Python. Completion of 5521 or graduate standing is recommended.

NLP is an interdisciplinary field grounded in linguistics, cognitive science, and social sciences. It focuses on building computational models for applications such as machine translation and conversational systems. Topics also address ethical AI design, such as fairness, interpretability, and bias mitigation. Students will read scholarly papers, build datasets, develop NLP models, and reflect on ethical implications in a semester-long project.

Grades are based on homework, participation, and a course project. All materials are posted on the class site. Canvas is used for submissions and grades, and Slack is used for Q&A and discussions. Email responses are not guaranteed.

Class Topics Overview

• Sep 2–16: Class Overview, Intro to NLP, Text Classification
• Sep 18–25: Word Embeddings & Language Models (Models, Search, Evaluation)
• Sep 30–Oct 9: Project Guidelines & Proposal Pitch
• Oct 14–23: Transformers, Pretraining, Prompting, LLMs
• Oct 28–Nov 6: Agents, Efficiency, Alignment, Reasoning
• Nov 11–20: Interpretability, Evaluation, Data, Human-AI Collaboration
• Nov 25–Dec 4: Human-centric NLP, Final Project Poster Presentations

Note: Full lecture slides, readings, and deadlines will be posted on the Schedule tab.

Instructors

Information

Class meets

Tuesday and Thursday, 11:15AM to 12:30PM, Akerman Hall 319

Office hours

DK: mostly over Slack (in-person by appointment)

Shirley: Wednesday 11:30-12 via Google Meet

Shuyu: Thursday 4:30-5:00 at Shepherd 443 or via Zoom

Drew: Tuesday 1-2 at Keller 1-213 Table #1 or via Zoom

Xiaxuan: Monday 1:30-2 via zoom

Class page

dykang.github.io/classes/csci5541/F25

Slack

csci5541f25.slack.com/

Canvas

canvas.umn.edu/courses/518535

Grading and Late Policy

Grading

• 60% Homework (hw1-4 for individual, hw5/6 for team)
• 30% Project (team)
• 10% Class Participation (individual)

Late policy for deliverables

Each student will be granted 5 late days to use for homeworks over the duration of the semester. After all free late days are used up, penalty is 1 point for each additional late day. The late days and penalty will be applied to all team members for group homework and project.

Schedule

We will cover basic NLP representations g(x), to build text classifiers P_theta(y|g(x)) , language models P_theta(g(x)), and large language models P_{theta is large}(g(x)). Based on knowledge you gain during the class, your team will develop your own NLP systems during the semester-long project. Pay attention to due dates and homework release. Lecture slides and homework/project description will be available in .

Date	Lectures and Dues	Readings
Sep 2	Class Overview HW1 out
Sep 4	Intro to NLP Tutorial on Notebook & Colab (Drew)
Sep 9	Text Classification	Determining the sentiment of opinions Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank From Tweets to Polls: Linking Text Sentiment to Public Opinion Time Series Style is NOT a single variable: Case Studies for Cross-Style Language Understanding Beyond Accuracy: Behavioral Testing of NLP Models with CheckList
Sep 11	HW2 out Text Classification: Practices Tutorial on Scikit-Learn and PyTorch (Shirley) Slides Scikit-Learn PyTorch Tutorial on Finetuning & vLLM (Shuyu) Slides Huggingface vLLM HW1 due (Sep 11 Thursday)	Text classifier with NLTK and Scikit-Learn Blog post on Pre-training vs Fine-tuning in LLM: Examples Tutorial on Text classification using HuggingFace's Transformers vLLM
Sep 16	Word Embeddings Project description out	From Frequency to Meaning: Vector Space Models of Semantics Efficient Estimation of Word Representations in Vector Space Linguistic Regularities in Continuous Space Word Representations GloVe: Global Vectors for Word Representation Retrofitting Word Vectors to Semantic Lexicons Gensim's word2vec tutorial
Sep 18	Language Models (1): Models HW3 out HW2 due (Sep 21, Sunday) Project Team Formation due	Chapter 3 of Jurafsky and Martin A Neural Probabilistic Language Model Recurrent neural network based language model Long Short-Term Memory Multivariable chain rule, simple version Long Short-Term Memory Sequence to Sequence Learning with Neural Networks
Sep 23	Language Models (2): Search	The Curious Case of Neural Text Degeneration Mutual Information and Diverse Decoding Improve Neural Machine Translation Sequence Level Training with Recurrent Neural Networks An Actor-Critic Algorithm for Sequence Prediction
Sep 25	Language Models (3): Evaluation Project Brainstorming due	Neural Machine Translation by Jointly Learning to Align and Translate Perplexity of fixed-length models BLEU: a Method for Automatic Evaluation of Machine Translation ROUGE: A Package for Automatic Evaluation of Summaries
Sep 30	Project Guideline HW4 out
Oct 2	Contextualized Word Embeddings HW3 due (Oct 5, Sunday)	Deep contextualized word representations BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding A Primer in BERTology: What we know about how BERT works
Oct 7	Project Proposal Pitch (1)	Slides Deck for Group A
Oct 9	Project Proposal Pitch (2)	Slides Deck for Group B
Oct 14	Transformers Proposal Report due	Attention is All you Need Tutorial on Illustrated Transformer Language Models are Unsupervised Multitask Learners Language Models are Few-Shot Learners Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Oct 16	Pretraining and Scaling HW4 due (Oct 19, Sunday)	Scaling Laws for Neural Language Models On the Opportunities and Risks of Foundation Models On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
Oct 21	Prompting HW5 out	Chain-of-Thought Prompting Elicits Reasoning in Large Language Models Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing Prefix-Tuning: Optimizing Continuous Prompts for Generation
Oct 23	Instructing and Augmenting LLMs	Training language models to follow instructions with human feedback Augmented Language Models: a Survey Toolformer: Language Models Can Teach Themselves to Use Tools Internet-augmented language models through few-shot prompting for open-domain question answering
Oct 28	LLMs as Agents	ReAct: Synergizing Reasoning and Acting in Language Models MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework Generative Agents: Interactive Simulacra of Human Behavior WebArena: A Realistic Web Environment for Building Autonomous Agents
Oct 30	Efficiency HW5 due (Nov 2, Sunday)	Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer The Power of Scale for Parameter-Efficient Prompt Tuning
Nov 4	Alignment HW6 out	Learning to summarize from human feedback Deep Reinforcement Learning from Human Preferences Proximal Policy Optimization Algorithms Direct preference optimization: Your language model is secretly a reward model Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback Dynamic Multi-Reward Weighting for Multi-Style Controllable Generation Benchmarking Cognitive Biases in Large Language Models as Evaluators
Nov 6	Reasoning	Understanding Reasoning LLMs DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning WebGPT: Browser-assisted question-answering with human feedback Training Verifiers to Solve Math Word Problems Easy-to-Hard Generalization: Scalable Alignment Beyond Human Supervision
Nov 11	Interpretability	Interpreting Language Models with Contrastive Explanations BERT Rediscovers the Classical NLP Pipeline Zoom In: An Introduction to Circuits INTERPRETABILITY IN THE WILD: A CIRCUIT FOR INDIRECT OBJECT IDENTIFICATION IN GPT-2 SMALL Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
Nov 13	Evaluation HW6 due (Nov 16, Sunday) Project midterm office hour due	HellaSwag: Can a Machine Really Finish Your Sentence? MEASURING MASSIVE MULTITASK LANGUAGE UNDERSTANDING Training Verifiers to Solve Math Word Problems Evaluating Large Language Models Trained on Code
Nov 18	Data	Annotation Artifacts in Natural Language Inference Data Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics Everyone's Voice Matters: Quantifying Annotation Disagreement Using Demographic Information ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks
Nov 20	Human-AI Collaboration
Nov 25	Human-centric NLP Concluding Remark
Nov 27	No class (Thanksgiving)
Dec 2	Final Project Poster (1)	Project Pages for Group B
Dec 4	Final Project Poster (2) Final Project Report due (Dec 12)	Project Pages for Group A

Homework Details (60%)

All questions regarding homework MUST be communicated with the lead TA via Slack homework channels (e.g., #hw1, #hw2) or during their office hours. Homework 1, 2, 3, and 4 must be completed individually, while Homework 5 and 6 are team-based (maximum of 4 people). The same team must be used for both Homework 4/5 and the course project.

The use of external resources (books, research papers, websites, etc.) or collaboration (students, professors, ChatGPT, etc.) must be clearly acknowledged in your report. See the notes for academic integrity guidelines.

All homework is due by midnight (11:59 PM) on the due date. There are no extensions due to a tight schedule, but late days can still be used. For team assignments, late days will be deducted from each team member. Refer to the homework description and Canvas link for submission:

Homework Assignments & Deadlines

• HW1: Building MLP-based text classifier with PyTorch (10 points, Individual, due: Sep 11 Thursday) (, )
• HW2: Fine-tuning text classifier using HuggingFace (10 points, Individual, due: Sep 21 Sunday) (, )
• HW3: Authorship attribution using language models (LMs) (10 points, Individual, due: Oct 5 Sunday) (, )
• HW4: Generating and evaluating text from pretrained LMs (10 points, Individual, due: Oct 19 Sunday) (, )
• HW5: Prompting with large language models (LLMs) (10 points, Team, due: Nov 2 Sunday) (, )
• HW6: Post training LLMs for alignment (10 points, Team, due: Nov 16 Sunday) (, )

Project Details (30%)

Please carefully read the project description (To be updated) , as it contains essential details about expectations, deadlines, grading rubric, and FAQs. It is your responsibility to ensure you do not miss any of the information provided.

Each team (maximum 4 members) must submit their report, a link to the code (or a zipped archive), and presentation slides/poster on Canvas before the final deadline. Use the official ACL style templates via Overleaf or GitHub.

Below are the required deliverables and deadlines (some fall on weekdays):

• Team formation (1 point, Due: Sep 18)
• Project brainstorming (1 point, Due: Sep 25)
• Proposal pitch (3 points, Due: Oct 7 and 9) – Slide decks: Group A | Group B
• Proposal report (5 points, Due: Oct 14)
• Midterm office hour participation (5 points, Due: Nov 14)
• Poster presentation (5 points, Due: Dec 2 and 4)
• Final report (10 points, Due: Dec 12) | Evaluation rubric

Below are selected reports and posters from previous semesters. Some were extended into publications at top-tier venues:

• [CSCI 5541 S23] Simulating Everyone's Voice: Exploring ChatGPT's Ability to Simulate Human Annotators
• [CSCI 5541 S23] Vision & Language-guided Generalized Object Grasping
• [CSCI 5541 S23] Generalizability of FLAN-T5 Model Using Composite Task Prompting
• [CSCI 5541 S23] Comparing the Effectiveness of Fine-tuning vs. One-Shot Learning on the Kidz Bopification Task
• [CSCI 5980 F22] Generating Controllable Long-dialogue with Coherence → Published in AAAI 2024
• [CSCI 8980 S22] Understanding Narrative Transportation in Fantasy Fanfiction → Published in Workshop on Narrative Understanding (WNU) @ACL 2023

Class Participation (10%)

Your class participation is carefully evaluated. Please upload a profile picture on Canvas and Slack so we can identify you during final evaluations.

The following criteria will be used to assess participation:

• Active involvement and discussion during class
• Engagement on Slack and during office hours with both the instructor and TAs
• Participation and Q&A during the project proposal and poster presentations

We explicitly track both offline and online participation and apply min-max normalization at the end of the course.

If you do not participate in class, Slack, or related discussions, your participation score will be zero.

Prerequisites

Required: CSCI 2041 – Advanced Programming Principles

Recommended: CSCI 5521 – Introduction to Machine Learning, or any course that covers core machine learning algorithms.

This course also assumes the following background:

• Strong programming skills, comparable to a third- or fourth-year undergraduate CS major. All assignments will be in Python.
• Familiarity with basic probability, linear algebra, and calculus.

Notes to Students

Academic Integrity

Unless group work is explicitly allowed, all assignments and project reports must reflect your individual effort. Verbal collaboration is acceptable, but anything you submit must be your own work. Please list the names of collaborators and cite any resources used. If you're uncertain whether something counts as cheating, consult the instructor beforehand. Academic dishonesty will result in a failing grade and will be handled according to University policies.

Students with Disabilities

If you have or suspect you have a disability and require accommodations, please contact your instructor and the Disability Resources Center (DRC).

COVID-19

All students are expected to follow University COVID-19 guidelines, including any masking or vaccination policies. This course is in-person and participation is expected. However, hybrid or online accommodations may be made if necessary. If you're feeling unwell, please stay home and catch up on the material remotely.

Textbooks, Related Courses, and Online Resources

Books

No textbook is required, but the following are recommended:

• Jurafsky & Martin, Speech and Language Processing, 3rd edition – online
• Jacob Eisenstein, Natural Language Processing

Resources

Course materials are inspired by work from Chris Manning (Stanford), Carlos Guestrin (Stanford), David Bamman (UC Berkeley), and Graham Neubig (CMU).