Sentiment analysis in Natural Language Processing (NLP) and Computational Linguistics (CL) involves identifying and analyzing polarity—positive, negative, neutral, or mixed—within text. This course provides a structured introduction to sentiment analysis methods, situating it within social meaning. The course covers mostly appplications of modern deep learning methods with a focus on large language models (LLMs), combining lectures with hands-on lab sessions to ensure a balance between theoretical understanding and practical implementation.
Slack Channel: cl-565_sentiment
Upon completion of this course, students are expected to:
- Become familiar with the foundations of sociopragmatic meaning with a focus on sentiment analysis
- Apply classical methods and large language models (llms) for sociopragmatics
- Design and develop real-world sociopragmatics systems using llms and agentic workflows
- Enhance text classification with explaination and retrieval-augmented models
- Critically evaluate societal considerations in text classification
This is a live presentation class. Students are expected to attend lectures when possible. Recordings will be available, but should be used as a study tool instead of a primary method of learning the material.
Classes meet Tuesday and Thursday at 9:30 a.m. - 11:00 am, in MCLD 3018. Labs are Tuesday from 2-6 p.m., in ORCH 4018.
The MDS-CL is an in-person program - you are expected to attend classes and labs. There will be no recordings of lectures. Interaction in class has been shown to improve understanding, and collaborative environments improve learning.
Attendance will be taken with iClicker. To receive full marks, you must attend 6 of 8 classes. Attendance in fewer classes than these will result in a grade of 0 for attendance. I understand that illness and emergencies happen (if you are ill, please do not come to class).
This is both an assignment- and project-based course. You'll be evaluated as follows:
| Assessment | Weight | Due Date | Submission Location |
|---|---|---|---|
| Lab Practice 1 | 16% | Feb 15, 11:59pm | Submit to GitHub |
| Lab Practice 2 | 16% | Mar 1, 11:59pm | Submit to GitHub |
| Short Reflection 1 | 6% | Mar 5, 11:59pm | Submit to GitHub |
| Lab Practice 3 + Project Milestone 1 | 24% | Mar 8, 11:59pm | Online |
| Short Reflection 2 | 6% | Mar 10, 11:59pm | Submit to GitHub |
| Lab Practice 4 + Project Milestone 2 | 24% | Mar 15 | Online |
| Engagement & Participation | 8% | Ongoing | In-Class: iClicker + Discussions |
| Position | Name | Slack Handle | GHE Handle |
|---|---|---|---|
| Main Instructor | Muhammad Abdul-Mageed | @Muhammad Mageed |
@amuham01 |
| Lab Instructor | Jungyeul Park | @jungyeul |
@jungyeul |
| Teaching Assistant | Yadong Liu | @Yadong Liu |
@yadliu |
- Wednesdays at 1:00-2:00 (or by appointment) at TFS Office.
| Lecture | Topic | Readings or Materials |
|---|---|---|
| 1 | Course Intro., Sentiment Task | xxxx |
| 2 | Classical Methods | xxxx |
| 3 | Social Meaning (e.g., emotion, toxic language) | xxxx |
| 4 | BERT, LLMs, CoT, Prompting, Instruction Finetuning | xxxx |
| 5 | Real-World Sentiment: LLM Agents I | xxxx |
| 6 | Real-World Sentiment: LLM Agents II | xxxx |
| 7 | Real-World Sentiment: Sentiment Explanation, RAG | xxxx |
| 8 | Societal Considerations: Low-Resource and Multilingual, Bias, Energy |
- Speech and Language Processing, 3rd edition (J&M)
- Lexicon-based methods for Sentiment Analysis (LMSA)
- Sentiment Analysis and Opinion Mining (SAOM)
- Modeling Arabic subjectivity and sentiment in lexical space(ArabicSSA)
- EmoNet (Abdul-Mageed and Ungar)(EmoNet)
- NLTK howtos for WordNet and SentiWordNet.
- Stanford Sentiment Treebank (StanfordSA)
- MPQA corpus and associated documentation (MPQA)
- GATE setup for MPQA
- Understanding Convolutional Neural Networks for Text Classification (Jacovi et al 2017)
- Target-dependent Twitter Sentiment Classification (Jiang et al. 2011)
- Adaptive Recursive Neural Network for Target-dependent Twitter Sentiment Classification (Deng et al. 2014)
- SVM Rank
- Argument Mining: A Survey (AMAS)
- Magnets for Sarcasm (Ghosh and Veale)
- Personality, Gender, and Age in the Language of Social Media (Schwartz et al)
- Computational Sociolinguistics
- The Risk of Racial Bias in Hate Speech Detection (Sap et al.)
- Cloroplaths in Plotly
- the datetime library
- Langid paper
- ekphrasis package
- Twitter POS tagger
Please see the general MDS policies.
Acknowledgements: Many of the readings are from a previous course by Dr. Garrett Nicolai.