National Taiwan University (NTU) / 國立臺灣大學
Semester: Fall 2026 (115-1)
Instructor: Yen-Chieh Liao 廖彥傑
Office hours: Instructor · TA (Matt Lee)
-
Git & GitHub guide for Mac users: Git 與 GitHub 新手實戰完全手冊 / Git & GitHub Practical Guide (PDF). 感謝 貴評（國發3） 提供！
-
Call for Proposals: 14th Annual Meeting of Asian Political Methodology. Submission deadline: October 9, 2026.
-
Research title + abstract due: October 23, 2026 (Week 7). You are welcome to book instructor office hours to discuss your research ideas, title, or abstract before the deadline.
Use HTML to view slides in your browser, PDF for a downloadable copy, and Source ZIP for editable slide files. The Week 4 Source ZIP also includes the classroom R scripts and required data; one download is enough.
For self-study, read the Lab HTML and run the accompanying R script one section at a time. Download the complete Lab ZIP to get the data and all required files. Worked answers are linked from the handout; no Quarto rendering is required.
| Week | Date | Slides | Lab |
|---|---|---|---|
| Week 1 | Sep 11 |  ·  · Source ZIP |
|
| Week 2 | Sep 18 | · · Source ZIP |
Lab HTML · R script · Lab ZIP · Instructions |
| Week 3 | Sep 25 | Mid-Autumn Festival | |
| Week 4 | Oct 2 | · · Source ZIP |
Lab HTML · R script · Lab ZIP · Instructions |
| Week 5 | Oct 9 | Compensation Holiday for Happy Double 10th Day | |
| Week 6 | Oct 16 | ||
| Week 7 | Oct 23 | ||
| Week 8 | Oct 30 | ||
| Week 9 | Nov 6 | ||
| Week 10 | Nov 13 | ||
| Week 11 | Nov 20 | NTU Sports Day | |
| Week 12 | Nov 27 | ||
| Week 13 | Dec 4 | ||
| Week 14 | Dec 11 | ||
| Week 15 | Dec 18 | ||
| Week 16 | Dec 25 | Constitution Day Holiday |
Introduces the basic concepts of quantitative text analysis and natural language processing, outlines course structure and expectations, and surveys the text-as-data paradigm in social science research.
Covers the use of R and Python for text analysis and sets up development environments (RStudio and Positron). Using supplied corpora, students practice text cleaning, tokenization, and building document-feature matrices in preparation for Week 4. They also check data coverage, metadata, and access for their research projects.
No class on September 25.
Connects preprocessing decisions and bag-of-words representations to dictionary-based measurement. A single workflow moves from a documented corpus to tokens, document-feature matrices, sentiment scores, and validation in political texts. Students also check corpus boundaries and access conditions for potential projects.
No class on October 9.
Covers human coding, training-set construction, supervised document classification, validation, and performance metrics. Students connect coding rules and annotation quality to the substantive claims a classifier can support.
Introduces Wordscores, Wordfish, and Latent Semantic Scaling (LSS) for estimating positions from text, with attention to anchors, assumptions, and external validation. Individual research abstract due: October 23.
Covers LDA, structural topic models, and keyword-assisted topic models, with emphasis on interpreting topics, comparing issue attention, and validating topics as social-science measures.
Combines readability and complexity measures with count-based similarity, word embeddings, and context-specific meaning. Students compare how different representations support research on political language across actors and groups.
Connects POS tagging, named entity recognition, and dependency parsing to contextual representations and transformer-based classification. Fine-tuning and interpretability are introduced through applications and demonstrations, with detailed implementation reserved for supplementary materials.
No class on November 20.
Introduces LLMs for annotation, classification, summarization, and multi-agent workflows. Students assess the reproducibility, validity, and bias of model outputs and consider when human evidence is still needed.
Explores research with multilingual texts, political speech recordings, and audiovisual sources. Students examine how machine translation, speech recognition, and model-generated annotations change the evidence available for political and social research, using published applications and a guided comparison of original and transformed text. Midterm research design report due: December 4.
This session focuses on poster format, visual presentation, and the 5-minute talk. We will review poster drafts together and discuss revisions before the Week 15 presentations.
Students present final research posters, deliver 5-minute presentations, join in-class discussion, and provide peer feedback.
No class on December 25. Students revise their projects independently after the poster session. Manuscript due: TBA. Bonus claim deadline: TBA. Submit the final research paper through Overleaf with instructor editing access; include any bonus claim and its supporting documents in a shared Overleaf project.
Acknowledgments: The design of this course draws primarily on Stefan Müller's quantitative text analysis syllabus (@stefan-mueller). The lab exercises and teaching materials are adapted from those developed by Martijn Schoonvelde (@hjmschoonvelde) for the 2026 Essex Summer School course “Introduction to QTA.”