Assigned: Feb. 02 Due: Feb. 09 (before class)

Complete the following Swirl R Programming lessons:

How to complete:

1. Open RStudio
2. Type library(swirl) then swirl() in the console
3. Select “R Programming” and work through each lesson listed above

Submission: You will confirm completion via an in-class poll at the start of Week 2. No screenshots or documentation required.

Assigned: Feb. 09 Due: Feb. 16 (before class)

R Programming:

Complete DataCamp: Introduction to Text Analysis in R – Chapter 1: Wrangling Text.

Optional: Replicate the in-class Orange demo

For extra practice, replicate the in-class demo by loading the presidential speeches corpus into Orange Data Mining and exploring it with the Corpus widget.

Download the corpus from the Data & Scripts page.

Steps:

1. Download the presidential speeches CSV from the Data page
2. Add it to a subfolder in your repo (e.g., /data/president_speeches/)
3. Commit and push via GitHub Desktop
4. Open Orange Data Mining
5. Add a Corpus widget and load the CSV from your data folder
6. Explore the corpus – browse speeches, filter, search
7. Save your workflow (.ows file)

Submission: Upload your .ows file and a screenshot of your Orange workflow to your GitHub repository.

Assigned: Feb. 16 Due: Feb. 23 (before class)

Required — R Programming:

Complete DataCamp: Introduction to R — Chapter 1 (Intro to Basics) and Chapter 2 (Vectors).

Optional — Preprocessing Practice (choose one or both):

Practice the full Korean text preprocessing pipeline on the presidential speeches corpus using Orange Data Mining, R, or both. To demonstrate your preprocessing, generate a word cloud from the result — this is a quick way to verify that your pipeline is working and producing meaningful output.

Option A: Orange Data Mining

Download the preprocessing script for your OS from the Data & Scripts page. Refer to the widget pipeline on the Presentations page (Orange column) for the full workflow.

1. Open Orange and create a new workflow
2. Load the presidential speeches corpus using the Corpus widget
3. Add a Preprocess Text widget — connect it to Corpus
4. Add a Python Script widget and paste the preprocessing script
5. Change TEXT_COLUMN to match your corpus column name
6. Add a Word Cloud widget — connect it to the output
7. Adjust settings until you have a meaningful word cloud
8. Save your deliverables (see below)

Saving your ODM work:

• Screenshot: Right-click the canvas background and select Save As Image, or use your system screenshot tool (Cmd+Shift+4 on macOS, Win+Shift+S on Windows)
• Workflow file: Go to File → Save As and save with the .ows extension

Option B: RStudio

1. Download the R script: week03_preprocessing.R
2. Open it in RStudio
3. Read the comments — the script walks you through each step
4. Run the script section by section (select lines and press Ctrl+Enter / Cmd+Enter)
5. The script will save wordcloud.png to your working directory

Note: The first run installs Python + Kiwi automatically (this takes a few minutes). After that, you can skip the installation steps.

Submitting your work:

1. Create a week03/ folder inside assignments/ in your repository
2. Add your deliverables to that folder:
   • ODM: screenshot (.png) + workflow file (.ows)
   • R: the saved word cloud image (wordcloud.png)
3. In GitHub Desktop: you will see the new files listed as changes
4. Write a short commit message (e.g., “Add week 3 word cloud”)
5. Click Commit to main, then Push origin
6. Confirm your files appear on github.com in your repository

The instructor has access to your repository and will review your submission there.

Assigned: Feb. 23 Due: Mar. 02 (before class) | R Programming extended deadline: Mar. 09, 15:15

Required — R Programming:

Complete DataCamp: Introduction to R — Factors Chapter and Matrices Chapter.

Orange Data Mining Workflow:

Reproduce the in-class Orange workflow to practice building a complete text analysis pipeline from corpus to visualization.

1. Corpus widget: load the presidential speeches CSV
2. Python Script widget: paste the preprocessing script for your OS
3. Corpus widget (second): reload to pick up processed text
4. Preprocess Text: tokenize by whitespace, load stopword list
5. Bag of Words: select TF-IDF weighting
6. Connect at least two visualization widgets (Word Cloud, Bar Plot, Distributions, or Statistics)
7. Take a screenshot of your workflow

Submitting your work:

1. Create a week04/ folder inside assignments/ in your repository
2. Add your deliverables:
   • Workflow file (.ows)
   • Screenshot of your Orange workflow (.png)
   • Two visualization figures (.png): one from raw counts (BoW Count) and one from TF-IDF (e.g., Word Cloud, Bar Plot, or another widget for each)
3. In GitHub Desktop: write a short commit message (e.g., “Add week 4 workflow”)
4. Click Commit to main, then Push origin
5. Confirm your files appear on github.com in your repository
6. Mark your completion on the shared Google Sheet

Assigned: Mar. 02 Due: Mar. 09 (before class)

Required — R Programming:

Complete DataCamp: Introduction to R — Factors Chapter and Matrices Chapter (extended deadline: Mar. 09, 15:15).

Midterm Preparation:

The Week 6 midterm assessment has two parts:

1. Online quiz (~20 min) — multiple-choice questions covering concepts from Weeks 1–5.
2. Hands-on task (~20 min) — download a small new corpus, preprocess it in Orange, and produce a clean Word Cloud. Upload your .ows file, a saved Word Cloud image, and a short .md file (research question + expected findings) to your GitHub repo.

Use the study guide to review all key concepts, the preprocessing pipeline, BoW/TF-IDF, and the Orange workflow.

Study Guide: Week 6 Assessment Study Guide (PDF)

How to prepare:

• Review the study guide — work through the self-check questions
• Practice building Orange workflows end-to-end (File → Corpus → Python Script → Corpus → BoW → Visualization)
• Make sure you understand why each preprocessing step exists, not just how to do it

Assigned: Mar. 09 Due: Mar. 16 (before class)

Midterm Answers

R Programming:

Complete DataCamp: Introduction to R — Lists Chapter and Data Frames Chapter.

Orange Data Mining Tutorials:

Watch the following four tutorials before next class. These cover the clustering methods we will use in Week 7:

• Hierarchical Clustering (Getting Started #05)
• k-Means (Getting Started #11)
• k-Means Explained (Getting Started #12)
• Document Clustering and Cluster Exploration (Text Mining #09)

Assigned: Mar. 16 Due: Mar. 30 (before class)

Orange Data Mining — Hierarchical Clustering:

Replicate the in-class hierarchical clustering demo using the NIKH clustering demo corpus (11 textbooks). Download it from the Data & Scripts page.

1. Load the corpus in Orange using the Corpus widget
2. Preprocess the text (Python Script → Preprocess Text → Bag of Words with TF-IDF)
3. Compute Distances (Cosine)
4. Run Hierarchical Clustering (Ward linkage)
5. Select two clusters from the dendrogram
6. Explore each cluster using descriptive tools of your choice (e.g., Word Cloud, Bar Plot, Corpus Viewer) — try to understand what makes these clusters different

Write-up: Create a short Markdown file (analysis.md) describing:

• What you did — which settings you chose and how you set up your workflow
• Why you did it — your reasoning for the choices you made
• What you found — which books ended up in each cluster, and what vocabulary or themes distinguish them

This does not need to be long — a few clear paragraphs is adequate. The goal is to reflect on the process and interpret what the clustering reveals.

Submitting your work:

1. Create a week07/ folder inside assignments/ in your repository
2. Add the following to that folder:
   • Your Orange workflow file (.ows)
   • Visualization screenshots (.png) showing your two-cluster comparison
   • Your analysis write-up (analysis.md)
3. In GitHub Desktop: write a short commit message (e.g., “Add week 7 clustering”)
4. Click Commit to main, then Push origin
5. Confirm your files appear on github.com in your repository
6. Mark your completion on the shared Google Sheet

R Programming — Swirl Tutorials:

Complete two Swirl lessons from the Exploratory Data Analysis course. You will need to install this course first — it is separate from the R Programming course you have been using.

Installing the course:

1. Open RStudio
2. Run the following in the console:

library(swirl)
install_course("Exploratory_Data_Analysis")
swirl()

1. Select Exploratory Data Analysis from the course list

For more details on installing Swirl courses, see the Swirl student page.

Complete these two lessons:

Documenting completion: Take a screenshot of the completion message for each lesson and save them as .png files in your week07/ folder alongside your Orange deliverables.

Optional — R Programming:

Complete DataCamp: Intermediate R — The apply family Chapter and Utilities Chapter.

Orange Data Mining Tutorials (preparation for Week 8):

Watch the following tutorials before the next class. These cover the word embedding methods we will use in Week 8:

• Word Embedding and Nearest Neighbors (Text Mining #01)
• Semantic Word Search (Text Mining #02)
• Document Embedding (Text Mining #05)

Assigned: Mar. 30 Due: Apr. 06

Orange Data Mining — Word Embeddings:

Replicate the in-class word embeddings demo using the presidential speeches corpus. Explore the results and try to identify patterns — which speeches or presidents appear similar, and what that might tell us about the content.

Submitting your work:

Add the following to a week08/ folder in your repository:

• Your Orange workflow file (.ows)
• Visualization screenshots (.png) of your choice — whatever helps illustrate the patterns you describe in your write-up
• A short write-up (analysis.md) describing what you see

Commit and push your changes, then confirm your files appear on github.com.

Required — R Programming:

Complete DataCamp: Introduction to the Tidyverse — Data Visualization Chapter and Data Wrangling Chapter.

Optional — R Programming:

Complete DataCamp: Introduction to Text Analysis in R — Sentiment Analysis Chapter and Intermediate R — Utilities Chapter.

Orange Data Mining Tutorials (preparation for Week 9):

Watch the following tutorial before the next class. This covers the sentiment analysis methods we will use in Week 9:

• Sentiment Analysis (Text Mining #10)

Assigned: Apr. 13 Due: Apr. 20 (before class)

Required — R Programming:

Complete the following two chapters from DataCamp: Introduction to the Tidyverse:

• Types of visualizations
• Grouping and summarizing

Orange Data Mining — Sentiment Analysis:

Replicate the in-class sentiment analysis demo using Moon Jae-in’s Twitter corpus and the KNU Korean sentiment dictionaries. All files are on the Data & Scripts page.

Build the following workflow:

1. Corpus widget: load moon_twitter.csv
2. Python Script widget: paste the sentiment preprocessing script for your OS — this runs Kiwi and keeps NNG, NNP, VV, and VA tags
3. Corpus widget (second): re-map the text field to processed_text
4. Sentiment Analysis widget: set Method to Custom Dictionary, load positive.txt as the positive word list and negative.txt as the negative word list
5. Box Plot widget: plot sentiment score by the period3 column

Extension — add a retweet or favorites filter:

Extend your workflow by adding a Select Rows widget between the first Corpus and the Python Script. Use it to filter the corpus to high-engagement tweets, then run the same sentiment pipeline on that subset.

Recommended thresholds (pick one):

Connect a second Box Plot to this filtered branch so you can compare sentiment in the high-engagement subset against the full corpus.

Write-up: Create analysis.md in your week09/ folder. In a few paragraphs, describe what the Box Plot by period shows, then address the engagement question: do Moon’s most-shared or most-liked tweets tend to be more positive or negative than the average? Does the pattern hold across all three periods?

Submitting your work:

Add the following to a week09/ folder in your repository:

• Your Orange workflow file (.ows)
• Two Box Plot screenshots (.png): one for the full corpus, one for the filtered subset
• Your write-up (analysis.md)

Commit and push, then confirm your files appear on github.com.

Optional — R Programming:

Complete the following two chapters from DataCamp: Introduction to Text Analysis in R:

• Sentiment Analysis
• Topic Modeling

Orange Data Mining Tutorials (preparation for Week 10):

Watch the following tutorial before the next class. It walks through topic modeling on tweets in Orange, which is what we will build on in Week 10:

• Twitter Widget and Topic Modeling (Text Mining)

Assigned: Apr. 20 Due: May 4 (before class)

This is the final assignment for the course. Run the LDA pipeline we used in class on one of the NIKH corpora and interpret what the model finds. You already know the Orange workflow from earlier weeks, so the focus here is reading the topics, not wiring up widgets.

Pick a corpus. Options, smallest to largest:

• 9-book NIKH History Textbooks (Demo) from the Data & Scripts page. Already tokenized in a processed_text column, so you can skip the Python Script step.
• 11-book NIKH Clustering Demo from the same page. Raw text in full_text; you preprocess it yourself (Kiwi, nouns only). This was the in-class corpus.
• Full 67-book NIKH corpus from the nlp_corpora repo. Optional and more advanced. Only try this if your machine has the memory to handle it. Orange can slow down or crash on a corpus this large, so save your workflow often and be ready to fall back to the 11-book sample if it hangs.

For k, try 3 on the 11-book sample (the in-class setting) or 5 to 6 on the 9-book and 67-book corpora. If you are unsure which to use, pick the 9-book: it is the simplest path.

Interpret. Pick one path, or do both:

• LDAvis view. Adjust the λ slider (start around 0.3 to see distinctive words), click through a few topics, and describe what you see. Which topics sit far apart on the map? Which overlap?
• Topic output. Read the top words for each topic in the Topic Modeling widget, give each one a short label of your own, and comment on which era seems to use which topics most.

Write-up. Create analysis.md in your week10/ folder. A few paragraphs is plenty. Say which corpus and which k you chose, what the topics look like, and what you took away. The Topic Modeling explorer shows the same kind of output on the full 67-book corpus if you want a reference while you work.

Submitting your work:

Add the following to a week10/ folder in your repository:

• Your Orange workflow file (.ows)
• Two screenshots (.png): one of the Topic Modeling widget output, one of the LDAvis view at the λ you picked
• Your write-up (analysis.md)

Commit and push, then confirm your files appear on github.com.

Assessment date: Monday, 4 May 2026 (in class)

Final Assessment page: Schedule, datasets, and rubric

Study Guide: Week 11 Assessment Study Guide (PDF)

The final assessment has two parts, completed in class on your laptop: an online concepts quiz (~15 min) and an Orange Data Mining application exercise (~45 min). It covers the second-half methods from Weeks 7–10: clustering, word embeddings, sentiment analysis, and LDA / topic modeling.

Workshop: Monday, 11 May 2026 (in class)

Paper due: Friday, 5 June 2026, 23:59 (Brightspace)

Final Paper page: Brief, rubric, and dataset menu

The final paper is a short research report (2,500–6,000 words) using one corpus from the curated dataset menu. Submit a single PDF; the URL of your public, FAIR-structured GitHub replication repository goes in a footnote on the paper’s title (suggested wording is in the brief). The paper is marked out of 10.

Come prepared to the 11 May workshop. Before class, pick one corpus from the dataset menu and draft a research question — phrased as a single sentence — that your chosen corpus can plausibly answer. The question and the dataset need to be linked: the question should be answerable with the corpus you pick.

In class, you will present your dataset and question to me for review, and then workshop your analysis plan with classmates working on different corpora.

If you would prefer a corpus from scdenney/nlp_corpora that is not on the menu, or another corpus, email me before the workshop. I will consider it, but it needs approval.