Week 4 Deliverable: Sentiment Analysis of Textbook Sentences

Course: Topical Reading: Digital Humanities (BA3 Korean Studies)
Instructor: Steven Denney Due: Thursday, November 14, 2025 by 17:00

Objective

Reproduce the in-class sentiment analysis workflow in Orange Data Mining (ODM) using a keyword other than 일본 (Japan) and present clear, concise results.

Tasks

1. Choose a Keyword

Pick one word or phrase relevant to Korean history (for example: 미국, 근대화, 독립, 민족, 여성, 평화).
Write 3–5 sentences explaining why you chose this word — it can be intuitive, tied to your interests, or motivated by something from the readings or a quick TF/DF/TF-IDF check.

2. Filter and Explore

Filter the corpus to sentences containing your keyword.
Create one simple exploratory visualization (Word Cloud, Bar Plot, or Heat Map) to understand how your word appears in context.

3. Sentiment Analysis

• Setup and run the Sentiment Analysis workflow from the class demonstration.
• Remember to…
  • Filter out zeros (sentiment ≠ 0).
  • Normalize sentiment values to the interval [–1, 1] (Preprocess → Normalize).
  • Create at least one data visual (e.g., Box or Violin plot) showing the sentiment distribution across periods.

4. Connect Scores to Actual Sentences & Exploration

Provide examples of the sentences behind your sentiment scores:

• Include a few (5-7) sentences** (mix of high, low, and moderate normalized sentiment, if you can).
• Briefly explain how you located these sentences (for instance, via the Corpus Viewer, Data Table, or by saving selections). Include the period and normalized sentiment score for each, if possible (it’s just good context).
• This is meant to be the challenging part — just do your best to associate results with real text. Don’t worry about it being perfect!
• Feel free but not obliged to do additional analysis on documents and words to supplement your analysis.

5. Short Write-Up

Write up everything in a brief report (a few hundred words total will suffice) that includes:

• Why you chose your keyword
• What you found about its tone or representation
• Any visualizations you created
• Any quick notes on normalization, zero filtering, or retrieving example sentences

Deliverables

Submit two files only:

• week04_report.pdf — your short write-up with plots and example sentences
• week04.ows — your Orange Data Mining workflow

Advice: Keep It Reasonable

One keyword. One exploratory visualization. One sentiment visualization.
A short explanation and a handful of example sentences — that’s all that’s required.

Optional R-Programming Track

Students who have been completing the optional R-programming exercises may choose to do this assignment entirely in R rather than in Orange Data Mining.
The same analytical logic applies: select a keyword, perform exploratory analysis, compute sentiment, normalize, visualize, and briefly interpret.
If you’re still gaining confidence in R, feel free to keep your workflow simple — the goal is to show understanding, not to produce complex code.
Those not following the R track should continue using Orange Data Mining.

For those completing the R version:

• Your deliverables will be a PDF report and an R script file (.R).
• You may use generative AI tools (e.g., ChatGPT, Claude) for guided assistance, but you must be transparent about how you used them.
• Please save and submit a downloaded copy of your gen AI chat history, so I can see what kind of help you received.
• You are expected to write the R code yourself to the extent possible, using AI only for guidance, clarification, or debugging support.