Linguistics 460: Textual Data Analysis with R
UNC-Chapel Hill Linguistics
Fall 2024
Elliott Moreton

Week 16, 2024.12.04.W

Week 16, 2024.12.02.M

Week 15, 2024.11.25.M

Week 14, 2024.11.20.W

Topics: Final-exam review

Class:

• HO on Canvas about final-project presentations
• Randomize order of presentations.
• Go over questions about course content
• Readings?
• Labs?
• Final-exam syllabus (on Canvas)?

Assignments:

• For your group's presentation day, 2 p.m.: Upload presentation slides to Canvas
• For 12/7 (F), 4 p.m.: Revise project draft in response to feedback.

Week 14, 2024.11.19.Tu (not a class day)

Assignments:

• Today: First draft of project due

Week 14, 2024.11.18.M

Topics: Project clinic in class.

Class:

• How to hand in the first draft of the project (Canvas, Assignments, by "project groups")
• Work together in groups on project while instructor circulates.
• Await visit from HVAC specialist.

Assignments:

• Press ahead on projects.
• Tomorrow: First draft of project due

Week 13, 2024.11.13.W

Topics: LASSO regression for document classification.

Class:

• Still using the IMDb database in the textdata library. (Can download imdb_jr.csv from the Canvas site)
• Linear models (reprise from 10/2)
• Logistic models for binary responses
  • Fit by maximum likelihood
  • Fit by LASSO
• How does performance compare on the database?

Assignments:

• Respond to email feedback about initial draft of project proposal if you haven't already.
• Press ahead on projects.
• For 11/19 Tu (note change!): First draft of project due.

Week 13, 2024.11.11.M

Topics: Evaluating classifier performance. Changing parameters to optimize performance

Class:

• Start Zoom
• Still using the IMDb database in the textdata library. (Can download imdb_jr.csv from the Canvas site)
• Fitting the Naive Bayes classifierl
• Using the classifier to make predictions
• Evaluating model performance: Cohen's kappa
• Adjusting model parameters to optimize performance
  • Varying document-term-matrix sparsity
  • Choice of weighting schemes: binary term frequency, term frequeny, tf-idf
  • Overfitting
• In-class project clinic

Assignments:

• For 11/13 (W): Review earlier reading (9/30) of Navarro 2017, Ch. 15, through end of Section 6, on linear regression
• For 11/13 (W): Install the glmnet package
• Respond to email feedback about initial draft of project proposal if you haven't already.
• Press ahead on projects.
• For 11/18 M:: First draft of project due.

Week 12, 2024.11.06.W

Topics: Naive Bayes classifiers for document classification.

Class:

• Start Zoom
• Still using the IMDb database in the textdata library. (Can download imdb_jr.csv from the Canvas site)
• The Inverse Fallacy again (from 9/18)
• Bayes's Rule
• Naive Bayes classifiers: theory
• Building and training a Naive Bayes classifier in R
• Making predictions with the classifier

Assignments:

• Respond to email feedback about initial draft of project proposal.
• Press ahead on projects.
• For 11/18 M:: First draft of project due.

Week 12, 2024.11.04.M

Topics: Document classification using supervised learning. Document-term matrices.

Class:

• Introduction to classification models (slides, Canvas):
  • Document classification
  • Supervised learning
  • Example: Internet Movie Database (IMDb) reviews (Maas et al, 2011)
  • Document-term matrices
• Check in about final projects.

Assignments:

• Respond to email feedback about initial draft of project proposal.
• For 11/06 W: Read Navarro 2017, Ch. 17, through the end of Section 17.2, about Bayesian statistics.
• For 11/06 W: Install the "e1071" package.
• For 11/06 W: Please try to install the "glmnet" package, and let me know in class if it works.
• For 11/18 M:: First draft of project due.

Announcementz: The Divine Comedy sentiment assignment was the last lab assignment in this class.

• Energy should now be going into the final project.
• TA office hours for final projects: M 9-9:50, Th 1-2 in Smith Bldg. 104

Week 11, 2024.10.30.W

Topics: Sentiment analysis.

Class:

• Finish sentiment-analysis slides and in-class problem from last time
• Discuss Lab 8 on tidy text
• Start Lab 9 on sentiment analysis

Assignments:

• For 11/04 M: Lab 9 on sentiment analysis.
• Respond to email feedback about initial draft of project proposal.
• For 11/18 M:: First draft of project due.

Week 11, 2024.10.28.M

Topics: Sentiment analysis.

Class:

• Sentiment
• The "bag-of-words" model
• Sentiment lexicons
• Using inner_join () to look text words up in a sentiment lexicon
• Checking the sentiment lexica against each other.
• (Time permitting:) Check-in about project proposals

Assignments:

• For 10/30 W: Read Chapter 2 of Silge & Robinson's book
• For 10/30 W: Do Quiz 9 on that reading
• For 10/30 W (3 p.m.) Submit an initial proposal for the final project

Announcement: Tuesday's office hours will be earlier than usual this week: 1-2 instead of 2-3.

Week 10, 2024.10.23.W

Topics: "Tidying" text.

Class:

• Finish examples from Chapter 1 of Silge & Robinson's book, in great detail.
• Important functions: unnest_tokens (), group_by (), anti_join (), bind_rows (), pivot_wider (), pivot_longer ().
• Start Lab 8 ("tidy" text).

Assignments:

• For 10/28 M: Lab 8 on "tidy" text.
• For 10/30 W: Submit an initial proposal for the final project (via the "Assignments" area on Canvas).

Week 10, 2024.10.21.M

Topics: "Tidying" text

Class:

• How can we break a text document down into individual words?
• Go through examples from Chapter 1 of Silge & Robinson's book, in great detail.
• Important functions: unnest_tokens (), group_by (), anti_join (), bind_rows (), pivot_wider (), pivot_longer ().
• (Last 10--15 minutes) Go over
  • Lab 7 on regular expressions.
  • Question 12 on the midterm (the Old Faithful geyser eruptions)

Assignment:

• Before 10/30 W: Introduce yourself in the Discussions area of Canvas. This is optional, but it could help with the next assignment:
• For 10/30 W: Find two or three partners for your final-project group.
• For 10/30 W: Submit an initial proposal for the final project.

Week 9, 2024.10.16.W

MIDTERM

• In this room, at this time, with computer
• Please see Midterm Syllabus for details

Assignment:

• For 10/21 M: Lab 7, regular expressions.
• For 10/21 M: Read Chapter 1 of Silge & Robinson's Text Mining with R: a tidy approach
• For 10/21 M: Install the gutenbergr package.
• Before 10/30 W: Introduce yourself in the Discussions area of Canvas. This is optional, but it could help with the next assignment:
• Before 10/30 W: Find two or three partners for your final-project group.
• For 10/30 W: Submit an initial proposal for the final project.

Week 9, 2024.10.14.M

Topics: Midterm review

Class:

• Start Zoom.
• Go over questions about course content
• Readings?
• Labs?
• Any unanswered questions from the quizzes?

Week 8, 2024.10.09.W

Topics: Regular expressions.

Class:

• Continue slides on Canvas: Liquid co-occurrence in U.S. baby names (Martin, 2007, Ch. 3), exemplifying use of regular expressions to describe strings
• Some places to look for data sets:
  • Compendium of R data sets (more than just in the datasets package)
  • Words in the World
  • Experimental Linguistics in the Field
  • A page of psycholinguistic databases and miscellaneous resource
  • This list of text corpora (en.wikipedia.org)
  • Project Gutenberg, of which more after Fall Break
  • Web scraping (see Lab 7 and references therein)
• Visit to Odum Institute.

Assignment:

• For 10/14 M: Read over the Midterm Syllabus on Canvas.
• For 10/16 W: Prepare for midterm.
• For 10/21 M: Lab 7, regular expressions.
• For 10/21 M: Read Chapter 1 of Silge & Robinson's Text Mining with R: a tidy approach
• For 10/21 M: Install the gutenbergr package.
• Before 10/30 W: Introduce yourself in the Discussions area of Canvas. This is optional, but it could help with the next assignment:
• Before 10/30 W: Find two or three partners for your final-project group.
• For 10/30 W: Submit an initial proposal for the final project.

Announcement: I've updated the "Final Projects" handout on Canvas to correct the dates.

Week 8, 2024.10.07.M

Topics: Final projects. Regular expressions.

Before class:

• (Instructor:) Start Zoom.
• (Students:) Install babynames library. There may be a lot of dependences; please let RStudio install them too.

• Start slides on Canvas: Liquid co-occurrence in U.S. baby names (Martin, 2007, Ch. 3)
• Structure of a final project
• Introduction to the tidyverse
• Time permitting: Introduction to regular expressions
• Handout on final projects (on Canvas)

Assignment:

• For 10/09 W (3/20 W): Read the chapter on regular expressions in Wickham et al. 2023.
• Before 10/30 W: Introduce yourself in the Discussions area of Canvas. This is optional, but it could help with the next assignment:
• For 10/30 W: Find two partners for your final-project group.
• For 11/06 W: Submit an initial proposal for the final project.

Week 7, 2024.10.02.W

Topics: Linear regression

Class:

• Linear associations
• The simple linear regression model (y = mx + b)
• Model fitting: ordinary least-squares regression
• Interpreting the fitted model
• Start Lab 6, linear regression.

Assignment for 10/7 (M):

• Read Chapters 1--3 of Handling and Processing Strings in R, by Gaston Sanchez.
• Lab 6, linear regression

Announcement: Office hours tomorrow (Thursday the 3rd), 2:30-3:30, instead of Friday. (That's just for this week.)

Week 7, 2024.09.30.M

Topics: Tests of association between categorical variables

Class:

• Categori(c)al vs. scalar variables
• Association and independence
• Mosaic plots
• The chi-squared test of association
• Fisher's exact test
• Time permitting: Discuss Lab 5

Assignment:

• For 10/02 W: Reading: Navarro 2017, Chapter 15, through the end of Section 6
• For 10/02 W: Reading Quiz 8, on Canvas

Week 6, 2024.09.25.W

Topics: Testing hypotheses about a population mean.

Class:

• Start Zoom.
• General caveats about null-hypothesis significance testing (from last time's slides)
• Hypotheses about a population mean: t-tests
  • Assumptions
  • Worked example
  • In-class problem
  • Start on Lab 5.

Assignment for 9/30 M: Lab 5, comparison of reaction times in a synonymy-choice task

Announcement: Schedule for Midterm:

• 10/14 M: Midterm review day, for questions about course content
• 10/16 W: MIDTERM, in class.

Week 5, 2024.09.18.W

Topics: Null-hypothesis significance testing

Class:

• Overview of NHST
• Logic of NHST
• Work through an example (binomial test)
• Do an in-class problem (chi-squared test for equal proportions)
• Discuss caveats

Assignment:

• For 9/25 W: Please read Navarro 2017, Chapter 5, Section 6, and Chapter 13, through the end of Section 3
• For 9/25 W: Reading Quiz 7.

Announcement: Office hours on Friday, 9/20, will be 1:30-2:30 instead of 2:30-3:30.

Reminder: Midterm is the week of 10/14-10/16 (one of those two days).

Week 5, 2024.09.16.M

Topics: Confidence intervals

Class:

• Computing a z or t confidence interval for a sample from a normally-distributed population.
• Frequentist interpretation of confidence intervals
• Some misinterpretations of CIs.
• Factors determining the width of CIs. How much influence does the experimenter have?
• What if the population isn't normally-distributed?
• Other kinds of CIs.
• Uses of CIs

Assignment:

• No lab this week (and no class on Monday 9/23).
• Reading: Navarro 2017
  • For 9/18 W: Chapter 11, through the end of Section 7
  • For 9/25 W: (Chapter 5, Section 6, and) Chapter 13, through the end of Section 3
• For 9/18 W: Reading Quiz 6.

Week 4, 2024.09.11.W

Topics: Sampling theory.

Class:

• Statistical inference (finish slides from last time)
• Sample statistics and the sampling distribution of the mean
• The Central Limit Theorem
• Start on Lab 4, sampling distributions and confidence intervals.

Assignment for 9/16 M:

• Read Navarro 2017, Ch. 10.5, which will introduce confidence intervals. There is no associated quiz.
• Do Lab 4, sampling distributions and confidence intervals

Week 4, 2024.09.09.M

Topics: Frequentist vs. Bayesian statistics. Probability distributions. Samples.

Class:

• Lab 3, descriptive statistics
  • Deciphering authors' description of what they did
  • dplyr not needed yet (coming in 2nd half)
• Inferential vs. descriptive statistics
• Frequentist vs. Bayesian inference
• Probability distributions. Examples:
  • Binomial
  • Normal
• Parameters and inference

Assignment for 9/11 W::

• Read Navarro 2017, Ch. 10, through the end of 10.4.
• Reading Quiz 5

Week 3, 2024.09.04.W

Topics: Descriptive statistics: Linear relationships and correlation.

Class:

• Comments on Lab 2
  • Need to follow authors' account of what they did with the data. What observations did they exclude?
  • Shape of RStudio graphics window can affect shape of saved graphic.
  • Useless libraries loaded by a number of people. (Why?)
  • Use line breaks, indentation, mnemonic variable names, and in-line comments to make the code more human-readable (and hence more maintainable later on).
• Describing relations between two numeric variables
  • Linear relationships
  • Correlation coefficients: Pearson's, Spearman's
  • Cautionary tale: Anscombe's Quartet
• Start on Lab 3 (descriptive statistics).

Assignment for 9/9 M:

• Lab 3 (descriptive statistics)
• Read Navarro 2017, Ch. 9 through the end of 9.5. There is no associated quiz.

Week 2, 2024.08.28.W

Topics: Data frames and basic visualization.

Class:

• Intro to data graphics in R (continued from last time; slides on Canvas)
  • Saving graphics to a file
• A couple of useful data-summarization functions.
• Start on descriptive statistics (slides on Canvas)
  • Descriptive vs. inferential statistics
  • Measures of central tendency
  • Measures of variability
  • Describing data grouping using formulas
  • Collapsing within groups using aggregate ()
• Start on Lab 2 (data frame and scatterplot).

Assignment for 9/4 Wed:

• Lab 2 (data frame and scatterplot). (This is due on Wednesday, rather than Monday, because of the holidays.)
• Read Navarro 2017, Chapter 5, Sections 5.1, 5.2, 5.4, 5.5, 5.7--5.10, and do Reading Quiz 4 (on Canvas)

Week 2, 2024.08.26.M

Topics: Data frames. Start visualization.

Class:

• Data frames (continued from last time; slides on Canvas Modules)
• Intro to data visualization (slides; Canvas Modules)

Assignment: Read Navarro 2017, Chapter 6, through the end of Section 6.2, about base R graphics ("traditional" R graphics). Do Quiz 3 on the reading.

Week 1, 2024.08.21.W

Topics: R data types

Class:

• Comments on Lab 1
  • Functions are reusable code, so you can save effort by reusing them.
  • As we start seeing more complex code in our textbook, watch how it is formatted and commented to make it easier for humans to follow.
• Data types in R (slides, Canvas Modules).
• Time permitting: Start data frames (slides, Canvas Modules).

Assignment for 8/26 M:

• Reading, from Navarro (2017), Learning Statistics with R: Ch. 3.10 through end of Ch. 4
• Quiz 2 (Canvas Quizzes)

Week 1, 2024.08.19.M

Topics: Course organization. R and RStudio.

Class:

• Course organization
  • Syllabus
  • UNC Honor Code.
• Read script on preparedness in the classroom.
• Installing R and RStudio
• Start on Lab 1, "Hello, World!" (Canvas, Assignments).

Assignment for 8/21 W:

• Reading, from Learning Statistics with R (one of our free on-line textbooks):
  • Ch. 1, Sections 1.1 and 1.2, on what statistics does for us
  • Ch. 3, Sections 3.1 through 3.9, about R
• Quiz 1 (Canvas, Quizzes)
• Lab 1 (Canvas, Assignments)