Ling H286 -- Analyzing the sounds of languages, Au 2007

Contents

• Course description
• Administration
• Syllabus
• Textbooks and associated ...
• Homework assignments
• Data analysis reports
• Term project
• Useful links

Course description

Speech is a very complex behavior. Saying even a simple sentence such as Luce hit another home run. takes more motor coordination than actually hitting the home run. Yet we are not surprised when we talk with a child as young as three or four who can say a sentence like this very naturally and intelligibly. Understanding speech also is a very complex skill. We take it for granted that we can listen to this sentence and correctly identify the first word as Luce rather than as Ruth and the second word as hit rather than as heat. Yet speakers of some other languages besides English have a great deal of difficulty distinguishing these two pairs of words. How can we describe the similarities between the consonants in Luce versus Ruth or between the vowels in hit versus heat that make these two word pairs difficult for speakers of some other languages? Can we analyze speakers’ behavior in saying words and sentences, and listeners’ behavior in listening to them, well enough to be able to construct models that predict which sounds will be difficult for children acquiring their first language or for adults acquiring a second language? In this course, we will introduce pertinent ideas and results from research in the various disciplines that have contributed to our understanding of the sounds of languages. We will introduce some of the quantitative analytical tools that are used in the phonetic sciences, and do several experiments in class, to give a flavor of the diverse research methods that speech scientists have developed to try to determine how speech is produced and perceived by humans. We will also take a brief look at how speech engineers have applied this knowledge to develop computer speech synthesis systems.

Administration

Syllabus

A full and detailed syllabus is available here in pdf format.

Textbooks

The readings for the course are from:

• Vowels and Consonants, 2nd ed., by Peter Ladefoged (2005, Blackwell).
  This one is available at the main campus bookstore and other outlets. (Be sure to get the 2nd edition.)
• Notes on Probability and Statistics for Analyzing the Sounds of Languages. Grant McGuire and Mary E. Beckman (2007).
  This one is a book in progress, the assigned parts of which will be posted in the appropriate place below as they come ready.

McGuire and Beckman (2007) was originally designed and is being developed as a companion textbook to Ladefoged (2005). When completed, it will be a series of homework assignments and data analysis problems to accompany the successive chapters of that book. Our original purpose was to develop this book as a way to enable linguists and speech scientists to use Peter's book in a data analysis course -- i.e., an introductory course on probability and statistics as applied in the analysis of real data. On January 24, 2006, the book acquired a second purpose. Like everyone else in our field, we learned many things from Peter Ladefoged, and hope that this book will be accepted by our colleagues as a suitable tribute to his legacy as a scientist and one of the world's greatest teachers and mentors.

Homework assignments

The readings from McGuire & Beckman (2007) and the homework assignments that accompany the readings in Ladefoged (2003) will be posted here as they become ready. If a link does not work, that means that the reading or homework assignment is not ready yet.

• Homework assignment 1 -- Questions related to Ladefoged (2005), Chapter 1 on "Sounds and Languages".
  (Use these questions to guide your reading of the chapter and turn in your answers to the questions at the beginning of class on Monday, September 24.)
• Try to download the R program and install it on your laptop or home computer before class on Wednesday, September 26, so that you can ask any questions about problems you are having in class or in my office hour on Thursday. Download this R script and try running it to see whether your install was successful and you can use R to do the first Data Analysis problem. (If you use a PC, right-click on the link to download it.)
• McGuire & Beckman (2007) Chapter 1 -- Chapter on "Introduction to data analysis".
  (We will go over the concepts in this chapter in class in Week 2. You can download the chapter and read it as a review of the concepts before we do quiz #1 in class on Wednesday, September 26.)
• Homework assignment 2 -- Questions related to Ladefoged (2005), Chapter 2 on "Pitch and Loudness".
  (Use these questions to guide your reading of the chapter and turn in your answers to the questions at the beginning of class on Wednesday, October 3.)
• McGuire & Beckman (2007) Chapter 2 -- Chapter on "Counting things over time".
  (We will go over the concepts in this chapter in class starting in Week 3. You can download the chapter and read it as a review of the concepts before we do quiz #2 in class on Wednesday, October 10.)
• Homework assignment 3 -- Questions related to Ladefoged (2005), Chapter 3 on "Vowel Contrasts" and Chapter 4 on "The Sounds of Vowels".
  (Use these questions to guide your reading of these two chapters and turn in your answers to the questions at the beginning of class on Monday, October 15.)
• Homework assignment 4 -- Questions related to Ladefoged (2005), Chapter 5 on "Charting Vowels".
  (Use these questions to guide your reading of the chapter and turn in your answers to the questions at the beginning of class on Wednesday, October 22.)
• Homework assignment 5 -- Questions related to Ladefoged (2005), Chapter 6 on "The Sounds of Consonants".
  (Use these questions to guide your reading of the chapter and turn in your answers to the questions at the beginning of class on Wednesday, November 7.)
• Homework assignment 6 -- Questions related to Ladefoged (2005), Chapter 9 on "Listening Computers".
  (Use these questions to guide your reading of the chapter and turn in your answers to the questions at the beginning of class on Monday, November 19.)

Data analysis reports and term project

The instructions and any data sets for the 5 reports from the data analysis exercises will be posted here as the they become ready. If a link does not work, that means that the assignment is not ready yet.

Data analysis reports

• Report 1 -- On "Counting words of different lengths"
  the associated wave file and an addendum
• Report 2 -- On "Pitch patterns and time plots"
• Report 3 -- On "Vowel inventories and vowel spaces"
• Report 4 -- On "Vowel differences and speaker differences"
  (Here is a link to the Central Limit Theory demo applet that we worked with in class on Monday, November 5. Here's a second demo with dice that does the same thing but with more explication. to see this second demo that more explicitly talks about the sampling mean.)
• Report 5 -- On "Word lengths and segment inventories"

Term project

The instructions and data sets for the 3 parts of the term project will be posted here as the they become ready. If an existing link does not work, that means that the assignment is not ready yet. There will also be links added as some parts become ready.

• Part 1 -- Instructions for doing the perception experiment and analyzing the results.
• Part 2 -- Instructions for making the recordings and measuring the vowel formants.
• Term project recordings -- a directory where the data files that we recorded on Monday, Oct. 1, were placed, for you to download and start analyzing.
• Term project measurements -- a directory where your formant measurement Table files were placed, for you to download and start analyzing. (There are also some files of R code there.)
• Part 3 -- Questions to answer as you relate the perception results from Part 1 to the vowel measures from Part 2.

Useful links

• You can download Praat off the Web.
• Here is the homepage to go to in order to get R, the "GNU S" software package for statistical analysis and graphics.
• Here is a link to a folder of more scripts that we have worked on together in this class.
• Sound files and video clips that accompany Peter Ladefoged's books Vowels and Consonants and A Course in Phonetics are available at the UCLA Phonetics Lab Archive.
• The home page for the International Phonetic Association also has a good deal of useful material, including an archive of recordings that go with the "language illustrations" in the Handbook of the International Phonetic Association.
• To insert IPA symbols into html code, you can use the Unicode encodings given at John Well's page on the IPA in Unicode. Alternatively, you can use an ASCII encoding of the IPA, such as SAMPA or WorldBet
• Here is a link to the directory where you can get the files and tcl/tk code for Choose Vowels experiment that we will use to get the perception data to analyze for the term project.
• Here is a link to the Detroiter vowel data that we looked at in class on October 31.
• A link to the article How to understand statistics by "Uncle Heavy" on the BBC homepage.
• A link to Alexei A. Sharov's wonderful lecture notes on "Models as analytical tools".
• Here is a web page that lets you change the skew of a histogram by painting a histogram with a mouse, and another one that lets you explore skew using a line drawing. The second one is on a page that also lets you look at various the normal distributions. Here is a web page that may help you appreciate the Central Limit Theorem.