Linguistics 580:
Computational Methods in Linguistic Analysis

Spring 2012

Course Info

• Lecture: MW 10:30-11:50, Savery 169
• Text: Readings linked below
• For discussion, etc, we'll be using Canvas
• Lab space: PCs running linux are available for student use in the Linguistics Treehouse (PDL B-202-A). For access, take your Husky card to David Brodbeck (PDL B-5-L). Contact David: linghelp @ u.
• This webpage: courses.washington.edu/cmling/

Instructor Info

• Emily M. Bender
• Office Hours: Fridays 12-2 & by appt.
• Office: Padelford A-219
• Phone: +1 (206) 543-6914
• Email: ebender at uw
• Alicia Beckford Wassink
• Office Hours: Mondays 12:30-1:30pm & by appt.
• Office: Padelford A-217
• Phone: +1 (206) 616-9589
• Email: wassink at uw

Links

• Database of corpora installed on patas cluster
• our shared corpus list CorpusList.rtf
• The Sociolinguistics Lab's R tutorial materials

Syllabus

Description

Linguistic research across all subfields involves the testing of hypotheses against data. Computational methods allow linguists to greatly extend the size of datasets and range of data brought to bear on their hypotheses. The purpose of this course is to introduce students to existing computational resources including text and speech corpora, annotations over those corpora and software for manipulating them. The course is project-driven: The target audience is graduate students who are engaged in research projects which can be enhanced with the computational methods we study. No background in computational linguistics or computer programming is assumed.

Course goals

By the end of this course students will:

• be able to frame research questions such that linguistic databases can be brought to bear on them
• be more familiar with existing linguistic databases and how to access them
• understand how annotated databases are created and how to cope with variability in annotations and other noise in the data
• be familiar with existing NLP tools which can automatically add certain kinds of annotation to corpora and how to run them
• be able to write simple computer programs to find and count things in large datasets
• understand the issues (IRB, IPR, etc) that arise in making data reusable
• understand the connection between data sampling and inferential statistics as well as issues around choosing the correct statistical tests for a given study

Practicum groups

We will get "hands-on" experience working with large corpora of interest to students. Students (in pairs or small groups) will form practicum groups, and will work together to accomplish specific database-related tasks across the quarter. Most of this work will be "behind the scenes", and groups will independently arrange how to conduct their work. There are two planned points in the quarter when practicum work will be presented in class: each practicum group will be responsible one syllabus day (a Wednesday) in the first half of the quarter for doing an in-class presentation to introduce us to their chosen database, show us how it is annotated, demonstrate how it is navigated, and show us how it might be of use to linguists. Note: a list of database resources of interest to linguists, CorpusList.rtf is available online in the Sociolinguistics Wiki.

Ideally, each student will have a laptop available to use in class on Wednesdays when we do hands-on exercises.

1. Choose a database with which to work for the quarter.
2. Work together to learn how to access the contents of this database.
3. Present their database to the class, show how to access it, write one research question that this database may be used to address.
4. Individually annotate some portion of the database.
5. Measure inter-annotator agreement on this portion selected in step (4).

Disability accommodations

To request academic accommodations due to a disability, please contact Disabled Student Services, 448 Schmitz, 206-543-8924 (V/TTY). If you have a letter from Disabled Student Services indicating that you have a disability which requires academic accommodations, please present the letter to the instructor so we can discuss the accommodations you might need in this class.

Requirements

• Weekly problem sets: 45% Includes both practicum work and individual assignments. For the practicum work, students are encouraged to work on the problem sets collaboratively in their practicum groups, but answers should be written up and submitted individually
• Term project: 50% Term projects are expected to be done individually
• Reading questions: 5%
• Grades may be adjusted up to 2% for in-class or Canvas participation.

Late homework policy

• Unless you've made prior arrangements with us, homework turned in within one day of the due date will receive 80% credit, two days 70% credit. No credit after that, though we will still be willing to look it over and make comments.
• By prior arrangements, we mean contacting us no later than the day before the homework is due (i.e., Sunday for homework due Monday) with the reason you feel you can't complete your homework on time. At that time, we will decide whether or not to grant an extension, and for how long.
• No late final projects will be accepted.

Schedule of Topics and Assignments (may be updated)

Each week has a technical topic and a theoretical topic. We seek to build skills in linguistic analysis, drawing on some issue from linguistic theory (Mondays) as well as technical computing skills (Wednesdays). You will, therefore, see the days of each week in the schedule labelled either "linguistics" or "technical" depending on their focus. This schedule may change in light of the projects that students are working on.

Date	Topic	Reading(s) to be discussed this week	Due
Wk 1 3/26, 3/28	Linguistics: Intro; Projects of interest; What database resources exist Technical: Getting around in Unix		Corpus scavenger hunt
Wk 2 4/2, 4/4	Linguistics: Processing resources: POS taggers, parsers, forced aligners Technical: Running software	Bender & Langendoen 2010	Running a parser/POS tagger/forced aligner Presentation by Practicum Groups Term project: Define research questions
Wk 3 4/9, 4/11	Linguistics: Discussion of project questions and how corpora can be brought to bear on them Technical: Python basics, SVN	Bird et al 2009 Chapters 0 and 1 [R: Gries 2009, ch. 3]	Hello World Term project: Identify relevant resources
Wk 4 4/16, 4/18	Linguistics: Lexical frequency Technical: Python (cont)	Jurafsky (skip or skim pp 63-88) [R: Gries, 2009, ch. 4]	Simple word frequency counter in Python (run over multiple corpora) Term project: Identify relevant resources
Wk 5 4/23, 4/25	Linguistics: Metadata (DCMI, OLAC); Publishing data with papers Technical: Python continued		Merging demographic data with transcripts (Python exercise)
Wk 6 4/30, 5/2	Linguistics: Annotation: Inter-annotator agreement; Annotation guidelines Technical: Annotation software (xtdf; Excel; ELAN)	Bird & Liberman 1999 Morgan et al. (skip or skim Sec.5,pgs. 22-31)	Annotation (in pairs to produce dual-annotation)
Wk 7 5/7, 5/9	Interim project reports, open discussion of normalized frequencies for student databases
Wk 8 5/14, 5/16	Linguistics: Inter-annotator agreement Technical: Computing Cohen's Kappa and Krippendorff's Alpha	Artstien & Poesio 2008 Clopper 2011	Practicum groups compute Kappa/Alpha (Python)
Wk 9 5/21, 5/23	Linguistics: Sampling; Choosing statistical tests Technical: Basic exploratory statistics in R	Hay 2011	R assignment Term project: Choose statistical tests
Wk 10 5/30	No class Monday, Project presentations
6/6			Final projects due No late projects accepted.