[wiki:HlpLab/LSA09/Syllabus Syllabus] | [wiki:HlpLab/LSA09/Assignments Assignments] | [wiki:HlpLab/LSA09/People People] | [wiki:HlpLab/LSA09/CorporaTutorials Corpora & Tutorials] | [wiki:HlpLab/LSA09/References Readings] | [http://lsa2009.berkeley.edu/courses/lsa125.html Offical LSA course page]

Syllabus

The class has three themes:

To introduce you to psycholinguistic theories of variation in production. What drives speakers' choices between different word orders (e.g. in the ditransitive alternation), different morphological variants (e.g. contracted vs. full auxiliaries)? Do they avoid potential ambiguities or try to ease production or comprehension processing in some other form?

To provide you with an introduction to at least one syntactic search software (TGrep2) and tools we developed to combine output of this search software into a database that you can import into excel or a statistics program (these tools are in the developmental stage and have not been release previously). We also plan to give you a super fast intro to the basics of the statistical methods employed in modern corpus-based research on language production.

To provide a forum for a discussion of the relative trade-offs of corpus-based vs. experiment-based psycholinguistic research. What does corpus-based work to offer? What are the current limitations?

This is an introductory class. I was asked by the LSA to provide a basic introduction and since many of you have not worked with corpora before, we will probably not get into the more advanced statistical aspects of corpus-based work, but I hope that we can give you enough of a background to get you started on corpus-based work and to connect to the community of researchers that is doing this type of work.

Date	Title	Description/Notes	Readings	Assignment due
7/7/2009	Psycholinguistic accounts of processing an communicative efficiency	Availability, Alignment, Domain Minimization, Ambiguity Avoidance, and Uniform Information Density	Required reading: Branigan et al. (2007), Hawkins (2007), Haywood et al. (2005), Frank and Jaeger (2008), Optional: Jaeger and Norcliffe (in press), , Yamashita and Chang (2001), Arnold et al. (2000), Choi (1997), Gildea and Temperley (2008), Arnold et al. (2004),Kraljic and Brennan (2005), Levy and Jaeger (2007), Jaeger (submitted)	12pm, Fr, 7/10/09
7/8/2009		You are highly encouraged to attend the [http://www.hlplab.wordpress.com/lsa09regression/ evening tutorial on logistic regression] (Note: different time & location: TBA, 7-9pm)
7/9/2009	An introduction to TGrep2: a syntactic search software	laptop highly company encouraged
7/14/2009	The relation between psycholinguistic and other approaches	more on psycholinguistic accounts; also grammaticalization, semantic accounts, sociolinguistic accounts	Required reading: Tagliamonte et al. (2005) OR Tagliamonte and Smith (2005), Bresnan and Hay (2007) Optional: Torres Calcoullos and Walker (2009), Bresnan et al. (2007)	12pm, Fr, 7/17/09
7/15/2009		If you're interested in the analysis of continuous corpus outcomes such as word durations, pitch accents, etc., consider attending the [http://www.hlplab.wordpress.com/lsa09regression/ evening tutorial on linear mixed models] (Note: different time & location: TBA, 7-9pm)
7/16/2009	An introduction to logistic regression in R	laptop highly company encouraged	Required reading: Baayen et al. (2008), Jaeger (2008) Optional: Johnson (2009), Bresnan et al. (2007)
7/21/2009	What to do if there is (seems to be) no right corpus for your purpose? Don't panic!	We discuss the process of selecting a corpus, trade-off between more available annotation and more available data, adding additional annotation, using text-corpora for syntactic searches, using the web as a corpus, and how to further enrich your corpus data base (e.g. with ngram information) .... bascially, we'll go through a bag of tricks. We will also use this class to catch up with left-overs from previous classes.		12pm, Wed, 7/22/09
7/23/2009	Corpora vs. Experiments: Trade-offs (... and a bit more statistics)	We'll discuss the advantages and disadvantages of corpora compared to running experiments in the lab. This will lead us to look a bit more at mixed models which provide control for individual speaker effects which make the analysis of corpus data more complicated. I hope that most of you can attend the evening tutorials, so that we can use this session to go through more advanced questions and for questions you have regarding your final projects.