Today

Term projects
Collostructions
Mid-term evaluation

Term projects

Term projects are scheduled as due to be presented on Dec 7th, and submitted on Dec 10th.

Appropriate topics

building a new, useful corpus of some language, language variety, register, etc.
investigating a question of theoretical interest (syntax, semantics, morphology, phonology, sociolinguistics, acquisition, discourse, etc.) using an existing corpus
norming stimuli (for experimentation, elicitation, etc.) using a corpus
using a corpus for an applied linguistic purpose, such as preparation of language instruction materials or lexicography

The term paper

(Since you're going to ask anyway:) It will probably be 10-20 pages. Much less or much more and I'll be suspicious.
It's can take the form of a research paper or a project description. In any case, you should include documentation of what you did, and a clear description of what it's good for.

You will also need to submit a term project proposal

1-3 pages
When should it be due? How about Nov. 1 (after the main linguistic content in the class)? Is this early enough? Too early?
Meet with me this week or next to talk about your project: My office hours: M 2-3, T10-11.

Questions

You can work on a project with someone else, but each of you will be required to put in as much work as you would on a singly-authored project. And it will have to be twice as good.

You can turn in your proposal early, but you're unlikely to get feedback before having to resubmit it unless you give it to me really soon.

Collostructions

There are significant associations among words that co-occur in a language - collocations. (We saw statistical means to measure these.) Are there significant associations between words and grammatical constructions?

Constructions: Meaning-bearing linguistic structures of any level of abstractness, whose meaning is not compositionally derived.

words (barbarian, mussel)
morphemes (-s, -ish)
fully-fixed expressions (tit for tat, Bob's your uncle)
variable idioms (be given to X, hoisted by X's own petard)
partially filled constructions (X Y's way Z, the Xer the Yer)
abstract grammatical constructions (ditransitive, transitive)

Collostructional analysis "starts with a particular construction and investigates which lexemes are strongly attracted or repelled by a particular slot in the construction (i.e. occur more frequently or less frequently than expected)" (p. 6)

collexemes: lexemes attracted to a particular slot in a construction
collostruct: construction associated with a particular lexeme

Methodology:

Start from a construction: define it, including slots, and search for n instances in the corpus
Pare these down (manually, in part) to include only those that are instances of the target construction
Calculate the independent frequencies of (1) the collexeme and (2) the collostruct, (3) their joint frequency, and (4) the total corpus size minus all instances of either

Perform a Fisher Exact test on these numbers, yielding a measure of the association between the collexeme and collostruct
Do this for all potential collexemes, and order them by strength of association.

Questions

In a collostructional analysis, you have to separate attracted and repelled collexemes. How and why do you do this?
Where do you draw the line for significant collexemes?
Footnote 3 says: "For the moment, we will only consider as repelled items those which do occur, but occur less frequently than expected, although it would of course also be possible to include items that should have occurred on statistical grounds, but did not." Explain this.
Do you really go through all the instances of the given construction by hand?
Do you have to run a separate Fisher Exact test for each collexeme? Isn't there a better way?