Paper/project plan

Due 2/4, in class or by email before class

This document is a proposal for what you plan to do for your term paper or term project. It should be a page long, and address the following:

1. What research question are you asking?
2. How does that question relate to the course topic?
3. What methods will you use to answer your question?
4. Which (min 2-3) of the course assigned readings do you plan to include as background?
5. Which other references do you plan to include (min 3)?

The purpose of this document is a) to get you thinking about your term paper/project relatively early in the quarter and b) to give me a chance to give you feedback relatively early in the process. If you would like to turn this document in earlier, I will do my best to get you feedback earlier.

General ideas for term papers:

1. Analysis of topic/focus/information structure and its expression in some language.
2. Translation of analysis from article to another framework and exploration of consequences.
3. Application of analysis from article to one or more other languages.

General ideas for term projects:

1. Annotate small portion of e.g., PTB for information structure using analysis from some paper. Develop initial guidelines, have another annotator use them and measure inter-annotator agreement. What further information needed to be added not available in paper?
2. Apply annotation scheme from existing work (cf. workshop at LSA Jan 2010) to new data. Explore how the annotation scheme relates to the concepts of topic and focus as developed in the literature we read in class.
3. If you can find a sufficient amount of annotated data, explore whether/how joint learning of information structure with some other kind of annotation can improve performance. (Candidate other kinds of annotation include coreference chains, named entity recognition, and topic boundaries.)
4. Evaluate MT in terms of information structure. Take existing system and see where the information structure translation is off. Is this reflected in BLEU score?
5. such as Question Answering. Assuming oracle marking of information structure. Can this improve the performance of an existing algorithm on this task?