ESSAY/PROJECT GUIDELINES September 23, 2010 It's time for 447 students to get underway with the project (as sketched above). I would like to see a project title and approx. half-page abstract, plus some references, no later than Friday, Nov. 5 (by email to LS, please). This presupposes that in the prior weeks you will have spent some time looking through the course text and references and searching the web and library for topics and papers or books that might interest you, and doing enough reading to acquire a sense of what the topic you are choosing is all about, and what the most relevant references are. Feel free to talk to me along the way, about possible titles, feasibility, scope, possible references, etc. A typical project is an EXPOSITION of, with critical COMMENTARY and your own SUGGESTIONS on, some (fairly recent) articles from the research literature on a subtopic touched on in this course. The length is likely to be between 3500 and 7000 words, and the number of articles might range from two to half a dozen, depending on their technical depth. (Of course, you may also have various additional citations of sources that you made less use of but that are relevant.) The critical commentary can be mingled with the exposition and/or follow it, but BE CLEAR WHAT IS TAKEN FROM THE REFERENCES (USING PROPER CITATIONS) AND WHAT ARE YOUR OWN THOUGHTS. Although it's not usual in scientific writing, in order to get credit for your own ideas feel free to use phrases like "I think that ..."; if you write impersonally, it won't be clear where you're presenting your own ideas and where you're summarizing/paraphrasing what you've read. (I'll assume ideas are not your own unless otherwise indicated.) Also, be sure that when you take material verbatim from some source, put it in quotes and cite the source, using conventional scientific citation style. Projects that involve programming rather than pure essay-writing are equally welcome (especially if using Lisp code). For such projects, less emphasis on background literature would be expected in the write-up, though you should still sketch a state-of-the-art framework for your effort, and of course use good code description/ documentation/commenting practices. The code developed might be an extension of something that exists (here in the department or elsewhere), or it might be entirely new. IN THE EXPOSITION, TRY TO CONVEY THE GIST OF THE IDEAS AS MUCH AS POSSIBLE THROUGH SPECIFIC EXAMPLES, rather then vague high-level descriptions. A carefully worked example is worth a thousand generalities. Grading will be based on (a) substantiveness of your effort; (b) clarity and coherence of the written presentation; and (c) originality, in the sense of showing an independent, informed perspective on the subject matter, and coming up with some specific ideas of your own, at least on how the work studied could be taken further or applied, or done differently and better. There are some trade-offs between the extent of literature coverage on the one hand and the amount of original work and commentary on the other. In other words, some projects might consist primarily of a small survey (with a personal perspective and commentary), while others might be more in the nature of a small research project, possibly (but not necessarily) including some programming/experiments. SOME POSSIBLE TOPICS: integrating parsing with interpretation; computational approaches to "situated" language learning; building knowledge bases for lexical meaning; learning word meanings from corpora or lexicons; induction of general world knowledge from text corpora; ontologies for NLU; paraphrasing and linguistic entailment/implicature; kinds of inference (e.g., abductive) needed for language understanding; determining referents of referring expressions (using semantic criteria); interpreting noun-noun compounds; interpreting tense and aspect; ellipsis; vagueness (and computational implications); metaphor (and computational implications); scoping quantifiers and other operators; biologically/psycholinguistically plausible approaches to language processing (e.g., parsing or disambiguation); aspects of language generation; morphological analysis (with semantic implications); speech act or plan recognition in NLU; etc. (The text, including chapters not covered yet, provides starting points for many other topics.) Other sources: the texts on reserve, various journals and conferences (Computational Linguistics, ACL, NAACL, EACL and COLING conferences, and many specialized workshops). There are also some useful encyclopedias and handbooks (e.g., Oxford Handbook of Computational Linguistics, the Handbook of Natural Language Processing, the Handbook of Logic and Language), and collections of readings. Of course, nowadays quick access to lots of online materials is possible, and good overviews on many topics exist in Wikipedia and on the AAAI website, and other AI/computational linguistics/NLP websites. But do not rely primarily on these. USE REFEREED, PUBLISHED WORK AS YOU PRIMARY SOURCES. Pick something that looks interesting to you! Best wishes, -Len