MIME-Version: 1.0
Server: CERN/3.0
Date: Monday, 16-Dec-96 23:12:34 GMT
Content-Type: text/html
Content-Length: 9413
Last-Modified: Friday, 12-Apr-96 22:27:57 GMT
What to Turn in for the Project
Official due date: Friday 5/3. We'll accept projects without penalty
until 5:00, Friday 5/10.
Programming projects:
Final Writeup for Programming Projects (just a few pages):
- Problem description
- Description of general approach
- Description and results of evaluation
- Discussion. (what worked; what didn't work; options you'd
like to have tried; analysis of the results; etc.)
- I'd also like to see any code that you wrote and a few short
traces of the system in action (if that makes sense for your project).
In-Class Presentation: ~7 minutes in length. Should include
an overview of the problem; your solution; evaluation method and
results.
Non-Programming projects:
Final Writeup for Non-programming Projects. This will vary for
each person, but in general, the final writeup for non-programming
projects will probably contain a description of the problem that you
looked at; a summary of the papers that you read; a critique of the
existing approaches; your attempt at an evaluation of the the
theory/algorithms presented in the papers on real text.
In-Class Presentation: For non-programming projects, the
in-class presentation should be a synopsis of what you'll include in
the paper. You'll have to leave a lot out of course...
Programming Projects from Last Year's Class
- Grammar Induction using Genetic Algorithms
This project implements a grammar generator using genetic algorithms.
- AutoSlog and FALCON: Automated Lexicon Construction
This paper evaluates the performance of AutoSlog and FALCON as lexicon
constructors.
- A Preposition Attacher
This project uses a matching algorithm for preposition attachment.
- Statistical Word Sense Disambiguation
This project implements a number of related statistical methods for word-sense
disambiguation and evaluates the method using examples from the Brown corpus.
- Automated English-Esperanto Translation
This project implements a limited English-Esperanto translator by extending
James Allen's bottom-up parser.
- An Implementation of a Method for Word Sense Disambiguation
This project implements a relaxation network approach (like CIRCUS's) for word-sense
disambiguation. It is based on: Large Neural Networks
for the Resolution of Lexical Ambiguity by J. Veronis and N. Ide.
- The Viterbi Algorithm and Smoothing
This project is an implementation of a part-of-speech tagger using a Hidden
Markov Model, a smoothed bigram language model, and the Viterbi algorithm.
Reading Projects from Last Year's Class
- Topics in Intelligent Multimedia/Multimodal Interface
This paper examines two existing systems that incorporate intelligent
multimedia/multinodal interfaces (WIP and COMET).
- Using FOIL for WH-Phrase Disambiguation
This paper examines ways to use FOIL for WH-phrase disambiguation. (FOIL is a
supervised learning algorithm for acquiring concepts in first-order logic.)
- Natural Language Understanding of the Japanese Language
This paper summarizes issues for understanding Japanese.
- Approaches to Automated Question Answering
This paper addresses issues in understanding and replying to questions given
in English.
- Word Boundary Detection in Continuous Speech Recognition: Application of
Phonological Constraints
This paper looks at three models that use broad class, phonemic sequence, and
allophonic constraints to facilitate word boundary detection.
- Computer Understanding of Conventional Metaphoric Language
This paper looks at MIDAS (Metaphor Interpretation, Denotation and Acquisition System)
for understanding conventional metaphor.
- An Analysis of Transformation-Based Part of Speech Tagging
This paper critiques the an article on transformation-based
part of speech tagging.
- Neural Networks and NLU
This paper looks at neural network applications for natural language understanding.
- Pronominal Anaphora Resolution
This paper describes four approaches to the computational task of resolving
the pronoun referents.
- Harmonic Grammars and Unaccusativity
This paper looks at Harmonic Grammars developed by Smolensky, Legendre and
Miyata.