Hopefully Final Proposal

Outline

• Plan Project (total 30% of Plan):
  • Obtain many (200+) samples of English text, mainly from Marlboro community members
    • Questionnaire – see page 2.
  • Put through Computer Aided Translation (CAT) program (Google Translate, Yahoo! Babel Fish, Bing Translator)
    • Using languages for which I can find near-native speakers (French… (and maybe one more – Italian?))
  • Compare CAT to human translation
    • Why the difference?
  • How does the CAT program do the translation? Introduce each and pros/cons.
    • Statistical? – (eg. Google)
      • Bayesian probabilities (Phrase-based translation)
    • Rule-based? – (eg. Systran, Yahoo! Babelfish)
      • (Dictionaries/grammars)
    • Hybrid? – (eg. Yahoo! Babelfish, Bing Translator)
      • What’s the algorithm?
    • Something else? Are there any of these? I think so.
  • Linguistic perspective: Look at syntax, tree structures
    • Discuss: What’s going on? Ambiguities, if applicable.
    • Evaluation
  • Computer Science perspective: Bayesian probabilities
    • Discuss: What’s going on?
    • How are they being used?
    • Evaluation

• “Other Plan Components” (total 40% of Plan):
  • Two papers:
    • One on Phonology and its significance in language.
    • One on Neural Networks, with toy code and other examples.

• Independent Portion (total 30% of Plan):
  • Examinations:
    • In Bayesian probabilities, Markov models, and Artificial Intelligence (15%)
    • In syntax and grammar. (15%)

Stuff to fill out on FPA

1. Plan Summary:
   "A study of Syntax and Bayesian probability with respect to Natural Language Processing and Machine Translation."