Course Overview

1
Course Overview

• Ling 571
• Week 1 9/29/05

2
Outline

• Syllabus
• Parsing
• CFG
• Parse tree
• Penn treebank
• Evaluation metrics
• Quick intro to conditional probability
• Misc

3
Syllabus

• General info office hour, text books, website
• Prerequisites
• C/C or Java, Perl, Unix/Linux shell
• Linguistics
• Probability
• Grading
• Content
• EPost and ESubmit
• Overview of the projects

4
Context-free grammar

• Definition of CFG (see Page 331 in JM)
• Examples
• S ? NP VP
• NP ? Det N
• NP ? N
• VP ? V NP
• N ? John / book
• Det ? a / the
• V ? bought

5
Parse tree

• John bought a book.
• Creating the parse tree using top-down and
  bottom-up methods.

6
Penn treebank annotation

• (S (NP (N John))
• (VP (V bought)
• (NP (Det a)
• (N book) ) ) )

7
Penn treebank tagset

• POS tags see P297 in JM
• Syntactic tags e.g., NP, VP
• Empty category (EC) e.g., trace (T)
• Function tags e.g., SBJ, OBJ, TMP,
• ? In this course, we use a simplified tagset.

8
Evaluation metrics for parsing

• Precision
• Recall
• F-measure
• A of correct brackets in the system output
• B of brackets in the system output
• C of brackets in the gold standard

9
Quick intro to conditional prob

• P(AB) the prob of A given B
• Example
• The car killed the duck
• P(w iduck wi-1the) 1/2
• P(wi-1the w iduck) 1

10
Conditional Prob (cont)

• (1) The/Det car/N killed/V the/Det duck/N
• (2) The/Det duck/N died/V under/P her/Pron car/N
• P(w2duck w1the) 2/3
• P(wduck t N) 2/4 1/2
• P(t2N w2duck) 2/2 1
• P(t2N t1Det) 3/3 1

11
Misc

• Homework 1 due on 10/6
• Pongo account ssh
• LING 570 questionnaire
• Registration