Table of Contents
Finite-State NLP - Lecture 1 Motivation and Foundations
Computational Linguistics
Some clean reusable math ...
OUTLINE
Finite state machines
Finite state machines
Function from strings to ...
Sample functions
A warning!
Functions and relations
Functions and relations
Functions and relations
Example: Unweighted acceptor
Example: Weighted acceptor
Example: Weighted acceptor
Example: Unweighted transducer
Building a lexical transducer
Finite-state “programming”
Finite-state “programming”
Ambiguities
Weighted version of transducer: Assigns a weight to each string pair
OUTLINE
Statistics and Language
More Weighted Transducers: Part-of-Speech Tagging
More Weighted Transducers: Part-of-Speech Tagging
Markov Model
Markov Model
Markov Model
Markov Model = Weighted FSA
Hidden Markov Model
Hidden Markov Model = WFST
Finite-state “programming”
OUTLINE
Hand-Coded Example: Parsing Dates
Source code: Language of Dates
Object code: All Dates from 1.1.1 to 31.12.9999
Parser for Dates
Problem of Reference
Refinement by Intersection
Defining Valid Dates
Parser for Valid and Invalid Dates
Observations
Historical Context
Some Reasons
More Reasons
OUTLINE
Xerox Regular Expression Calculus
Symbols
Common Regular Expression Operators
Xerox Extensions
Containment
Restriction
Replacement
Marking
Directed Replace Operators
@-> Left-to-right, Longest-match Replacement
Syllabification
Conditional Replacement
Merge Operators
OUTLINE
How to define transducers?
How to implement basic operators on transducers?
What are the “basic” transducers?
OUTLINE
Function from strings to ...
Weight semiring
Semiring
OUTLINE
|