Mitsubishi Electric Research Laboratories

Introduction to Finite-State Devices in Natural Language Processing

MERL Report:  TR96-13
Where Published: Finite-State Devices for Natural Language Processing, Roche and Schabes (Editors), MIT Press

The theory of finite-state automata (FSA) is rich and finite-state automata techniques have been used in a wide range of domains, such as switching theory, pattern matching, pattern recognition, speech processing, hand writing recognition, optical character recognition, encryption algorithm, data compression, indexing and operating system analysis (Petri-net). In this chapter, we describe the basic notions of finite-state automata and finite-state transducers. We also describe the fundamental properties of these machines while illustrating their use. We give simple formal language examples as well as natural language examples. We also illustrate some of the main algorithms used with finite-state automata and transducers.

 Read the full technical report (PDF: 511.7 kB)