Reviews

Syntactic pattern recognition: an introduction
by Rafael C. Gonzales and Michael G. Thomason, 1978
in Current Science, Gift Siromoney

There are two main approaches to the area of pattern recognition. The first is the decision- theoretic approach which utilizes decision functions for classifying pattern vectors. It is based on classical statistical theory. The second called the syntactic approach was proposed by Professor R. Narasimhan of the Tata Institute of Fundamental Research, Bombay and the book under review is primarily about this second approach. This new approach to pattern recognition is also referred to as linguistic or grammatical pattern recognition and this approach possesses structure-handling capability lacked by the decision theoretic approach.

Syntactic pattern recognition relies heavily on two branches of computer mathematics, viz., automata and formal language theories. The book deals with both these theories at an introductory level.

Formal language theory deals with different kinds of grammars that generate specific kinds of patterns in the form of strings of symbols. Basic ideas from one-dimensional formal languages have been suitably modified to generate two dimensional patterns.

Automata theory deals with string language acceptors of different kinds of capabilities. The book deals with some of the simpler models that have been found to be useful in the area of pattern recognition.

A whole chapter is devoted to higher dimensional grammars such as Tree grammars, Plex grammars and Rosenfeld's model of web grammars. Grammatical models incorporating probability are dealt with in a separate chapter.

Given a set of patterns, is it possible to find a grammar that generates the patterns? The chapter on grammatical inference deals with this problem. Now that we have Iravatham Mahadevan's concordance of the Indus script, it should be possible to make use of grammatical inference techniques to find grammars for the set of known strings of symbols of the Indus script.

The book will serve as a good introductory text in the area of syntactic pattern recognition.