유한상태 오토마타를 이용한 수식 표현 인식

Abstract

There have been many trials to parse sentences in text to search complete and exact parses, but it is very hard because of unavoidable incompleteness of lexicon and grammar. Recently, to alleviate these difficulties, partial parsing appears as an alternative in the field. Partial parsing aims to recover syntactic information efficiently and reliably from unrestricted text, by sacrificing completeness and depth of analysis.

As a part of partial parsing, the identification of Korean numerical expressions in text is described in this paper. Numerical expressions are required in several systems such as information extraction systems and question-answering systems. One of desired characteristics of these systems is the fastness. To achieve this goal, we use a finite-state automaton, for which we could use a tool like lex. So that we could rapidly implement the system. We observed that the system is fast and correct through several experiments. To evaluate our system, we used newspaper as test collection. We achieved the recall of 90.8%, and the precision of 86.9%. Experiments show that our system is comparatively correct.