Two Notions of Parsing

The term parsing, derived from Latin pars orationis (parts of speech), was originally used to denote the grammatical explication of sentences, as practiced in elementary schools. The term was later borrowed into computer science and linguistics, where it has acquired a specialized sense in connection with the theory of formal languages and grammars. However, in practical applications of natural language processing, the term is also used to denote the syntactic analysis of sentences in text, without reference to any particular formal grammar, a sense which is in many ways quite close to the original grammar school sense. In other words, there are at least two distinct notions of parsing that can be found in the current literature on natural language processing, notions that are not always clearly distinguished. I will call the two notions grammar parsing and text parsing, respectively. Although I am certainly not the first to notice this ambiguity, I feel that it may not have been given the attention that it deserves. While it is true that there are intimate connections between the two notions, they are nevertheless independent notions with quite different properties in some respects. In this paper I will try to pinpoint these differences through a comparative discussion of the two notions of parsing. This is motivated primarily by an interest in the problem of text parsing and a desire to understand how it is related to the more well-defined problem of grammar parsing. In a following companion paper I will go on to discuss different strategies for solving the text parsing problem, which may or may not involve