Emerging Database course Grammar Models Grammar Models For

Emerging Database course: Grammar Models

Grammar Models For those types of strings which represent a part of any language with strong syntactic rules, the above models are not usable. For example, in computer al gebra(like MAPLE or MATHEMATICA) if the parentheses are balanced in a formula, the next character may not be a")". Similarly, for those of computer languages where the delimiter is a special character (e. g. , ". " or "; ") this char actermay not occur "inside" a statement. For the above type of strings with theoretically-infinite nesting possibilities, the so-called grammar models are adequate. In a grammar model we have well-defined productions (e. g. , condi tionalor unconditional jump statements and subroutine calling sequences), and we can decide the probabilities of each sub-production in a given production. This gives us a multilevel hierarchy, containing in each level a discrete proba bilitydistribution.

Now, if we have a string to be modeled by a grammar model it is parsed according to the grammar, and its probability may be computed by multiplying together the probability of each production that is used. The use of grammars-principally for compressing Pascal programs-has been explored by Katajainen eta/. [39] and Cameron [14]. Compressing strings from a formal-computer-language, this type of model is very successful, but it is almost usable for natural languages. It follows from pragmatic reasons that it is almost impossible to find a grammar for natural languages: their syntactic rules are very complicated and are not modelable "by hand. " Constructing them mechanically from samples is also impossible because we cannot decide the exact boundary of the given language. It is worth mentioning that some heuristics are useful for modeling natural languages.

REFERENCES • Timon C. Du. , Emerging Database System Architectures • Bochmann, G. Concepts for Distributed Systems Design • Capron, H. L. Computers: Tools for an Information Age