Method of Statistical classification normally depends on a bets single model to predict in an accurate way. Model of this type is designed to get maximum accuracy by balance of recall and precision. The performance of method of Model Switching which has been discussed in the paper shows good performance with higher accuracy of prediction and recall rate of 100 per cent through usage of a set of models that can be decomposed in the place of single model. MS1, the system of implementation, is examined in a case study that predicts Prepositional Phrase Attachment (PPA). Its outcomes illustrate the accuracy of IV better than other methods of statistics so as to choose single statistical classification models and that are better with other successful approaches of NLP in disambiguation of PPA. The method of Model Switching might be preferred on other methods due to its generalization (i.e., its applicability in wider range), and its comparatively better prediction accuracy. It might be applied as a tool for analysis as well to examine the domain nature and the data characteristics with the assistance of models that are being generated.
The classical definition of decision problems is “those difficulties whose answers comes in one of the two categories: No or Yes (Garey and Johnson, 1979). Another problem class, the optimization problems, are those problems that minimize or maximize some worth; though, they may also be casted as decision problem (Cormen, et. al., 1990). In problems of classification, both the traits are incorporated: A problem of classification is a problem of decisions, in which decision making takes place (selection of a class is made), which increases a function of utility (Neumann.Y and Morgenstern. O 1953). The method of Switching of Model, as suggested in this study can is applicable by means of any kind of function of utility (criterion of decision) for any kind of problem regarding decision with data categories, which is symbolized like a tuple (C, F1, F2, ..., Fn) of a class C variable and a number of variables of features F {1 .... }.
Resolution of problem of PPA is a common issue in any system of NLP that treats with text understanding or syntactic parsing. The classifier of Naive Bayes and learning system of leading machine, such as C-4.5, CN-2 & problem of PEBLS, failed to give forecast with accurateness rates of competition on the problem. A sentence or phrase can observed to be so confusing that it might not be likely to identify the right connection devoid of extra information of context. (Ratnaparkhi.A, et al., 1994) stated that experts of human can attain 93 per cent accuracy, if out of context, the cases are provided as entire sentences.
In the study, we deal with just the kind of PPA problem that have been demonstrated above and do not take the other PPA problems of PPA under consideration, which are less frequent. For the details of problem linguist, the person who reads can get help ...