A System For Health Document Classification Using Machine Learning

2.2.2    STEMMING
In linguistic morphology and information collection, stemming is the process for decreasing deviated (or sometimes derived) words to their stem, original form. The stem need not be identical to the morphological root of the word; it is usually enough if it is concern words map of similar stem, even if this stem is not a valid root. In computer science algorithms for stemming have been studied since 1968. Many search engines consider words with the similar stem as synonyms as a kind of query broadening, a process called conflation.
2.2.3    STOP WORD REMOVAL
Typically in computing, stop words are filtered out prior to the processing of natural language data (text) which is managed by man but not a machine. A prepared list of stop words do not exist which can be used by every tool. Though any stop word list is used by any tool in order to support the phrase search the list is ignored.
Any group of words can be selected as the stop words for a particular cause. For a few search machines, these is a list of common words, short function words, like the, is, at, which and on that create problems in performing text mining phrases that consist them. Therefore it is needed to eliminate stop words contains lexical words, like "want" from phrases to raise performance.
2.2.4    VECTOR REPRESENTATION OF THE DOCUMENTS
Vector denotation of the documents is an algebraic model for symbolizing text documents (and any objects, in general) as vectors of identifiers, like, for example, index terms which will be utilized in information filtering, information retrieval, indexing and relevancy rankings where its primary use is in the SMART Information Retrieval System.
A sequence of words is called a document (Leopold, 2002). Thus every document is generally denoted by an array of words. The group of all the words of a training group is called vocabulary, or feature set. Thus a document can be produced by a binary vector, assigning the value 1 if the document includes the feature-word or 0 if there is no word in the document.
2.2.5    FEATURE SELECTION AND TRANSFORMATION
The main objective of feature-selection methods is to decrease of the dimensionality of the dataset by eliminating features that are not related for the classification (Forman, 2003). The transformation procedure is explained for presenting a number of benefits, involving tiny dataset size, tiny computational needs for the text categorization algorithms (especially those that do not scale well with the feature set size) and comfortable shrinking of the search space. The goal is to reduce the curse of dimensionality to yield developed classification perfection. The other advantage of feature selection is its quality to decrease over fitting, i. e. the phenomenon by which a classifier is tuned also to the contingent characteristics   of   the   training   data   rather   than   the   constitutive characteristics of the categories, and therefore, to augment generalization.
Feature Transformation differs considerably from Feature Selection approaches, but like them its aim is to decrease the feature set volume. The approach does not weight terms in order to neglect the lower weighted but compacts the vocabulary based on feature concurrencies.