Extracting document features to improve classification and clustering
Péter Schönhofen · 2008
ion, applied sciences, arts, belief, culture, events, geography, health, health science, history, mathematics, nature, natural sciences, people, personal life, philosophy, religion, science, self, self-care, social sciences, society, surnames, technology, thought Category sizes are wildly varying, some comprise of a single article, others carrying more than 1000, although there is a constant effort to split large categories into more manageable ones. In addition to categories, there are also so called portals, on one hand classifying the above main categories into twelve groups (for example joining “Health”, “Self-care” and “Health science” under the title “Health and fitness”), and on the other hand providing a short list of the most popular and important sub-categories. Let us see a document about the chemical compound pantopon, in its original Wiki markup form: ’’’Pantopon’’’ is a preparation of [[opiates]] made up of all of the [[alkaloids]] present in [[opium]] in their natural proportions as [[hydrochloride]] salts. It can sometimes be tolerated by persons who are allergic to [[morphine]]. Pantopon is just [[opium]] with all of the tar and other insoluables removed in an injectable form which is nearly as potent as [[morphine]] is by weight. Other drugs of the same type have include opium alkaloid sulphates (Omnopon), phosphates, and valerates. ==References== *[[Compendium of Pharmaceuticals and Specialties]] {{pharma-stub}} [[Category:Opioids]] Wikipedia is an optimal choice in almost all aspects: its text reflects current language usage, contains a large number of documents (approximately 1,600,000 at the start of 2007), documents are well written, are richly linked together, and cover almost every conceivable topic, even the most recent events, persons, technical devices. However, it also has its disadvantages: categorization and linkage of articles is sometimes inconsistent or wrong, the classification system is ad-hoc, does not follow any established