European University of Lefke (EUL) Computer Engineering Faculty Member Asst. Prof. Dr. Zafer Erenel made statements on the subject of “Classification of Documents in Electronic Environment”.

Erenel: The number of documents on the internet has been increasing rapidly in the last decade

“Considering the past ten years, it is seen that the number of documents in the internet environment has been increasing rapidly” said Erenel, who said that billions of users from 7 to 70 have reached this environment with personal computers, smart phones and tablets. He stated that it is natural and some problems that arise with many documents need to be solved. Erenel stated that one of these problems is the problem of classifying the texts in the electronic environment and said, “The text classification problem means determining which or which of the predetermined classes a text falls into. Such classifications can produce very efficient solutions even on the pages of popular search engines. As of today, if we look at the page of the www.yahoo.com search engine, many documents are grouped in a tree structure under the titles such as 2020 America’s Choice, News, Finance, Sports, Coronavirus, Entertainment, Life, Shopping. Apart from search engines, to distinguish unsolicited e-mail messages from normal e-mails, automatic determination of the language of a text, determination of suitable materials for different age groups or types of readers, sentiment analysis, determination of a speaker or author’s attitude towards a subject, text classification they are usage areas ”he said.

Erenel: Vector machines are often used to classify text

Erenel pointed out that the classifiers used quite frequently to classify text in the 1990s and early 2000s are support vector machines, and recently the learning of word representations that include the data of how often words occur together to establish a semantic or conceptual relationship between words has come to the fore. “These representations have succeeded in producing performance on support vector machines by using convolutional neural networks under the title of deep learning,” Erenel said.