Text segmentation for Language Identification in Greek Forums

In this paper, we examine the benefit of applying text segmentation methods to perform language identification in forums. The focus here is on forums containing a mixture of information written in Greek, English as well as Greeklish. Greeklish can be defined as the use of Latin alphabet for rendering Greek words with Latin characters. For the evaluation, a corpus was manually created by collecting web pages from Greek university forums and most specifically, pages containing information that combines Greek with English technical terminology and Greeklish. The evaluation using two well known text segmentation algorithms leads to the conclusion that despite the difficulty of the problem examined, text segmentation seems to be a promising solution.