Computational detection of Uyghur multiword expressions

Abstract

This paper describes some important properties of Uyghur multiword expressions (MWEs). MWEs are significantly difficult to handle in any languages for natural language processing. There are many sub-fields in natural language studies such as automatic word correction, sentence correction, sentence fragmentation, machine translation, bilingual MWEs alignment and text mining etc. In all these fields, it is very important to detect a word in a sentence correctly. By their structure, MWEs are very complex and difficult to be defined or identified properly. In general, MWEs consists of more than one word or phrases. In some cases, these words that create a MWE are written together as a single (compound) word, however sometimes these words are written separately as completely different words (non-compound) each other. In this paper, Uyghur MWEs are analyzed by their structure and some practical approaches suggested detecting them effectively at the initial stage. End of this paper, these approaches evaluated some scientific results and given some general for future studies.