The Research Of Uyghur FrameNet Target Word Recognition Based On Conditional Random Fields

Uyghur frame target word recognition is the premise and foundation of the targetword frame identification and frame disambiguation, occupies an importantproportion in automatic labeling of semantic roles in Uyghur frame semantics.Compared with the target word recognition achievements in English and Chinese,Uyghur frame target word recognition is still in a preliminary study on the stage.Because of Uyghur distinctive syntactic and semantic features, simple transplantrelated technology into Uyghur frame semantic target word recognition is notavailable.This paper studied the method of Uyghur frame semantic target word recognition,analyzed the characteristics of Uyghur FrameNet semantic knowledge baseand features of target words, using conditional random field model of Uyghur frametarget word recognition are discussed.When identification, the recognition problem as a sequence labelingproblem to solve, namely with the statistical method for target word recognition. AsCRFs can not only make use of the context information as the characteristic, also canadd more complex features for training and inferencing. At present, It is naturallanguage processing fields using sequence comparison of ectensive annotation model.Therefore, use the conditional random field model to identify the target word. In thispaper,such the word form, part of speech, word stem as features, different featureselection of different window sizes, to identify the target word to formulate anoptimal template. Experimental results show that, using conditional randomfields model to identify the target words, effectiveness of these features isalso good, the F value reached86.17%.