Coreference Resolution Of Uyghur Noun Phrase

Anaphoricity refers to the process of seeking for anaphora to substitute indicative pronoun in the chapter.It consists of two subtasks: 1)Anaphoricity Determination: anaphoricity determination refers to the prophase work of coreference determination.Based on the identified language units: personal pronouns,noun phrases,entity zero anaphora item,further determine which language unit is anaphoricity item and which one is not in the Anaphroa.2)Coreference Determination: coreference determination refers to the process of seeking for anaphora to substitute indicative pronoun in the text.The types of anaphora are divided into the following: noun phrase coreference determination,the pronoun coreference determination and zero-style coreference determination according to anaphora objects.Most of current studies on coreference determination focus on English and Chinese with rich corpus resources,and achieved fruitful results.However,there are few studies about the language with rare corpus resources like Uyghur.Therefore,the paper makes the following researches about noun phrase coreference determination in Uyghur language.(1)Under the guidance of Uyghur language experts,according to the characteristics and noun phrase categories of Uyghur language,the paper extracts the characteristic vector of Uyghur noun phrase.15 characteristics are extracted as characteristic vector of anaphoricity determination.With the strong text learning ability of deep learning in natural language processing and the characteristics of the extracted deep semantic information,this paper uses SAE(Staked Autoencoder Method)to recognize the anaphoricity item of Uighur noun phrases so as to test the effectiveness of deep learning.The nonnegative constraints weight is introduced on the basis of autoencoder method in order to build SNCAE(Staked Nonnegative Constrained Autoencoder Method).The paper uses SNCAE to complete the task of anaphoricity determination of Uighur noun phrases.(2)Under the guidance of Uyghur linguistic experts,by analyzing the phenomenon of Uyghur nominal noun phrases,this paper summarizes the five nominal noun phrases and extracts the eigenvectors of Uyghur nouns.At the same time,the word vector with vocabularies semantic and contextual position is introduced into eigenvector,and the unidentified items are added before the test samples are generated.Using the staked nonnegative constrained autoencoder method(SNCAE)extract the deep semantic features to complete the Uyghur noun phrase coreference determination based on deep semantic and syntactic information.