A Study On The Construction And Dependency Analysis Of Naxi Language Dependent Trees

Naxi language dependency parsing is one of the most important basic wor k in Naxi language information processing,The purpose is based on the depend ency grammar to derive grammatical structure of Naxi sentences.Parsing results can affecting machine translation.Nowadays,Naxi Dependency Treebank resourc es is scarcity,Labor Standards is difficult,so,firstly,two different ways to constru ction of Naxi Dependency Treebank; secondly,According generated treebank to b uild dependency parser:(1)As corpus of Naxi language is rare, Marked difficulty, compared to mat ure Chinese corpus,its Syntactic Analysis is much more difficult. Therefore, we propose to begin the experiment with the Chinese-Naxi Languages align relati on-ship.firstly,we solve aligned word processing by Chinese-Naxi sentence pairs; secondly,we do the dependency parsing with Chinese sentences; finally,Naxi D ependency Treebank generated by Chinese-Naxi Languages align relationship an d Chinese Dependency Tree. Experimental results show that:This approach can simplify the process of manual collection and annotation of Naxi Dependency Treebank,save manpower and time to build the treebank of Naxi Language.(2)Combined with the characteristics of Co-training algorithm, firstly,mark small samples;Secondly,Building a weak learner by a small amount of labeled samples with two redundant views;lastly,through a large number of unlabeled s amples to improve the learner,use of learner to build Naxi Dependency Treeba nk with15K Naxi sentence.Co-training does not need the Chinese sentences as the medium, Only need a small number of marked Naxi corpus and a large number of unmarked Naxi corpus,avoid the negative result of marked inadequat e Naxi corpus.(3)After we had the treebank, the next job is to build Naxi Dependency P arser.Firstly.Using SVM methods to build root finder.Use the root Finder to fin d out the root,use the root divided the sentence into clause;secondly,analyzed o n clause by the deterministic algorithm, This method reduces the complexity of the sentence, Avoiding errors by that the sentence complexity is too high.