基于SVM的汉语决策式依存分析

决策式分析有着贪婪的特性，容易引起错误增殖。针对该问题，提出一种基于SVM的汉语决策式依存分析算法。利用SVM构建根查找器，用根结点将句子划分为2个子句。从子句中识别出介词短语，采用改进后的Nivre算法分析子句。该算法在分析句子之前做预处理从而降低句子复杂度，减少错误增殖，分析准确率也相应得到提高。实验结果表明，该分析策略的准确率比Nivre算法提高了3.38%。

Deterministic Dependency Parsing for Chinese Based on SVM

Deterministic parsing has the greedy characteristic that easily brings the error propagation. Aiming at this question, this paper proposes a deterministic dependency analysis algorithm for Chinese including three steps. It utilizes SVM to construct a root finder to divide a sentence into two sub-sentences and extracts the prepositional phrases from sub-sentence. Improved Nivre＇s algorithm is adopted to parse sub-sentence. It does pre-processing before parsing sentences to decrease the complexity of the sentence and reduce the error propagation, and improve the parsing accuracy consequently. Experimental evaluation shows the accuracy of this parsing strategy is higher by 3.38% than Nivre＇s.