论文部分内容阅读
针对汉语简单名词短语提出一种混合的识别模型,该模型包括组合分类器方法和一种并列结构识别算法.分析简单名词短语与其他类型名词短语的异同,进一步确定其识别任务;利用词、词性和词义信息构建层叠的组合分类器对其进行识别;最后,创新性地利用词性组合模板和基于词向量的语义相似度构造一种并列结构识别算法,提高了简单名词短语的识别精度的同时,保持了其内部结构的清晰.简单名词短语识别的F-值为91.19%,比目前最好结果提高了0.85%,验证了该方法识别简单名词短语的有效性;内部并列结构左右边界的识别精确率分别为80.93%和82.11%,在一定程度上解决了目前多名词并列结构难以识别的问题.
This paper proposes a hybrid recognition model for Chinese simple noun phrases, which includes a combination of classifier and a parallel structure recognition algorithm.It analyzes the similarities and differences between simple noun phrases and other types of noun phrases to further identify their recognition tasks. And the semantic information to build a cascade of classifiers to identify them; Finally, the innovative use of part-of-speech combination template and semantic similarity based on word vectors to construct a parallel structure recognition algorithm to improve the recognition accuracy of simple noun phrases, And its internal structure is clear. The F-value of simple noun phrase recognition is 91.19%, which is 0.85% higher than the best result so far. It verifies the validity of this method in recognizing simple noun phrases. The recognition of the left and right boundary of the internal parallel structure The rates were 80.93% and 82.11%, respectively, to a certain extent, solved the problem that the parallel structure of multiple nouns is difficult to identify.