Location:
Search - prop200406
Search list
Description: 概率句法分析器对于统计自然语言处理的很多高层应用，如统计机器翻译、问答系统、信息抽取、文本挖掘等都是至关重要的，直接决定这些应用系统的最终性能。本系统是一个概率型的Chart分析器。系统的分析算法是采用了多种优化策略。分析结果是概率最大的一棵分析树。在概率模型方面，本系统在一定程度上突破了pcfg的上下文无关假设，引入了结构上下文条件，使得分析结果正确率有了明显提高。在使用宾州中文树库进行的实验中，我们的分析器的标记召回率和标记精确率平均在75%-80%左右。在使用一个短句树库进行的实验中，两个指标都在90%以上。概率句法分析既需要建立合理的概率模型，又需要积累树库等语言资源。我们把所做的一点工作进行开放，就是希望抛弃闭门造车的做法，集思广益，推动这个基础领域的发展，使汉语的句法分析尽早实现实用化-probability syntax analyzer for statistical natural language processing of many senior applications, such as statistical machine translation, quiz systems, information extraction, text mining are essential, these applications directly determine the final performance. The system is a probability - based Chart analyzer. Systematic analysis algorithm is optimized using a variety of strategies. Results of the analysis is the greatest probability of a tree. The probability model, the system to some extent breakthrough in the context of pcfg unrelated to the assumption that the introduction of the context of the structural conditions, making results of the analysis accuracy rate has markedly improved. The use of Chinese tree of Pennsylvania library experiments, the analyzer markers recall rate a
Platform: |
Size: 565168 |
Author: 江鹏 |
Hits:
Description: 概率句法分析器对于统计自然语言处理的很多高层应用，如统计机器翻译、问答系统、信息抽取、文本挖掘等都是至关重要的，直接决定这些应用系统的最终性能。本系统是一个概率型的Chart分析器。系统的分析算法是采用了多种优化策略。分析结果是概率最大的一棵分析树。在概率模型方面，本系统在一定程度上突破了pcfg的上下文无关假设，引入了结构上下文条件，使得分析结果正确率有了明显提高。在使用宾州中文树库进行的实验中，我们的分析器的标记召回率和标记精确率平均在75%-80%左右。在使用一个短句树库进行的实验中，两个指标都在90%以上。概率句法分析既需要建立合理的概率模型，又需要积累树库等语言资源。我们把所做的一点工作进行开放，就是希望抛弃闭门造车的做法，集思广益，推动这个基础领域的发展，使汉语的句法分析尽早实现实用化-probability syntax analyzer for statistical natural language processing of many senior applications, such as statistical machine translation, quiz systems, information extraction, text mining are essential, these applications directly determine the final performance. The system is a probability- based Chart analyzer. Systematic analysis algorithm is optimized using a variety of strategies. Results of the analysis is the greatest probability of a tree. The probability model, the system to some extent breakthrough in the context of pcfg unrelated to the assumption that the introduction of the context of the structural conditions, making results of the analysis accuracy rate has markedly improved. The use of Chinese tree of Pennsylvania library experiments, the analyzer markers recall rate a
Platform: |
Size: 565248 |
Author: 江鹏 |
Hits: