In Chinese Question Answering System, in order to improve the accuracy of the passage retrieval which includes answers, this paper presents a method to compute the weight of passage based on the analysis of current pa...
详细信息
In Chinese Question Answering System, in order to improve the accuracy of the passage retrieval which includes answers, this paper presents a method to compute the weight of passage based on the analysis of current passage retrieval methods. The word frequency of query words and query expansion words in passage, the length of passage;the distribution density and minimum match span of query words and query expansion words in passage are considered. The weight calculation of the answer passage is realized. And candidate passage set is retrieved on the basis of the weight Calculation. The experiment result on passage retrieval shows that this method has good effects that the MRR value equals 0.50.
In the Chinese question answering system, the answer extraction is generally realized by extracting answers from answer sentence pattern which is corresponded to the question. But the extraction of answer sentence pat...
详细信息
In the Chinese question answering system, the answer extraction is generally realized by extracting answers from answer sentence pattern which is corresponded to the question. But the extraction of answer sentence pattern greatly depended on expert's language knowledge, so the patterns can't be listed fully that the precision of the answer extraction will be affected. A method that the answer sentence pattern can be obtained by machine learning is proposed in this paper. All types of Chinese question answer pairs are statistically analyzed, the query of question is constructed, and then the corresponding documents are recalled by search engines. Next, the sentence patterns with the maximum weights are selected as the candidate answer sentence pattern by calculating the weight of sentence patterns in corresponding document. At last, the answer to the question is extracted from the Internet based on the answers sentence pattern which has been obtained. The experiment on extracting answer for Chinese factoid question is done. The experiment results make a very good effect that the MRAR value equals 0.28.
This paper proposes a new interpolatory subdivision scheme for triangular meshes with arbitrary topology. This scheme generates C1-continuous limit surfaces with bounded curvature property, which leads to improved qua...
详细信息
This paper proposes a new interpolatory subdivision scheme for triangular meshes with arbitrary topology. This scheme generates C1-continuous limit surfaces with bounded curvature property, which leads to improved quality surfaces. To compute newly inserted vertices, a group of rules are designed. In order to generate smooth surfaces near extraordinary vertices, new masks for the irregular cases are also constructed. The method for dealing with boundaries is presented and the limit curves on the boundaries is C1-continuous also. Eigen-structure analysis demonstrates the smoothness of the limit surfaces generated by the proposed scheme. Finally, results of refined models with the new scheme are presented and the difference of refined models between the new scheme and other schemes is shown also. In most cases, the new scheme generates more pleasure surfaces, especially near the irregular settings.
Predicted particle swarm optimization is an enhanced version aiming to increase the utilization ratio of velocity information, and the performance heavily relies upon the parameters settings. According to control theo...
详细信息
Predicted particle swarm optimization is an enhanced version aiming to increase the utilization ratio of velocity information, and the performance heavily relies upon the parameters settings. According to control theory, the trajectories of position and velocity vectors of each particle can be both viewed as oscillatory links, and the relationship between inertia weight and accelerator coefficients is obtained. Thus, a self-adjusting parameter strategy is proposed. Simulation results show the new proposed strategy is powerful and useful.
In this paper, we present a scalable parallel simulator - MPUS - for verifying the design of our next generation high performance parallel computer - RedNeurons(RN) parallel computer. The RedNeurons parallel computer ...
详细信息
ISBN:
(纸本)9781595937575
In this paper, we present a scalable parallel simulator - MPUS - for verifying the design of our next generation high performance parallel computer - RedNeurons(RN) parallel computer. The RedNeurons parallel computer is based on CMP technology, and it adopts an advanced but maybe some complicated architecture and topology. This paper mainly describes the design and implementation of the MPUS.
The accuracy of localization is a significant criterion to evaluate the practical utility of localization algorithm in wireless sensor networks. In mostly localization algorithms, one of the main methods to improve lo...
详细信息
The interference imposes a potential negative impact on the performance of a wireless network. A device is interfered if it receives a transmission not intended for it. In this paper, we introduce an explicit computat...
详细信息
Wireless sensor networks have been widely used in many important fields. Medium Access Control (MAC) protocols have a significant effect on the function and performance of sensor networks. At present, most MAC protoco...
详细信息
Hosts in mobile ad hoc networks (MANET) are capable of self-configuration and they can migrate at will, which makes the network topology highly dynamic. Thus, routes may break easily since the data session fails when ...
详细信息
This paper proposes a filtering technology based on content. This technology constructs an estimation function, and estimates the candidate properties, then chooses several properties with the highest value which are ...
详细信息
This paper proposes a filtering technology based on content. This technology constructs an estimation function, and estimates the candidate properties, then chooses several properties with the highest value which are used by all documents in the document set. At the same time, it can get the characteristics which have low frequency in counter samples but high frequency in consistent samples, and those which have low frequency in consistent samples but high frequency in counter samples by calculating increase of information. And then, it can choose several unlisted words as new keywords which have higher information increase and have not been used as keywords, and audit and filtering of network information at last.
暂无评论