Beijing TRS Information Technology

www.trs.com.cn
Beijing, China
SEARCH FILTERS
Time filter
Source Type

Li S.,Beijing Information Science and Technology University | Lv X.,Beijing Information Science and Technology University | Lv X.,Beijing TRS Information Technology | Li Y.,Beijing Information Science and Technology University | And 3 more authors.
International Journal of Information Processing and Management | Year: 2010

Text classification is the key technology for topic tracking, and vector space model (VSM) is one of the most simple and effective model for topics representation. Feature selection algorithm in VSM is an important means of data pre-processing, and it can reduce vector space dimension and improve the generalization ability of the algorithm. Therefore, it is necessary for feature selection algorithms to be in-depth and extensive research. So we study how feature space dimension and feature selection algorithm affect topic tracking. Then we get the variation law that they affect topic tracking, and add up their optimal values in topic tracking. Finally, TDT evaluation methods prove that optimal topic tracking performance based on weight of evidence for text increases by 8.762% more than mutual information.


Liu Y.W.,Beijing Information Science and Technology University | Xiao S.B.,Beijing TRS Information Technology | Wang T.,Beijing TRS Information Technology | Shi S.C.,Beijing TRS Information Technology
Advanced Materials Research | Year: 2011

Judging the sentiment orientation of Chinese words is the basic work of the passage sentiment orientation research. Using Chinese basic sentiment words and corpus, we can identify sentiment words in the passage and expand sentiment lexicon effectively in order to improve the result of text semantic orientation analysis. With the basis of HowNet [1] sentiment words, we construct a Chinese sentiment lexicon by analyzing sentence structure and calculating the score of semantic similarity. We conduct Chinese text sentiment orientation classification experiment with this lexicon, the result shows the accuracy has achieved above 70% and obtained quite good classification effect. © (2011) Trans Tech Publications.


Li S.,Beijing Information Science and Technology University | Lv X.,Beijing Information Science and Technology University | Lv X.,Beijing TRS Information Technology | Zhan C.,Lab Tech Support | And 2 more authors.
Journal of Computers | Year: 2011

Through researching and analyzing adaptive strategy and fuzzy C-means (FCM) clustering algorithm, we put them together to form an adaptive FCM clustering algorithm. It is a good solution to the problem of local optimum as well as sensitivity to the initial value for the traditional FCM clustering algorithm. Finally, the new algorithm has been used in the divided region of police patrols in a city. In the division of the region, it has been proved by experiments that the sum of distance between a police vehicle and each possible accident scene can achieve the minimum value, which shows a significant effect of police patrols. And through the improved dijkstra algorithm to calculate shortest path length between a police vehicle and an accident scene, it proves that a police vehicle in the division of the region arrives at an accident scene within three minutes after accepting the warnings, whose proportion is 90.2%. On the base of the divided region, we put parameter adaptive thinking and MMAS together to form adaptive MMAS, which is used to calculate optimal patrol circuit in the divided region. It proves in the experiment that algorithm efficiency in adaptive max-min ant system increases by 26.34% more than non-adaptive max-min ant system under the same condition and has a good prospect for the optimal circuit of police patrols. © 2011 ACADEMY PUBLISHER.


Wang L.,Beijing Information Science and Technology University | Shi S.-C.,Beijing Information Science and Technology University | Shi S.-C.,Beijing TRS Information Technology | Lv X.-Q.,Beijing Information Science and Technology University | And 3 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2010

Based on the study of TF-IDF, information gain and information entropy, the paper proposes an improved method of weight calculation, which combines the TF-IDF Normalization with information gain, to extract key words. Moreover, to abstract indexing words with counting semantic similarity of the key words in order to finish a process of automatic indexing. Through the comparative experiment shows that the comprehensive assessment value of indexing words which are obtained by the modified method of weight calculation are higher than obtained by the traditional TF-IDF method. © 2010 Springer-Verlag.


Liu Y.,Beijing Information Science and Technology University | Xiao S.,Beijing Information Science and Technology University | Xiao S.,Beijing TRS Information Technology | Lv X.,Beijing Information Science and Technology University | And 3 more authors.
2010 International Conference on Computing, Control and Industrial Engineering, CCIE 2010 | Year: 2010

Through research on K-means algorithm of text clustering and semantic-based vector space model, a semantic-based K-means text clustering model is proposed to solve the problem on high-dimensional and sparse characteristics of text data set. The model reduces the semantic loss of the text data and improves the quality of text clustering. Experiments prove that semantic-based text clustering increases by more 6 percent than non-semantic-based one in the final evaluation of the F1 index value. © 2010 IEEE.


Yang C.,Beijing Information Science and Technology University | Shi S.,Beijing Information Science and Technology University | Shi S.,Beijing TRS Information Technology | Li L.,Beijing Information Science and Technology University | And 3 more authors.
Proceedings - 2011 10th International Symposium on Distributed Computing and Applications to Business, Engineering and Science, DCABES 2011 | Year: 2011

In the condition of single training sample, traditional methods get low recognition accuracy, or even can not be used. In view of this situation, this paper proposes a method to slove this problem. Firstly, face image is decomposed by image pyramids. Then, each layer image segmentation into sub images with th same size. After that, the feature of each sub image, which got with (W2DPC)2A, gets a weight through the adaptive method. Finally, Euclidean distance is used to classify face images. Experimental results on ORL and Yale show that the presented method can achieve a certain degree of recognition accuracy. © 2011 IEEE.


Zhang K.,Beijing Information Science and Technology University | Du Y.,Beijing Information Science and Technology University | Du Y.,Beijing TRS Information Technology | Lv X.,Beijing Information Science and Technology University | And 3 more authors.
Proceedings of the 2010 2nd International Conference on Future Computer and Communication, ICFCC 2010 | Year: 2010

Through introducing and analyzing the new form of media called micro-blog, we have concluded the characteristics of micro-blog and website structure features.Then we gave concrete realization of the search engine based on nutch technology, and transformed the existing Chinese word segmentation system. Finally the search engine was built completely. After the analysis of the collected data, we found micro-blog site features. What is more, we found that five-depth was the best depth for micro-blog. ©2010 IEEE.


Liu Z.,Beijing Information Science and Technology University | Lv X.,Beijing Information Science and Technology University | Lv X.,Beijing TRS Information Technology | Liu K.,Beijing Information Science and Technology University | And 3 more authors.
2nd International Workshop on Education Technology and Computer Science, ETCS 2010 | Year: 2010

Based on the text information processing, we have made a study on the application of support vector machine in text categorization. Through introducing the basic principle of SVM, we described the process of text classification and further proposed a SVM-based classification model. Finally, experimental data show that F1 value of SVM classifier has reached more than 86.26%, and the classification results comparing to other classification methods have greatly improved, and it also proves that SVM is an effective machine learning method. © 2010 IEEE.


Yue L.,Beijing Information Science and Technology University | Lv X.,Beijing Information Science and Technology University | Lv X.,Beijing TRS Information Technology | Xiao S.,Beijing Information Science and Technology University | And 3 more authors.
Proceedings 2011 International Conference on Mechatronic Science, Electric Engineering and Computer, MEC 2011 | Year: 2011

Topic Detection is a sub-task of Topic Detection and Tracking, its main task is to find and organize topics that system didn't know. By analyzing hundreds of website news reports, we find that usually there exist some keywords in text, and early study didn't pay enough attention to this, we propose a topic detection algorithm according this. The algorithm is based on K-means clustering algorithm, choose keywords and enhance the weight of keywords. Experiment proves it can improve the efficiency of system. © 2011 IEEE.


Loading Beijing TRS Information Technology collaborators
Loading Beijing TRS Information Technology collaborators