%0 Conference Proceedings
%T PWA-PEM for Latent Tree Model and Hierarchical Topic Detection
%+ Beijing University of Posts and Telecommunications (BUPT)
%+ China Institute of Marine Industrial Systems Engineering [Beijing]
%A Liu, Zhuchen
%A Chen, Hao
%A Li, Jie
%A Yu, Yanhua
%Z Part 5: Natural Language Processing
%< avec comité de lecture
%( IFIP Advances in Information and Communication Technology
%B 10th International Conference on Intelligent Information Processing (IIP)
%C Nanning, China
%Y Zhongzhi Shi
%Y Eunika Mercier-Laurent
%Y Jiuyong Li
%I Springer International Publishing
%3 Intelligent Information Processing IX
%V AICT-538
%P 183-191
%8 2018-10-19
%D 2018
%R 10.1007/978-3-030-00828-4_19
%K Hierarchical Latent Tree Analysis
%K Topic detection
%K Aitken acceleration
%K PEM
%Z Computer Science [cs]Conference papers
%X Hierarchical Latent Tree Analysis (HLTA) is a new method of topic detection. However, HLTA data input uses TF-IDF selection term, and relies on EM algorithm for parameter estimation. To solve this problem, a method of accelerating part of speech weight (PWA-PEM-HLTA) is proposed based on Progressive EM-HLTA (PEM-HLTA). Experimental results show that this method improves the execution efficiency of PEM-HLTA, averaging 4.9 times speed, and improves the speed of 6 times in the best case.
%G English
%Z TC 12
%2 https://inria.hal.science/hal-02197798/document
%2 https://inria.hal.science/hal-02197798/file/473854_1_En_19_Chapter.pdf
%L hal-02197798
%U https://inria.hal.science/hal-02197798
%~ IFIP
%~ IFIP-AICT
%~ IFIP-TC
%~ IFIP-TC12
%~ IFIP-IIP
%~ IFIP-AICT-538