A Novel Fast Framework for Topic Labeling Based on Similarity-preserved Hashing

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3339–3348, Osaka, Japan, December 11-17 2016. A Novel Fast Framework for Topic Labeling Based on Similarity-preserved Hashing Xian-Ling Mao♠, Yi-Jing Hao♠, Qiang Zhou♠, Wen-Qing Yuan♥, Liner Yang♣, Heyan Huang♠∗ ♠Department of Computer Science, Beijing Institute of Technology, China ♥Beijing YanZhiYouWu Technology Co., LTD, China ♣Department of Computer Science and Technology, Tsinghua University, China {maoxl, hyj, qzhou, hhy63}@bit.edu.cn yuan@yzyw-inc.com, lineryang@gmail.com Abstract Recently, topic modeling has been widely applied in data mining due to its powerful ability. A common, major challenge in applying such topic models to other tasks is to accurately interpret the meaning of each topic. Topic labeling, as a major interpreting method, has attracted significant attention recently. However, most of previous works only focus on the effectiveness of topic labeling, and less attention has been paid to quickly creating good topic descriptors; Meanwhile, it’s hard to assign labels for new emerging topics by using most of existing methods. To solve the problems above, in this paper, we propose a novel fast topic labeling framework that casts the labeling problem as a k-nearest neighbor (KNN) search problem in probability distributions. Our experimental results show that the proposed sequential interleaving method based on locality sensitive hashing (LSH) technology is efficient in boosting the comparison speed among probabil- ity distributions, and the proposed framework can generate meaningful labels to interpret topics, including new emerging topics. 1 Introduction A wealth of topic models have been proposed to extract interesting topics in the form of multinomial distributions from the corpus automatically, which are useful data mining tools for the statistical analysis of document collections and other discrete data. A common, major challenge in applying all such topic models is to accurately interpret the meaning of each topic. In general, it is very difficult for users to understand a topic merely based on the multinomial word distribution, especially when they are not familiar with the background knowledge. Topic labeling, which generates meaningful labels for a topic so as to facilitate topic interpretation, has attracted increasing attention recently. Early research on topic labeling generally either select top words in the distribution as primitive labels (Blei et al., 2003; Ramage et al., 2009), or generate labels manually in a subjective manner (Mei et al., 2006; Mei and Zhai, 2005). However, it is highly desirable to automatically generate meaningful labels. Several automatical labeling methods have been proposed recently (Mei et al., 2007; Lau et al., 2010; Lau et al., 2011; Magatti et al., 2009; Mao et al., 2012; Hulpus et al., 2013; Mehdad et al., 2013; Cano et al., 2014). These existing approaches generally take following steps to generate meaningful labels for a given topic: (1) extract candidate labels; (2) rank candidate labels. First, most of existing methods extract candidate labels from a document collection by natural language processing techniques for a given topic, which is time-consuming; for example, the candidate labels in the method proposed by Mei et al. (2007) are extracted from …

Content truncated to 500 words