Siheng Li, Kejiao Li, Zenan Xu, Guanhua Huang, Kun Li, Haoyuan Wu, Wujiajia, Zihao Zheng, Chenchen Zhang, Kun Shi, Xue Gong, Qi Yi, Ruibin Xiong, Tingqiang Xu, Yuhao Jiang, Jianfeng Yan, Yuyuan Zeng, Guanghui Xu, Jinbao Xue, Zhijiang xu, Zheng Fang, Shuai LI, Qibin Liu, Xiaoxue Li, Zhuoyu Li, Yangyu Tao, Fei Gao, Cheng Jiang, Bochao Wang, Kai Liu, Jianchen Zhu, Wai Lam, Bo Zhou, Di Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Inference algorithms of latent Dirichlet allocation (LDA), either for small or big data, can be broadly categorized into expectation-maximization (EM), variational Bayes (VB) and collapsed Gibbs sampling (GS). Looking for a unified understanding of these different inference algorithms is currently an important open problem. In this paper, we revisit these three algorithms from the entropy perspective, and show that EM can achieve the best predictive perplexity (a standard performance metric for LDA accuracy) by minimizing directly the cross entropy between the observed word distribution and LDA's predictive distribution. Moreover, EM can change the entropy of LDA's predictive distribution through tuning priors of LDA, such as the Dirichlet hyperparameters and the number of topics, to minimize the cross entropy with the observed word distribution. Finally, we propose the adaptive EM (AEM) algorithm that converges faster and more accurate than the current state-of-the-art SparseLDA [20] and AliasLDA [12] from small to big data and LDA models. The core idea is that the number of active topics, measured by the residuals between E-steps at successive iterations, decreases significantly, leading to the amortized σ(1) time complexity in terms of the number of topics. The open source code of AEM is available at GitHub.

Latent Dirichlet allocation (LDA) is a widely-used probabilistic topic modeling tool for content analysis such as web mining. To handle web-scale content analysis on just a single PC, we propose multi-core parallel expectation-maximization (PEM) algorithms to infer and estimate LDA parameters in shared memory systems. By avoiding memory access conflicts, reducing the locking time among multiple threads and residual-based dynamic scheduling, we show that PEM algorithms are more scalable and accurate than the current state-of-the-art parallel LDA algorithms on a commodity PC. This parallel LDA toolbox is made publicly available as open source software at mloss.org.

XML is a more desirable format for modeling and storing clinical data in EMR (Electronic medical record) applications for its extendibility; however, existing EMR systems either are built on top of RDBMS or file systems or lack of support for complex and large scale healthcare applications, such as treatment effectiveness analysis and procedure optimization. SAP Technology Lab, China is developing a clouds-enabled information appliance, Xbase, built on top of Hadoop, which is the first XML-based information appliance designed specifically for large scale and complex healthcare applications. XML presents a different set of challenges for query processing, indexing, parallelism, and distributed computing using existing Hadoop's APIs as well as its HDFS storage infrastructure and MapReduce framework. In this paper, we describe system architecture and internal designs of Xbase as well as how the indexing is mapped to RDBMS and Hadoop. We also discuss why we select Hadoop over other candidates, such as Hbase, Google's Bigtable, and Hive.

Update of applications in SaaS is expected to be a continuous efforts and cannot be done overnight or over the weekend. In such migration efforts, users are trained and shifted from a existing version to a new version successively. There is a long period of time when both versions of applications co-exist. Supporting two systems at the same time is not a cost efficient option and two systems may suffer from slow response time due to continuous synchronization between two systems. In this paper, we focus on how to enable progressive migration of multi-version applications in SaaS via evolving schema. Instead of maintain two systems, our solution is to maintain an intermediate schema that is optimized for mixed workloads for new and old applications. With a application migration schedule, an genetic algorithm is used to find out the more effective intermediated schema as well as migration paths and schedule. A key advantage of our approach is optimum performance during the long migration period while maintaining the same level of data movement required by the migration. We evaluated the proposed progressive migration approach on a TPCW workload and results validated its effectiveness of across a variety of scenarios; Experimental results demonstrate that our incremental migration proposed in this paper could bring about 200% performance gain as compared to the existing system.