Efficient methods for topic model inference on streaming document collections
Limin Yao, David Mimno, Andrew McCallum · ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) · 2009
Topic models provide a powerful tool for analyzing large text collections by representing high dimensional data in a low dimensional subspace. Fitting a topic model given a set of training documents requires approximate inference techniques that are computationally expensive. With today's large-scale, constantly expanding document collections, it is useful to be able to infer topic distributions for new documents without retraining the model. In this paper, we empirically evaluate the performance of several methods for topic inference in previously unseen documents, including methods based on Gibbs sampling, variational inference, and a new method inspired by text classification. The classification-based inference method produces results similar to iterative inference methods, but requires only a single matrix multiplication. In addition to these inference methods, we present SparseLDA, an algorithm and data structure for evaluating Gibbs sampling distributions. Empirical results indicate that SparseLDA can be approximately 20 times faster than traditional LDA and provide twice the speedup of previously published fast sampling methods, while also using substantially less memory.
Cited by 410 works. doi:10.1145/1557019.1557121 Open-access PDF
Authors
References
- Latent dirichlet allocation
- A maximum entropy approach to natural language processing
- Finding scientific topics
- A Collapsed Variational Bayesian Inference Algorithm for Latent Dirichlet Allocation
- Fast collapsed gibbs sampling for latent dirichlet allocation
- Distributed Inference for Latent Dirichlet Allocation
- Modeling and predicting personal information dissemination behavior
- Dynamic mixture models for multiple time series
- Topic Models over Text Streams: A Study of Batch and Online Unsupervised Learning