Practitioners dealing with large text collections frequently use topic models such as Latent Dirichlet Allocation (LDA) and Non-negative Matrix Factorization (NMF) in their projects to explore trends. Despite twenty years of accrued advancement in natural language processing tools, these models are found to be slow and challenging to apply to text exploration projects. In our work, we engaged with practitioners (n=15) who use topic modeling to explore trends in large text collections to understand their project workflows and investigate which factors often slow down the processes and how they deal with such errors and interruptions in automated topic modeling. Our findings show that practitioners are required to diagnose and resolve context-specific problems with preparing data and models and need control for these steps, especially for data cleaning and parameter selection. Our major findings resonate with existing work across CSCW, computational social science, machine learning, data science, and digital humanities. They also leave us questioning whether automation is actually a useful goal for tools designed for topic models and text exploration.

We present a general and modular method for privacy-preserving Bayesian inference for Poisson factorization, a broad class of models that includes some of the most widely used models in the social sciences. Our method satisfies limited-precision local privacy, a generalization of local differential privacy that we introduce to formulate appropriate privacy guarantees for sparse count data. We present an MCMC algorithm that approximates the posterior distribution over the latent variables conditioned on data that has been locally privatized by the geometric mechanism. Our method is based on two insights: 1) a novel reinterpretation of the geometric mechanism in terms of the Skellam distribution and 2) a general theorem that relates the Skellam and Bessel distributions. We demonstrate our method’s utility using two case studies that involve real-world email data. We show that our method consistently outperforms the commonly used naive approach, wherein inference proceeds as usual, treating the locally privatized data as if it were not privatized.

Duplicate documents are a pervasive problem in text datasets and can have a strong effect on unsupervised models.Methods to remove duplicate texts are typically heuristic or very expensive, so it is vital to know when and why they are needed.We measure the sensitivity of two latent semantic methods to the presence of different levels of document repetition.By artificially creating different forms of duplicate text we confirm several hypotheses about how repeated text impacts models.While a small amount of duplication is tolerable, substantial over-representation of subsets of the text may overwhelm meaningful topical patterns.

Rule-based stemmers such as the Porter stemmer are frequently used to preprocess English corpora for topic modeling. In this work, we train and evaluate topic models on a variety of corpora using several different stemming algorithms. We examine several different quantitative measures of the resulting models, including likelihood, coherence, model stability, and entropy. Despite their frequent use in topic modeling, we find that stemmers produce no meaningful improvement in likelihood and coherence and in fact can degrade topic stability.