Familia: A Configurable Topic Modeling Framework for Industrial Text Engineering

In this paper, we propose a configurable topic modeling framework named Familia. Familia supports an important line of models that are widely applicable in text engineering scenarios. order to relieve burdens software engineers without knowledge Bayesian networks, is able conduct automatic parameter inference for variety models. Simply through changing the data organization Familia, easily explore broad spectrum existing or even design their own models, and find one best suits problem at hand. With its superior extendability, has novel sampling mechanism strikes balance between effectiveness efficiency inference. Furthermore, essentially big parallel distributed storage. The utilities necessity demonstrated real-life industrial applications. would significantly enlarge engineers’ arsenal pave way utilizing highly customized problems. Source code have been released Github via https://github.com/baidu/Familia/.