Loading...
Please wait, while we are loading the content...
Similar Documents
Multi-dimensional Topic Modeling with Determinantal Point Processes
| Content Provider | Semantic Scholar |
|---|---|
| Author | Cohen, Jeremy |
| Copyright Year | 2015 |
| Abstract | Probabilistic topics models such as Latent Dirichlet Allocation (LDA) provide a useful and elegant tool for discovering hidden structure within large data sets of discrete data, such as corpuses of text. However, LDA implicitly discovers topics along only a single dimension. Recent research on multi-dimensional topic modeling aims to devise techniques that can discover multiple groups of topics, where each group models some different dimension, or aspect, of the data. For example, applying a multi-dimensional topic modeling algorithm to a text corpus could result in three dimensions that turn out to represent “subject,” “sentiment,” and “political ideology.” In this work, we present a new multi-dimensional topic model that uses a determinantal point process prior to encourage different groups of topics to model different dimensions of the data. Determinantal point processes are probabilistic models of repulsive phenomena which originated in statistical physics but have recently seen interest from the machine learning community. Though topic models are usually applied to text, we motivate our method with the problem of discovering “tastes” (topics) in a data set of recipes that have been rated by users of a cooking web site. We present both an unsupervised algorithm, which discovers dimensions and their tastes automatically, and a semisupervised algorithm, which allows the modeler to steer the tastes (topics) towards ones that will be semantically meaningful, by “seeding” each taste with a small number of representative recipes (words). Our results on the recipe data set are mixed, but we are hopeful that the general technique presented here might very well prove useful to multi-dimensional topic modeling in other domains. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.cs.princeton.edu/sites/default/files/uploads/jeremy_cohen.pdf |
| Alternate Webpage(s) | https://www.cs.princeton.edu/sites/default/files/uploads/jeremy_cohen.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
