Viscovery: Trend Tracking in Opinion Forums based on Dynamic Topic Models

Espinoza, Ignacio; Mendoza, Marcelo; Ortega, Pablo; Rivera, Daniel E.; Weiss, Fernanda

doi:10.6025/jdim/2019/17/1/13-24

Cited by 1 publication

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Methods such as Patchinko Allocation topic model [51], correlated topic model [52][53][54], supervised topic model [55,20,[56][57][58], dynamic topic model [59,29,60,61], hierarchical topic model [62], spherical topic model [63], all characterize these alternatives provided to the LDA architecture. Currently, within the framework of LDA-based topic models, the advancement of social media platforms [64] and online services such as Q&A (questions and answers) [34] communities are having some serious impacts on extensions such as dynamic topic model [65,66,64,29,60], correlated topic model [53,52], supervised topic model, and online topic model schemes [67][68][69][70]41]. Current topic models also provide improvement in semantic analysis [30,44,71,72,29] to enhance coherence in the topics estimated and the relationship between documents [73].…”

Section: Related Work and Backgroundmentioning

confidence: 99%

“…Current topic models also provide improvement in semantic analysis [30,44,71,72,29] to enhance coherence in the topics estimated and the relationship between documents [73]. Some current hot topics in research (within topic modeling framework) include social network analysis, bioinformatics [74], emotion, sentiment analysis [65,75,66], and information retrieval [76,35]. It is important to notice that the generative setting, through the BoW representation including its derivates and topic models, have provided tremendous success in computer vision for object learning and categorization [32, 77-79, 4, 5].…”

Section: Related Work and Backgroundmentioning

confidence: 99%

See 1 more Smart Citation

Efficient integration of generative topic models into discriminative classifiers using robust probabilistic kernels

Ihou

Bouguila

Bouachir

2020

Pattern Anal Applic

View full text Add to dashboard Cite

A direct implementation of supervised topic modeling using a Naive Bayes classifier is mainly characterized by the formulation of robust generative topic models that utilize prior distributions such as Dirichlet in LDA (latent Dirichlet allocation), where the classification ultimately follows the Bayes theorem. Though, in large scale applications, SVM (support vector machine) seems to outperform Naive Bayes. In this paper, we propose a classification framework that combines the flexibility of the generative topic models and the strong performance of the SVM. We therefore present a generative-discriminative collapsed variational Bayes technique for text documents and visual classification. Our collapsed variational Bayes topic model implements simultaneously two different and asymmetric conjugate priors within the same generative process as it specifically draws the document and corpus parameters using both GD (generalized Dirichlet) and BL (Beta-Liouville) distributions. Each of these flexible priors generalizes the Dirichlet in LDA. The proposed hybrid model results in a much improved inference that contributes to more accurate estimates, coherent (topic) generative features, a robust formulation of probabilistic kernels, and a much improved classification rate. Experiments in image and text documents classification show the merits of the proposed approach.

show abstract