[ABE-L] Seminários do Programa de Pós Graduação em Estatística/UFRGS – Identifying topics in text using parametric and non-parametric Bayesian models

[Márcia Barbian]

Olá,

Temos o prazer de convidar a todas e todos para mais um seminário do
Programa de Pós-Graduação em Estatística da UFRGS.

Nosso próximo encontro ocorrerá de forma *REMOTA* no *dia 14 de julho* e
terá como palestrante a pesquisadora *Daiane Aparecida Zuanetti* -
Professora Departamento de Estatística - Universidade Federal de São Carlos
e Programa Interinstitucional de Pós-Graduação em Estatística (PIPGEs) da
UFSCar-USP
*Título:* Identifying topics in text using parametric and non-parametric
Bayesian models

*Abstract:* One branch within text analysis is topic modeling, whose
methodologies aim to understand the topic structure that forms a document
and segmenting multiple documents by their dominant topics (subjects). One
of the pioneering methods in this context is the mixture model (MM), which
assumes that each document is composed of words from a single topic. Given
this limitation, the methodology of latent Dirichlet allocation (LDA) has
gained considerable visibility due to its greater flexibility. The number
of topics needs to be predetermined for both MM and LDA estimation, and if
unknown, the best number can be chosen via selection criteria. One
alternative for selecting the number of topics and estimating the model
simultaneously is through non-parametric Bayesian methodologies. In this
work, in addition to contrasting probabilistic methodologies for topic
analysis and proposing a non-parametric marginal Bayesian mixture model for
this purpose, we compare criteria that generally measure the quality of the
models and are used for choosing the number of topics. To do this, we apply
the models and selection criteria to two sets of real data. Joint work with
Edvaldo Capobiango Coelho Filho.


*Data: 14/07/2026Horário: 13h30*
*Local: *https://mconf.ufrgs.br/webconf/ppgest

Abraço,

Márcia Barbian
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20260711/cd47d61a/attachment.htm>