An inhomogeneous Poisson process model for text time-series signals
| When |
Jan 28, 2025
from 01:00 to 02:00 |
|---|---|
| Where | M7 101 |
| Attendees |
Ioannis Chalkiadakis |
| Add event to calendar |
vCal
iCal
|
Ioannis Chalkiadakis
Title: An inhomogeneous Poisson process model for text time-series signals
The talk will present a stochastic model for count time-series data observed at regular or irregular time intervals, using text data signals as covariates. First, we will consider how to think of text as a time-series signal suitable for being statistically processed and modelled, and why it is important to do so. We will begin with a statistical framework for constructing text time-series signals from unstructured text data, before presenting our model for time-series of counts, which consists of an inhomogeneous Poisson process with a chi-squared process intensity function, driven by text-based covariates. An efficient optimisation procedure will be demonstrated which leverages control points and a variational inference framework. Finally, an example of the model's utility in capturing statistical signatures of texts will be illustrated via some preliminary analysis in the context of political science and US presidential rhetoric.
Website: https://scholar.google.fr/citations?hl=fr&user=d4PB7B4AAAAJ&view_op=list_works&sortby=pubdate
In Room M7 101, 1st floor, Monod campus, ENSL.