STATS JOB TALK - Xiaoyi Yang, Carnegie Mellon University

Learning Social Networks from Text Data using Covariate Information 
Describing and characterizing the impact of historical figures can be challenging, but unraveling their social structures perhaps even more so. Historical social network analysis methods can help and may also illuminate people who have been overlooked by historians, but turn out to be influential social connection points. Text data, such as biographies, can be a useful source of information to learn the structure of historical social networks but can also introduce challenges in identifying links. The majority of the research on network reconstruction is based on the co-occurrence of names in text, since if two people know each other, it is more likely that they show up in or near the same text section. However, even if two people are co-mentioned, they do not necessarily know each other; it may be that they share an acquaintance. Moreover, given historical tendency of frequently used or common names, without additional distinguishing information, we may introduce incorrect connections.

In this work, we extend the Local Poisson Graphical Lasso model with a (multiple) penalty structure that incorporates covariates giving increased link probabilities to people with shared covariate information. First, we will explain why the statistical graphical models that only relies on co-mention counts can not fully exploit information in the text data. Second, we will explain how to manipulate and incorporate multiple lasso penalties to include covariate information like sex, birth/death year, family name, and social group membership. We present results on data simulated with characteristics of historical networks and show that this type of penalty structure can improve network recovery as measured by precision and recall. We also illustrate the approach on biographical data of individuals who lived in early modern Britain, targeting the period from 1500 to 1575.

Tuesday, February 23, 2021 at 4:45pm

Virtual Event
Event Type

Lecture

Audience

Faculty, Students, Staff

Department
Division of Mathematical and Natural Sciences, Mathematics
Subscribe
Google Calendar iCal Outlook

Recent Activity