A Visual Analytics Framework for Identifying Topic Drivers in Media Events

Document Type

Article

Publication Date

9-14-2017

Publication Title

IEEE Transactions on Visualization and Computer Graphics

Volume

24

Issue

9

First page number:

2501

Last page number:

2515

Abstract

Media data has been the subject of large scale analysis with applications of text mining being used to provide overviews of media themes and information flows. Such information extracted from media articles has also shown its contextual value of being integrated with other data, such as criminal records and stock market pricing. In this work, we explore linking textual media data with curated secondary textual data sources through user-guided semantic lexical matching for identifying relationships and data links. In this manner, critical information can be identified and used to annotate media timelines in order to provide a more detailed overview of events that may be driving media topics and frames. These linked events are further analyzed through an application of causality modeling to model temporal drivers between the data series. Such causal links are then annotated through automatic entity extraction which enables the analyst to explore persons, locations, and organizations that may be pertinent to the media topic of interest. To demonstrate the proposed framework, two media datasets and an armed conflict event dataset are explored.

Keywords

Semantic similarity; Media annotation; Visual analytics; Causality modeling; Social media

Disciplines

Computer Sciences | Graphics and Human Computer Interfaces

Language

English


Search your library

Share

COinS