Entity-Centric Topic Extraction and Exploration: A Network-Based Approach.

Topic modeling is an important tool in the analysis of corpora and the classification and clustering of documents. Various extensions of the underlying graphical models have been proposed to address hierarchical or dynamical topics. However, despite their popularity, topic models face problems in the exploration and correlation of the (often unknown number of) topics extracted from a document collection, and rely on compute-intensive graphical models. In this paper, we present a novel framework for exploring evolving corpora of news articles in terms of topics covered over time. Our approach is based on implicit networks representing the cooccurrences of entities and terms in the documents as weighted edges. Edges with high weight between entities are indicative of topics, allowing the context of a topic to be explored incrementally by growing network sub-structures. Since the exploration of topics corresponds to local operations in the network, it is efficient and interactive. Adding new news articles to the collection simply updates the network, thus avoiding expensive recomputations of term and topic distributions.