Temporal corpus summarization using submodular word coverage

Abstract

In many areas of life, we now have almost complete electronic archives reaching back for well over two decades. This includes, for example, the body of research papers in computer science, all news articles written in the US, and most people's personal email. However, we have only rather limited methods for analyzing and understanding these collections. While keyword-based retrieval systems allow efficient access to individual documents in archives, we still lack methods for understanding a corpus as a whole. In this paper, we explore methods that provide a temporal summary of such corpora in terms of landmark documents, authors, and topics. In particular, we explicitly model the temporal nature of influence between documents and re-interpret summarization as a coverage problem over words anchored in time. The resulting models provide monotone sub-modular objectives for computing informative and non-redundant summaries over time, which can be efficiently optimized with greedy algorithms. Our empirical study shows the effectiveness of our approach over several baselines.

Self reinforcement for important passage retrieval

“…Closely related to our work are the unsupervised key phrase extraction approaches that have been explored to reinforce summarization [32,29,11,22,24]. Namely, Litval and Last [11] and Riedhammer et al [22] propose the use of key phrases to summarize news articles [11] and meetings [22].…”

Streaming Algorithms for Submodular Function Maximization

“…Constrained submodular maximization has found several new applications in recent years. Some of these include data summarization [LB11,SSSJ12,DKR13], influence maximization in social networks [KKT03, CWY09, CWW10, GBL11, SS13], generalized assignment [CCPV07], mechanism design [BIK07], and network monitoring [LKG + 07].…”

Streaming submodular maximization

“…Over the recent years, submodular optimization has been identified as a powerful tool for numerous data mining and machine learning applications including viral marketing [17], network monitoring [22], news article recommendation [10], nonparametric learning [14,29], document and corpus summarization [23,7,33], network inference [30], and Determinantal Point Processes [13]. A problem of key importance in all these applications is to maximize a monotone submodular function subject to a cardinality constraint (i.e., a bound on the number k of elements that can be selected).…”

“…pelling approach that has gained a lot of interest in recent years is data summarization: selecting representative subsets of manageable size out of large data sets. Applications range from exemplar-based clustering [8], to document [23,7] and corpus summarization [33], to recommender systems [10,9], just to name a few. A systematic way for data summarization, used in all the aforementioned applications, is to turn the problem into selecting a subset of data elements optimizing a utility function that quantifies "representativeness" of the selected set.…”