Tracking the dynamic evolution of participant salience in a discussion

Abstract

We introduce a technique for analyzing the temporal evolution of the salience of participants in a discussion. Our method can dynamically track how the relative importance of speakers evolve over time using graph based techniques. Speaker salience is computed based on the eigenvector centrality in a graph representation of participants in a discussion. Two participants in a discussion are linked with an edge if they use similar rhetoric. The method is dynamic in the sense that the graph evolves over time to capture the evolution inherent to the participants salience. We used our method to track the salience of members of the US Senate using data from the US Congressional Record. Our analysis investigated how the salience of speakers changes over time. Our results show that the scores can capture speaker centrality in topics as well as events that result in change of salience or influence among different participants.

Uncovering social semantics from textual traces: A theory‐driven approach and evidence from public statements ofU.S.Members ofCongress

“…When a set of words (or longer n ‐grams) frequently co‐occur it is possible topic models assume this is because of a systematic statistical tendency for these individual words to be drawn conditional on the presence of the “topic.” But it is also possible that their co‐occurrence is explained by the wholesale repetition of a chunk of text in which they are co‐present. This mimetic isomorphism in text has been observed in numerous cases, in the blogosphere (Leskovec, Backstrom, & Kleinberg, ; Simmons, Adamic, & Adar, ), social media (Adamic et al., ), political news reporting (Grimmer, ), discussions on Wikipedia and the Supreme Court (Danescu‐Niculescu‐Mizil, Lee, Pang, & Kleinberg, ), as well as within the U.S. Congress (Hassan et al., ).…”

A Low-Rank Approximation Approach to Learning Joint Embeddings of News Stories and Images for Timeline Summarization

“…Additionally, unsupervised techniques such as language modeling (Allan et al, 2001) have been used for temporal summarization. In recent years, ranking and graph-based methods (Radev et al, 2004b;Erkan and Radev, 2004;Mihalcea and Tarau, 2004;Fader et al, 2007;Hassan et al, 2008;Mei et al, 2010;Yan et al, 2011b;Yan et al, 2011a;Zhao et al, 2013;Ng et al, 2014;Zhou et al, 2014;Glavaš andŠnajder, 2014;Tran et al, 2015;Dehghani and Asadpour, 2015) have also proved popular for extractive timeline summarization, often in an unsupervised setting. Dynamic programming (Kiernan and Terzi, 2009) and greedy algorithms (Althoff et al, 2015) have also been considered for constructing summaries over time.…”

Entity Summarisation with Limited Edge Budget on Undirected and Directed Knowledge Graphs

“…More recent results include text summarization with latent semantic indexing [1], extraction of key phrases [3,23], and summarization of information from multiple documents [22]. Summarisation of scientific papers was studied by Hassan et al [9]. For all these approaches, their input is unstructured text, not semantic graphs.…”