On using a quantum physics formalism for multidocument summarization

Abstract

International audienceMultidocument summarization (MDS) aims for each given query to extract compressed and relevant information with respect to the different query-related themes present in a set of documents. Many approaches operate in two steps. Themes are first identified from the set, and then a summary is formed by extracting salient sentences within the different documents of each of the identified themes. Among these approaches, latent semantic analysis (LSA) based approaches rely on spectral decomposition techniques to identify the themes. In this article, we propose a major extension of these techniques that relies on the quantum information access (QIA) framework. The latter is a framework developed for modeling information access based on the probabilistic formalism of quantum physics. The QIA framework not only points out the limitations of the current LSA-based approaches, but motivates a new principled criterium to tackle multidocument summarization that addresses these limitations. As a byproduct, it also provides a way to enhance the LSA-based approaches. Extensive experiments on the DUC 2005, 2006 and 2007 datasets show that the proposed approach consistently improves over both the LSA-based approaches and the systems that competed in the yearly DUC competitions. This demonstrates the potential impact of quantum-inspired approaches to information access in general, and of the QIA framework in particular

“Potentialities or possibilities”: Towards quantum information science?

“…However, it remains unclear as to how we should regard the application of quantum ideas in the information sciences, and how their “quantumness” is regarded. To give just a few examples, it has been described as: A metaphor (Bruza, Kitto, Nelson, & McEvoy, ; Wittek & Darányi, 2011a, 2011b) An analogy (Arafat, ; Piwowarski et al., 2010b; Widdows, ; Zhao et al., ) Inspired by quantum theory (Piwowarski et al., , ; Zhao et al., ) Quantum‐like (Di Buccio et al., ; Haven & Khrennikov, ) An abstract framework (Bruza et al., ) A scientific mirror (Arafat, ) …”

“…One relatively long‐established method is latent semantic analysis, originally a model for experimental studies of use and ambiguity of words, later adapted for IR, and extended to incorporate latent semantic indexing (Dearwester, Dumais, & Harshman, ; Ding, ; Landauer, McNamara, Dennis, & Kintsch, ). In essence, these methods deal with “sparse” document‐term matrices, that is, where each document has only a very few of the terms present in all the collection, by reducing the dimensionality to a much smaller number of “latent variables.” A quantum probability model of IR can subsume these methods, using the weights (measures of the contributions of two terms for describing a document or query) to measure associations between the uses rather than the semantics of terms (Piwowarski, Amini, & Lalmas, ).…”

Measuring and visualizing the contributions of Chinese and American LIS research institutions to emerging themes and salient themes

“…For example, Piwowarski et al (2012) proposed an extended technique based on the quantum information access framework, in which salient themes were first extracted from documents and then a summary was generated by extracting salient sentences from different documents for each theme.…”

Sentence simplification, compression, and disaggregation for summarization of sophisticated documents

“…From the earliest days of automatic text summarization through today, extraction of sentences from the original document has been the preferred approach (see, e.g., Barzilay & Elhadad, 1997;Erkan & Radev, 2004;Haghighi & Vanderwende, 2009;Hovy & Lin, 1998;Luhn, 1958;Mihalcea & Tarau, 2004;Piwowarski, Amini, & Lalmas, 2012;Yang & Wang, 2008).…”