Exploring Content Models for Multi-Document Summarization

By Aria Haghighi, Lucy Vanderwende. North American Chapter of the Association for Computational Linguistics, 2009

573 citations recorded by the source corpus.

We present an exploration of generative probabilistic models for multi-document summarization. Beginning with a simple word frequency based model (Nenkova and Vanderwende, 2005), we construct a sequence of models each injecting more structure into the representation of document set content and exhibiting ROUGE gains along the way. Our final model, HierSum, utilizes a hierarchical LDA-style model (Blei et al., 2004) to represent content specificity as a hierarchy of topic vocabulary distributions. At the task of producing generic DUC-style summaries, HierSum yields state-of-the-art ROUGE performance and in pairwise user evaluation strongly outperforms Toutanova et al. (2007)'s state-of-the-art discriminative system. We also explore HierSum's capacity to produce multiple 'topical summaries' in order to facilitate content discovery and navigation.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/9065d03256bcc962938f81eb795e70db214c459c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Learning to Score System Summaries for Better Content Selection Evaluation. (2017, background): “JS divergence is simply the symmetric version of the well-known Kullback-Leibler (KL) divergence (Haghighi and Vanderwende, 2009).”
Autofolding for Source Code Summarization (2014, methodology): “We compare two different content models for this task: a simple vector space model (VSM) and a topic model that, building on work in NLP summarization [9], endows different scopes (files, projects, and the corpus) with separate topics, allowing the model to separate out those tokens that are used mo”
A Simple, Fast Diverse Decoding Algorithm for Neural Generation (2016, methodology): “Features we consider include (1) average tf-idf score of constituent words in the generated output; (2) KLSum (Haghighi and Vanderwende, 2009), which reflects the topic distribution overlap between the entire input document and the generated output;13 and (3) backward probability p(X|Y ), i.e., the…”
Machine Reading Tea Leaves: Automatically Evaluating Topic Coherence and Topic Model Quality (2014, methodology): “…means of preprocessing a document collection, and the topics and per-document topic allocations are fed into downstream applications such as document summarisation (Haghighi and Vanderwende, 2009), novel word sense detection methods (Lau et al., 2012b) and machine translation (Zhao and Xing, 2007).”
Distraction-Based Neural Networks for Modeling Document (2016, methodology): “…summarization algorithms, which we listed in the table as Luhn [Luhn, 1958], Edmundson [Edmundson, 1969], LSA [Steinberger and Jezek, 2004], Lex-rank [Erkan and Radev, 2004], Text-rank [Mihalcea and Tarau, 2004], Sumbasic [Vanderwende et al., 2007], and KL-sum [Haghighi and Vanderwende, 2009].”
BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization (2019, background): “While prior work uses probabilistic topic models (Barzilay and Lee, 2004; Haghighi and Vanderwende, 2009) or relies on classiﬁers trained with sophisticated features (Yang et al., 2017), we focus on salient words and their occurrences in the input.”
Data Augmentation for Abstractive Query-Focused Multi-Document Summarization (2021, methodology): “2004), graph-based (Erkan and Radev 2004) or hierarchical LDA-style models (Haghighi and Vanderwende 2009), and variants of query-focused summarization (Dang 2005), that orient the summary around a given query (Daumé III and Marcu 2006; Zhao, Wu, and Huang 2009).”
Semantic Matching in Search (2014, methodology): “Haghighi and Vanderwende [82] apply topic modeling techniques to multi-document summarization in which the topic model is utilized for representing the content of a set of documents.”
Automatic Caption Generation for News Images (2013, background): “, 2007; Haghighi and Vanderwende, 2009), syntactic features (such as clause information or dependencies) (Barzilay and McKeown, 2005), and topic representations (Daumé III and Marcu, 2006; Haghighi and Vanderwende, 2009).”
Pattern Learning for Relation Extraction with a Hierarchical Topic Model (2012, methodology): “This model is very similar to the one used by Haghighi and Vanderwende (2009) in the context of text summarization. w (the observed variable) represents a pattern between two entities.”
Multi-document Summarization via Deep Learning Techniques: A Survey (2020, methodology): “Existing methods using traditional algorithms are based on term frequency-inverse document frequency (TF-IDF) [11, 131], clustering [52, 159], graphs [104, 158], and latent semantic analysis [8, 60].”
Improving the Estimation of Word Importance for News Multi-Document Summarization (2014, background): “The KLSUM summarizer (Haghighi and Vanderwende, 2009) aims at minimizing the KL divergence between the probability distribution over words estimated from the summary and the input respectively.”
Non-Expert Evaluation of Summarization Systems is Risky (2010, methodology): “The same idea has also been employed in various manual evaluations, for example by Haghighi and Vanderwende (2009), to directly compare the summaries of two different systems.”
System Combination for Multi-document Summarization (2015, methodology): “Because ﬁnding the summary with the smallest KL divergence is intractable, we employ a greedy method that iteratively selects an additional sentence that minimizes the KL divergence (Haghighi and Vanderwende, 2009).”
A Class of Submodular Functions for Document Summarization (2011, methodology): “, Daum´e et al. (2009) 7.67 - Vine, Daum´e et al. (2009) 8.24 -”
Guiding Extractive Summarization with Question-Answering Rewards (2019, methodology): “These include LexRank (Radev et al., 2004), SumBasic (Vanderwende et al., 2007), and KLSum (Haghighi and Vanderwende, 2009).”
Automatically Assessing Machine Summary Content Without a Gold Standard (2013, methodology): “In fact, KL divergence was greedily optimized in Haghighi and Vanderwende (2009) to obtain a high performance summarizer.”
Multi-News: A Large-Scale Multi-Document Summarization Dataset and Abstractive Hierarchical Model (2019, background): “…summarization have been both extractive (Carbonell and Goldstein, 1998; Radev et al., 2000; Erkan and Radev, 2004; Mihalcea and Ta-rau, 2004; Haghighi and Vanderwende, 2009) as well as abstractive (McKeown and Radev, 1995; Radev and McKeown, 1998; Barzilay et al., 1999; Ganesan et al.,…”