On the Effectiveness of using Sentence Compression Models for Query-Focused Multi-Document Summarization

Proceedings of COLING 2012: Technical Papers, pages 457–474, COLING 2012, Mumbai, December 2012. On the Effectiveness of Using Sentence Compression Models for Query-Focused Multi-Document Summarization Y llias Chali Sadid A. Hasan University of Lethbridge, Lethbridge, AB, Canada ❝❤❛❧✐❅❝s✳✉❧❡t❤✳❝❛✱ ❤❛s❛♥❅❝s✳✉❧❡t❤✳❝❛ ABSTRACT This paper applies sentence compression models for the task of query-focused multi-document summarization in order to investigate if sentence compression improves the overall summariza- tion performance. Both compression and summarization are considered as global optimization problems and solved using integer linear programming (ILP). Three different models are built depending on the order in which compression and summarization are performed: 1) ComFirst (where compression is performed first), 2) SumFirst (where important sentence extraction is performed first), and 3) Combined (where compression and extraction are performed jointly via optimizing a combined objective function). Sentence compression models include lexical, syntactic and semantic constraints while summarization models include relevance, redundancy and length constraints. A comprehensive set of query-related and importance-oriented measures are used to define the relevance constraint whereas four alternative redundancy constraints are employed based on different sentence similarity measures using a) cosine similarity, b) syntactic similarity, c) semantic similarity, and d) extended string subsequence kernel (ESSK). Empirical evaluation on the DUC benchmark datasets demonstrates that the overall summary quality can be improved significantly using global optimization with semantically motivated models. KEYWORDS: Sentence compression, query-focused multi-document summarization, integer linear programming (ILP). 457 1 Introduction and Related Work Text summarization is a good way to compress large amount of information into a concise form by selecting the most important information and discarding redundant information (Mani and Maybury, 1999). Query-focused multi-document summarization aims to create a summary from the available source documents that can answer the requested information need (Chali and Hasan, 2012). Extraction-based automatic summarization has been a common practice over the years for its simplicity (Edmundson, 1969; Kupiec et al., 1995; Carbonell and Goldstein, 1998; Lin, 2003; Martins and Smith, 2009; Berg-Kirkpatrick et al., 2011). Extraction of the most important sentences to form a summary can degrade the summary quality if there exists a longer sentence with partly relevant information to prevent inclusion of other important sentences (due to summary length constraint) (Martins and Smith, 2009). Sentence compression can be a good remedy for this problem where the task can be viewed as a single-sentence summarization (Jing, 2000; Clarke and Lapata, 2008). Sentence compression1 aims to retain the most important information of a sentence in the shortest form whilst being grammatical at the same time (Knight and Marcu, 2000, 2002; Lin, 2003). Previous researches have shown that sentence compression can be used effectively in automatic summarization systems to produce more informative summaries by reducing the redundancy in the summary sentences (Jing, 2000; Knight and Marcu, 2002; Lin, 2003; Daumé III and Marcu, 2005; Zajic et al., 2007; Madnani et al., 2007; Martins and Smith, 2009; Berg-Kirkpatrick et al., 2011). However, most of these researches either focused on the task of single document summarization and generic summarization or did not consider global properties of …

Content truncated to 500 words