Query-focused multi-document summarization: automatic data annotations and supervised learning approaches

Abstract

In this paper, we apply different supervised learning techniques to build query-focused multi-document summarization systems, where the task is to produce automatic summaries in response to a given query or specific information request stated by the user. A huge amount of labeled data is a prerequisite for supervised training. It is expensive and time-consuming when humans perform the labeling task manually. Automatic labeling can be a good remedy to this problem. We employ five different automatic annotation techniques to build extracts from human abstracts using ROUGE, Basic Element overlap, syntactic similarity measure, semantic similarity measure, and Extended String Subsequence Kernel. The supervised methods we use are Support Vector Machines, Conditional Random Fields, Hidden Markov Models, Maximum Entropy, and two ensemble-based approaches. During different experiments, we analyze the impact of automatic labeling methods on the performance of the applied supervised methods. To our knowledge, no other study has deeply investigated and compared the effects of using different automatic annotation techniques on different supervised learning approaches in the domain of query-focused multi-document summarization.

Scientific document summarization via citation contextualization and scientific discourse

“…(ii ) solving an optimization problem [20,8,27]: these approaches cast the summarization problem as an optimization problem where an objective function needs to be optimized with respect to some constraints. (iii ) supervised models [60,25,18], where selection of sentences in the summary are learned using a supervised framework. (iv ) graph based [29,53,62]: these approaches seek to find the most central sentences in a document's graph where sentences are nodes and edges are similarities.…”

MR&MR-SUM: MAXIMUM RELEVANCE AND MINIMUM REDUNDANCY DOCUMENT SUMMARIZATION MODEL

“…Up to now, various extraction-based techniques have been proposed for generic document summarization. 28 In automatic document summarization, the selection process of the distinct ideas included in the document is called diversity. The diversity is very important evidence serving to control the redundancy in the summarized text and produce more appropriate summary.…”

Complex Question Answering: Homogeneous or Heterogeneous, Which Ensemble Is Better?

“…We use query-focused supervised extractive multi-document summarization technique for this purpose [1][2][3]. Ensemble methods are learning algorithms that construct a set of classifiers and then classify new data points by taking a (weighted) vote of their predictions [4].…”

“…Supervised classifiers are typically trained on data pairs, defined by feature vectors and corresponding class labels. We use an automatic labeling approach to annotate the training data using ROUGE [1,3,9]. From each sentence of the training (and testing) data, we extract different query-related features and importance-oriented features such as: n-gram overlap, Longest Common Subsequence (LCS), Weighted LCS (WLCS), skip-bigram, exact word overlap, synonym overlap, hypernym/hyponym overlap, gloss overlap, Basic Element (BE) overlap, syntactic tree similarity measure, position of sentences, length of sentences, Named Entity (NE) match, cue word match and title match [1,3,5,13].…”

“…We take no action for the negatively classified sentences so that they could fall back during ranking. The combined weighted votes of all the classifiers are used to rank the sentences to produce 250-word summaries [1].…”