From Text Summarisation to Style-Specific Summarisation for Broadcast News

Abstract

Abstract. In this paper we report on a series of experiments investigating the path from text-summarisation to style-specific summarisation of spoken news stories. We show that the portability of traditional text summarisation features to broadcast news is dependent on the diffusiveness of the information in the broadcast news story. An analysis of two categories of news stories (containing only read speech or some spontaneous speech) demonstrates the importance of the style and the quality of the transcript, when extracting the summary-worthy information content. Further experiments indicate the advantages of doing style-specific summarisation of broadcast news.

Broad coverage paragraph segmentation across languages and domains

“…Furthermore, sometimes the output of a speech recognizer needs to be processed automatically by applications such as information extraction or summarization. Most of these applications port techniques developed for written texts to spoken texts (e.g., Christensen et al [2004]) and therefore require input that is punctuated and broken into paragraphs. While there has been some research on finding sentence boundaries in spoken text [Stevenson and Gaizauskas 2000], relatively little research has examined the automatic insertion of paragraph boundaries.…”

Automatic Summarization

“…That can include features from the speech signal or features that capture the dialog nature of exchanges. Techniques for automatic summarization of spoken input have been developed for many applications, most notably for summarization of broadcast news [31,79,124], dyadic conversations [71,72], meetings [63,142,214] and lectures/presentations [68,77,93,168,230].…”

“…The most widely applied techniques for transcript summarization [31,77,142,229] have been Maximal Marginal Relevance [25] and Latent Semantic Analysis [69]. Variations of frequency features (total frequency, TF * IDF) and positional features, as well as length features, similar to the ones developed for text summarization are employed in much of the existing work on speech summarization.…”

How Ontology Based Information Retrieval Systems May Benefit from Lexical Text Analysis

“…Other methods focus on the identification of text portions related to the document title [29]. Most automatic text summary methods are based on a supervised learning process, that requires human intervention to set an adequate training corpus [57,7]. [47] proposes an unsupervised method to extract key phrases in a summarization context.…”