Generating Indicative-Informative Summaries with SumUM

Abstract

We present and evaluate SumUM, a text summarization system that takes a raw technical text as input and produces an indicative informative summary. The indicative part of the summary identifies the topics of the document, and the informative part elaborates on some of these topics according to the reader's interest. SumUM motivates the topics, describes entities, and defines concepts. It is a first step for exploring the issue of dynamic summarization. This is accomplished through a process of shallow syntactic and semantic analysis, concept identification, and text regeneration. Our method was developed through the study of a corpus of abstracts written by professional abstractors. Relying on human judgment, we have evaluated indicativeness, informativeness, and text acceptability of the automatic summaries. The results thus far indicate good performance when compared with other summarization technologies.

A Survey to Text Summarization Methods for Turkish

“…Saggion and Lapalme [26] present a text summarization system that takes a raw technical text as input and produces an indicative informative summary. SumUM first the topics of the document, and then elaborates on some of these topics according to the reader's interest.…”

Unsupervised Sentence Enhancement for Automatic Summarization

“…Previous studies on cut-and-paste summarization (Jing and McKeown, 2000;Saggion and Lapalme, 2002) investigate the operations that human summarizers perform on the source text in order to produce the summary text. Our previous work argued that current extractive systems rely too heavily on notions of information centrality (Cheung and Penn, 2013).…”

Automatic Summarization from Multiple Documents

“…Arguments take valid values from instances of an ontology (more on ontologies in section 2.5). A seemingly similar concept exists in [SL02], where we find the term template describing a structure indicative of a category of predicates. The categories of predicates and the templates themselves have been created using empirical examination based on a set of assumptions relative to the structure of technical articles: they are supposed to map unstructured text to structured, semantically important pieces of information.…”

“…For example, an agent is used to locate relevant context, another to represent text into propositions and another to remove redundancy. In [SL02], Saggion and Lapalme use conceptual information in terms of concepts and relations which they have extracted from a large corpus of scientific interdisciplinary articles and summaries. Cognitive analysis is also applied in [SYGH05], where the analysis returns a taxonomy as the required cognitive model.…”

“…In [SL02] we find an approach extracting structural information, depending on the formalisms used in technical articles concerning content order. This kind of formalism suggests that high level structure is in general Abstract, Introduction, Previous Work and so on, while in-paragraph text has a well-formed structure itself, with the first sentence being indicative of the paragraph meaning, and the last being a concluding remark for example.…”