SimSum: an empirically founded simulation of summarizing
Section snippets
Introduction and purpose of the paper
The main achievement of summarizing is the reduction of available contents to the most relevant items. Often, but not always, the task includes the goal-oriented acquisition of information from external sources such as texts. Otherwise, memory may supply the material for the summary. Most of the time, the resulting summary (or abstract) is a short text.
The SimSum (Simulation of Summarizing) system does what its name promises: it simulates summarizing of human experts and thus produces a
Current research in automatic summarization
The best place to start an exploration of current international research in summarizing is the webpage of the Ottawa Text Summarization Project (http://www.site.uottawa.ca/tanka/ts.html). It provides information about the local projects and it links to bibliographies, webpages of active researchers, text corpora, industrial systems, conferences and new papers in summarizing.
Summarization is currently a busy field. The increase in activity is mostly motivated by the Internet. Users need
SimSum development
The SimSum simulation has been added to the empirical model of summarizing for scientific and presentational purposes:
- •
As usual, the computer model serves to explain and check the empirical cognitive model which is its foundation.
- •
lt prepares a cognitively grounded approach to automatic summarizing, something like agents in the WWW accepting a user’s query and bringing up a reasonably short statement (a summary) of the knowledge available in response to it.
- •
To its users of today, SimSum shows in a
The test setting
During my regular introductory course to content analysis, polytechnic freshmen tested SimSum as an instrument for learning basic concepts about summarizing. Students did not enrol to a specific experiment. SimSum simply replaced the combined media I used in earlier courses: a tape recorder and the written thinking-aloud protocol, the source paper, the summarizer’s notes and the summary in different stages of progress. This makeshift multimedia show had provoked ‘aha’-experiences in students
Conclusion
Advancing the scientific frontiers of text summarization presupposes more knowledge about the way summarization works. The main fruit of the empirical investigation behind SimSum is an image of the summarization process which is detailed enough to lay the foundations for a simulation. Since the resulting summarization model incorporates the know-how of human experts, it has good prospects of presenting powerful techniques. Summarizing by cooperating cognitive agents seems to be such a principle.
Acknowledgements
The SimSum development was funded under grant F 916.00 by the German Federal Ministry of Education and Research. The German Science Foundation supported the empirical investigation under grant En 186/1-3.
References (53)
- et al.
Evolution of blackboard control architectures
Expert Systems with Applications
(1994) - et al.
How to implement a naturalistic model of abstracting: four core working steps of an expert abstractor
Information Processing & Management
(1995) Automated discourse generation using discourse structure relations
Artificial Intelligence
(1993)- et al.
Innovations in text interpretation
Artificial Intelligence
(1993) Generating summaries from event data
Information Processing & Management
(1995)- et al.
Generating concise natural language summaries
Information Processing & Management
(1995) Constructing literature abstracts by computer: Techniques and prospects
Information Processing & Management
(1990)- et al.
Causal relatedness and importance of story events
Journal of Memory and Language
(1985) - André E., Graf W., Müller J., Profitlich H.-J., Rist T., & Wahlster W. (1997). AiA: Adaptive communication assistant...
- Barker K., Chali Y., Copeck T., Matwin S., & Szpakowics S. (1998). The design of a configurable text summarization...
The psychology of written composition
Salience-based content characterization of text documents
Dynamic presentation of document content for rapid online skimming
Summarizing information
Verbal reports as data
Psychological Review
Protocol analysis: Verbal reports as data
Forward and backward reasoning in automatic abstracting
A propositional language for text representation
Evaluating importance: A step towards text summarization
Summarization of documents that include graphics
Knowledge-based indexing in SIMPR: Integration of natural language processing and principles of subject analysis in an automated indexing system
Journal of Document and Text Management
The discovery of grounded theory: Strategies for qualitative research
Cited by (10)
Summarizing court decisions
2007, Information Processing and ManagementSummarization from medical documents: A survey
2005, Artificial Intelligence in MedicineSummarizing speech by contextual reinforcement of important passages
2012, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)Human information interaction: An ecological approach to information behavior
2012, Human Information Interaction: An Ecological Approach to Information BehaviorLearning predicate insertion rules for document abstracting
2011, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)