- Journal of Computational Intelligence
- University of British Columbia Vancouver, Canada
- First Author: Giuseppe Carenini, Associate Professor of Computer Science, University of British Columbia
Work series on same topic of Giuseppe Carenini
- A framework for summarizing a corpus of evaluative
documents about a single entity by a natural language summary
Series of previous work by same first author
- first paper: extract & organize evaluative info
- 1st study: compared the two summarizers: extractive vs. abstractive. The summarizers have different strengths and weaknesses.
- 2nd study: when corpus is controversial, Evaluative text abstraction
tends to be more effective than extraction.
- 3rd study: assessed the effectiveness of our user tailoring strategy. User tailored summaries are more informative than un-tailored ones.
TASK of TEXT SUMMARIZATION
- a task of annual Document Understanding Conference, DUC (renamed to Text Analysis Conference (TAC) since 2008)
- most of previous work focused on factual, edited text and biographies, for which the goal is extract & order the most informative sentences in a domain-independent way, while avoiding repetition
when documents contain inconsistent info, the summarizer needs to identify overlaps and inconsistencies and produce summary that explains those inconsistencies.
- basic steps
- extract & analyze the evaluations expressed in the corpus (been studied extensively)
- summarize them (received less attention)
extraction-based summarizer: extract