Enumeration of Extractive Oracle Summaries
This work addresses the problem of evaluating and improving extractive summarization systems for researchers, but it is incremental as it builds on existing oracle summary methods.
The paper tackles the limitations of extractive summarization by proposing an Integer Linear Programming formulation to enumerate all oracle summaries based on ROUGE-N, revealing that there is still room for performance improvement and that F-measures from these enumerated summaries have significantly stronger correlations with human judgment than those from single oracle summaries.
To analyze the limitations and the future directions of the extractive summarization paradigm, this paper proposes an Integer Linear Programming (ILP) formulation to obtain extractive oracle summaries in terms of ROUGE-N. We also propose an algorithm that enumerates all of the oracle summaries for a set of reference summaries to exploit F-measures that evaluate which system summaries contain how many sentences that are extracted as an oracle summary. Our experimental results obtained from Document Understanding Conference (DUC) corpora demonstrated the following: (1) room still exists to improve the performance of extractive summarization; (2) the F-measures derived from the enumerated oracle summaries have significantly stronger correlations with human judgment than those derived from single oracle summaries.