It's hard to measure the summarization's quality totally based on ROGUE. For academic papers, it will always lose key points.

Cons of ROGUE

It's often used to measure the semantic similarity of the summarization and the whole document.

University of Michigan - Ann Arbor

1. Stream Document Summarization
2. Timeline Summarization
3. Extreme Long Document Summarization
4. Dialogue Summarization
5. Query-based Summarization

Text Summarization Incorporating Document Structure

Dialogue Discourse-Aware Graph Model and Data Augmentation for Meeting Summarization

QMSum: A New Benchmark for Query-based Multi-domain Meeting Summarization

It proposes a new paradigm of summarizing the meetings. It only collects 232 meetings. It separates the summarization generation task into several QA tasks, which is called query-based summarization.

QMSum

Dialogue discourse indicates relations between discourse units in a conversation. They propose a dialogue units to do data augmentation and a task to detect those Dialogue discourses

DDAMS

Recall-oriented understudy for gisting evaluation(ROGUE)

Hierarchical Transformers for Multi-Document Summarization

Scalable, Semi-Supervised Extraction of Structured Information from Scientific Literature

CitationAS: a summary generation tool based on clustering of retrieved citation content

The summary is hierarchically organized along with one or more organizational principles such as time, location, entities, or events.

Hierarchical Summarization

https://homes.cs.washington.edu/~mausam/papers/acl14.pdf

Hierarchical Summarization: Scaling Up Multi-Document Summarization

Toward Abstractive Summarization Using Semantic Representations

They generate some AMR Graphs based on some rules and use them as data augmentation for helping summarization.

Learn Before

Related

Learn After