Papers › EmailSum: Abstractive Email Thread Summarization

EmailSum: Abstractive Email Thread Summarization

30 Jul 2021ACL 2021 5arXiv:2107.14691archive 2025-07-28

Shiyue Zhang, Asli Celikyilmaz, Jianfeng Gao, Mohit Bansal

Recent years have brought about an interest in the challenging task of summarizing conversation threads (meetings, online discussions, etc.). Such summaries help analysis of the long text to quickly catch up with the decisions made and thus improve our work or communication efficiency. To spur research in thread summarization, we have developed an abstractive Email Thread Summarization (EmailSum) dataset, which contains human-annotated short (<30 words) and long (<100 words) summaries of 2549 email threads (each containing 3 to 10 emails) over a wide variety of topics. We perform a comprehensive empirical study to explore different summarization techniques (including extractive and abstractive methods, single-document and hierarchical models, as well as transfer and semisupervised learning) and conduct human evaluations on both short and long summary generation tasks. Our results reveal the key challenges of current abstractive summarization models in this task, such as understanding the sender's intent and identifying the roles of sender and receiver. Furthermore, we find that widely used automatic evaluation metrics (ROUGE, BERTScore) are weakly correlated with human judgments on this email thread summarization task. Hence, we emphasize the importance of human evaluation and the development of better metrics by the community. Our code and summary data have been made available at: https://github.com/ZhangShiyue/EmailSum

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

ZhangShiyue/EmailSum officialmentioned in paperpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Abstractive Text SummarizationEmail Thread Summarization

Datasets

Introduced by this paper, per the archive.

EmailSum

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Email Thread Summarization EmailSum (long) SemiSuptogether BertS 32.3 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) SemiSuptogether RLsum 40.67 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) SemiSuptogether ROUGE-1 44.08 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) SemiSuptogether ROUGE-2 14.06 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) SemiSuptogether ROUGE-L 31.17 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) T5base BertS 32.09 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) T5base RLsum 39.88 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) T5base ROUGE-1 43.81 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) T5base ROUGE-2 14.08 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) T5base ROUGE-L 30.47 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) Oracle BertS 26.31 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) Oracle RLsum 42.14 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) Oracle ROUGE-1 45.98 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) Oracle ROUGE-2 15.49 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (long) Oracle ROUGE-L 32.4 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) SemiSuptogether BertS 33.91 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) SemiSuptogether RLsum 33.7 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) SemiSuptogether ROUGE-1 36.98 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) SemiSuptogether ROUGE-2 11.21 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) SemiSuptogether ROUGE-L 28.76 #1 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) T5base BertS 33.9 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) T5base RLsum 32.76 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) T5base ROUGE-1 36.57 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) T5base ROUGE-2 10.56 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) T5base ROUGE-L 28.3 #2 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) Oracle BertS 22.32 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) Oracle RLsum 35.61 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) Oracle ROUGE-1 39.04 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) Oracle ROUGE-2 12.47 #3 of 3 Archive leaderboard report
Email Thread Summarization EmailSum (short) Oracle ROUGE-L 30.17 #3 of 3 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdafactorAttentionAttention DropoutBPEDense ConnectionsDropoutGated Linear UnitInverse Square Root ScheduleLayer NormalizationLinear LayerMulti-Head AttentionResidual ConnectionSentencePieceSoftmaxT5

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections