Chien-Sheng Wu, Caiming Xiong, Pontus Stenetorp, Wenhao Liu, Linqing Liu
We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
In this paper, we aim to improve abstractive dialogue summarization quality and, at the same time, enable granularity control. Our model has two primary components and stages: 1) a two-stage generation strategy that generates a preliminary summary sketch serving as the basis for the final summary. This summary sketch provides a weakly supervised signal in the form of pseudo-labeled interrogative pronoun categories and key phrases extracted using a constituency parser. 2) A simple strategy to control the granularity of the final summary, in that our model can automatically determine or control the number of generated summary sentences for a given dialogue by predicting and highlighting different text spans from the source text. Our model achieves state-of-theart performance on the largest dialogue summarization corpus SAMSum, with as high as 50.79 in ROUGE-L score. In addition, we conduct a case study and show competitive human evaluation results and controllability to humanannotated summaries. * Equal contribution. Work mainly done when Linqing Liu was an intern at Salesforce Research. Kelly: I still haven't received the rent money. Did you check with your bank? John: Yes. I definitely sent it last week. Kelly: But I still don't have it. Can you please check that you sent it to the right account. John: Ok. Give me 5 min. Kelly: OK John: I checked and the money did go out of my account last week. Kelly: What account number did you send it to? John: 44-1278 Kelly: No wonder! My account number is 44-1279. You sent it to someone else's account. John: ...! I'm really sorry! Kelly: I still need the rent money though. John: I'm really sorry I'll have to go to the bank tomorrow and ask if they can re-send it to the right account. Kelly: Thanks ! Longest-1 John said I'm really sorry I'll have to go to the bank tomorrow and ask if they can re-send it to the right account. BART-1 Kelly still hasn't received the rent money from John. John sent the rent money to the wrong account and will have to ask the bank to re-send it to the correct one tomorrow. BART Kelly still hasn't received the rent money. John sent it to the wrong account number 44-1278. John will go to the bank tomorrow and ask if they can re-send the money to the right account. Sketch: 1 #confirm haven't received the rent money check with your bank 2 none 3 #confirm check that you sent it to the right account 4 none 5 none 6 #abstain the money did go out of my account last week 7 #abstain did you send it to 8 none 9 #what sent it to someone else's account 10 none 11 #abstain need the rent money though 12 #abstain 'm really sorry i'll have to go to the bank tomorrow and ask if they can re-send it to the right account 13 none Summary: John sent the rent money to the wrong account last week. John will go to the bank tomorrow and ask if he can re-send the money to the correct account. Gold Kelly hasn't received the rent money, because John sent it to the wrong bank account. He will go to the bank to tackle the issue. Table 4: A test set example with generated summaries.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2105.14064")
get_code_for_paper("2105.14064")
have("2105.14064")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2105.14064.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2105.14064)
Connect an agent — have() is free.