公开(公告)号：US11741142B2

公开(公告)日：2023-08-29

申请号：US17589522

申请日：2022-01-31

Applicant: salesforce.com, inc.

Inventor： Haopeng Zheng ,  Semih Yavuz ,  Wojciech Kryscinski ,  Kazuma Hashimoto ,  Yingbo Zhou

IPC: G06F16/34 ,  G06F40/166 ,  G06N20/00 ,  G06F40/117 ,  G06F40/279

CPC classification number: G06F16/345 ,  G06F40/166 ,  G06N20/00 ,  G06F40/117 ,  G06F40/279

Abstract: Embodiments described herein provide document summarization systems and methods that utilize fine-tuning of pre-trained abstractive summarization models to produce summaries that more faithfully track the content of the documents. Such abstractive summarization models may be pre-trained using a corpus consisting of pairs of articles and associated summaries. For each article-summary pair, a pseudo label or control code is generated and represents a faithfulness of the summary with respect to the article. The pre-trained model is then fine-tuned based on the article-summary pairs and the corresponding control codes. The resulting fine-tuned models then provide improved faithfulness in document summarization tasks.

公开(公告)号：US20230054068A1

公开(公告)日：2023-02-23

申请号：US17589522

申请日：2022-01-31

Applicant: salesforce.com, inc.

Inventor： Haopeng Zheng ,  Semih Yavuz ,  Wojciech Kryscinski ,  Kazuma Hashimoto ,  Yingbo Zhou

IPC: G06F40/166 ,  G06F40/279 ,  G06F40/117 ,  G06N20/00

Abstract: Embodiments described herein provide document summarization systems and methods that utilize fine-tuning of pre-trained abstractive summarization models to produce summaries that more faithfully track the content of the documents. Such abstractive summarization models may be pre-trained using a corpus consisting of pairs of articles and associated summaries. For each article-summary pair, a pseudo label or control code is generated and represents a faithfulness of the summary with respect to the article. The pre-trained model is then fine-tuned based on the article-summary pairs and the corresponding control codes. The resulting fine-tuned models then provide improved faithfulness in document summarization tasks.