Academia.eduAcademia.edu

Figure 2 – uploaded by Harish Karnick

See full PDFdownloadDownload figure
Figure 3: Attention Probability for decoding on a SUMPUBMED example where the attention corresponding to word ‘present’ in the generated summary is shown.

Figure 3 Attention Probability for decoding on a SUMPUBMED example where the attention corresponding to word ‘present’ in the generated summary is shown.

Related Figures (8)

Figure 1: SUMPUBMED creation pipeline.
Figure 1: SUMPUBMED creation pipeline.
Table 1: Average number of sentences and words in the abstract and text in the three SUMPUBMED versions
Table 1: Average number of sentences and words in the abstract and text in the three SUMPUBMED versions
Table 3: Pearson’s correlation between ROUGE scores and human ratings on SUMPUBMED’s noun-phrase version
Table 3: Pearson’s correlation between ROUGE scores and human ratings on SUMPUBMED’s noun-phrase version
Table 5: ROUGE scores on CNN-Dailymail (CNN-DM) and DUC 2001 dataset (DUC) using seq2seq models
Table 5: ROUGE scores on CNN-Dailymail (CNN-DM) and DUC 2001 dataset (DUC) using seq2seq models
Table 6: ROUGE scores of noun-phrase SUMPUBMED version using a seq2seq model of varying decoding steps
Table 6: ROUGE scores of noun-phrase SUMPUBMED version using a seq2seq model of varying decoding steps
Table 7: Results for TextRank an Extractive Summarization approach on hybrid version of the SUMPUBMED.
Table 7: Results for TextRank an Extractive Summarization approach on hybrid version of the SUMPUBMED.
Table 8: ROUGE scores on hybrid version of the SUMPUBMED using Hybrid model: TextRank + seq2seq models
Table 8: ROUGE scores on hybrid version of the SUMPUBMED using Hybrid model: TextRank + seq2seq models
Table 9: ROUGE comparison on SUMPUBMED. seq?2seq abstractive methods’ target summary is of 250 words
Table 9: ROUGE comparison on SUMPUBMED. seq?2seq abstractive methods’ target summary is of 250 words
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved