2026-10-12T03:21:56Z https://data.mendeley.com/oai

oai:data.mendeley.com/rxhj7g6y2k.3 2025-05-21T07:57:25Z

HASAN, MAHMUDUL MASBA: A Large-Scale Dataset for Multi-Level Abstractive Summarization of Bangla Articles Mendeley Data Our research hypothesis is to evaluate the effectiveness of different Bangla text summarization methods compared to the original text ('main'). The data shows that: - The average length of the main text is 2482.72 characters. - The average length of the summaries are: - sum1: 293.75 characters, - sum2: 506.10 characters, - sum3: 688.50 characters. The compression ratio of each summary method (summary length divided by main length) reveals that: - sum1's mean compression ratio is 0.14, - sum2's mean compression ratio is 0.24, and - sum3's mean compression ratio is 0.33. Notable findings: - sum1 appears to be the shortest summary on average, with a higher degree of compression. - sum2 produces summaries of medium length, while sum3 tends to generate the longest summaries. Data Gathering and Interpretation: The data can be interpreted to assess which method produces the most concise, yet meaningful, summaries. Researchers can use these findings to evaluate the trade-offs between summary length and completeness of information conveyed. Natural Language Processing Natural Language Generation Arean, Abdullah Ibne Hanif Khan, Md Mosaddek Dataset doi:10.17632/rxhj7g6y2k.3 oai:data.mendeley.com/rxhj7g6y2k.3 Creative Commons Attribution 4.0 International http://creativecommons.org/licenses/by/4.0 https://data.mendeley.com/datasets/rxhj7g6y2k 2025-05-21T07:57:25Z