Artificial Intelligence-Based Multi-Document Extractive Summarization Using Sentence-Level Features
SPU - Journal of Science, Technology and Management Research · 01.07.2026 · 153-160
DOI: 10.63766/spujstmr.26.000082 ↗Kurz gefasst
Fachartikel in der Zeitschrift SPU - Journal of Science, Technology and Management Research, online erschienen am 01.07.2026. Verfasst von Karina Patel und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The goal of multi-document text summarizing is to automatically generate a succinct yet thorough summary from multiple connected sources. The growing volume of digital text data is making the task of document summarization more challenging. More effective summarization is now possible because to significant advancements in artificial intelligence techniques and pre-trained language models; nonetheless, the majority of deep learning approaches require a significant amount of processing power and large training datasets. This work suggests a multi-document text summarizing method powered by artificial intelligence that makes use of supervised learning and sentence feature modeling. The proposed system begins with heavy preprocessing that involves lemmatization, normalization, stop-word removal, and sentence segmentation. Statistical and structural features, including the TF–IDF representation, length of the sentence, normalized position of the sentence, and numeric token count, are extracted to denote sentence significance. Like other supervised extraction algorithms discussed in previous literature, a Logistic Regression classifier is created using a balanced dataset to classify sentences into significant or insignificant categories. Probabilities are used for ranking of selected sentences, while cosine similarity is applied under a certain word limit constraint to avoidredundancies. The suggested approach demonstrates effective accuracy while needing very little processing effort, according to experimental examination of the DUC 2003 corpus. Based on the ROUGE measure, the model produces ROUGE-1, ROUGE-2, and ROUGE-L scores of 0.1722, 0.0306, and 0.1048, respectively. It demonstrates how basic AI models combined with useful sentence-based features could be a reliable benchmark for information extraction from several publications. To increase the effectiveness of summaries, the authors intend to look into contextual representation algorithms like Sentence-BERT and implement human evaluation standards in subsequent studies.
Zitieren
APA
Patel, K., & Serasiya, S. (2026). Artificial Intelligence-Based Multi-Document Extractive Summarization Using Sentence-Level Features. SPU - Journal of Science, Technology and Management Research, 153-160. https://doi.org/10.63766/spujstmr.26.000082
Vancouver
Patel K, Serasiya S. Artificial Intelligence-Based Multi-Document Extractive Summarization Using Sentence-Level Features. SPU - Journal of Science, Technology and Management Research. 2026:153-160. doi:10.63766/spujstmr.26.000082
BibTeX
@article{Patel2026,
author = {Patel, Karina and Serasiya, Shilpa},
title = {Artificial Intelligence-Based Multi-Document Extractive Summarization Using Sentence-Level Features},
journal = {SPU - Journal of Science, Technology and Management Research},
year = {2026},
pages = {153--160},
doi = {10.63766/spujstmr.26.000082}
}
EUVIMED