Page 29 - 2024

Ph.D.
(Engineering & Technology)
METAHEURISTIC APPROACHES BASED TEXT SUMMARIZATION
FOR MULTIPLE DOCUMENTS

Ph.D. Scholar : Patel Praveshkumar Somabhai
Research Supervisor : Dr. Paresh M. Solanki

Regi. No.: 20276341005
Abstract :
The massive amount of textual information available on the Internet, including news,
blogs, websites, and user reviews continues to grow rapidly every day. Consequently,
users face significant challenges in finding the specific information they want to find, as it
is time-consuming to read and understand all the text they encounter in search results.
Moreover, much of the text contains repeated or irrelevant content. Therefore, there is a
need for text summarization and shortening text resources has become urgent and much
more important.

Multi-document summarization is a more challenging task compared to summarizing a
single document. It involves dealing with multiple documents that contain similar and
appropriate information on a specific topic or a single document that covers information
from various domains. The complexity arises from the presence of conflicting views,
biases, and thematic diversity in large document sets. Additional issues include handling
redundant information across multiple documents, compressing the content effectively
and selecting sentences efficiently for extraction at a reasonable speed and specific
amounts of words form multiple documents. These challenges can be addressed by
employing metaheuristic techniques in text summarization process which plays
important role to maintain control over relevance and coverage and remove redundancy
in summary. Therefore, there is a necessity to explore text summarization methods based
on metaheuristic approaches, as they can generate high-quality summaries from multiple
documents. In the proposed metaheuristic-based text summarization framework for
multiple documents, firstly identifying and implemented effective sentence scoring
features. Further based on selected sentence features a fitness function is constructed,
then firefly, cuckoo, ant colony and Teaching-learning based metaheuristic techniques are
applied for generation of precise text summary. Finally, generated summaries are
evaluated using ROUGE and BLEU matrices. Experiment results concluded that
incorporating metaheuristic techniques in multiple documents text summarization are
outperformed compared to existing approaches of multi-document text summarization.

Key words: Text Summarization, Natural language processing, Metaheuristic, Summary
Evaluation, Firefly, Cuckoo Search, Teaching Learning, ROUGE, BLEU.

01