HYBRID APPROACH FOR TEXT SUMMARIZATION USING DEEP LEARNING

HYBRID APPROACH FOR TEXT SUMMARIZATION USING DEEP LEARNING

1. VISVESVARAYA TECHNOLOGICAL UNIVERSITY BELAGAVI, KARNATAKA- 590018
GOVERNMENT ENGINEERING COLLEGE
CHALLAKERE -577522
Batch :15
23/05/2025 Hybrid approach for Text Summarization using Deep Learning 1
Major Project Phase – 1 Presentation on
“Hybrid approach for Text Summarization using
Deep Learning”
2025-26
Under Guidance of :
Asst. Prof. MANJUNATHA S
Presented By :
KIRAN MALI (4EG22CS017)
SUDEEP T TALWAR (4EG22CS044)
SUDEEP VISHWAKARMA (4EG22CS045)
VIVEK B M (4EG22CS053)

2. Contents
2
1.ABSTRACT
2.Introduction
3.Objectives
4.Literature review
5.Conclusion
6.References
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

3. 1. ABSTRACT
3
▪Text summarization is the process of condensing large amounts of text into a concise and
meaningful summary while retaining key information.

▪Traditional summarization techniques are classified into extractive and abstractive methods.
Extractive summarization selects important sentences from the original text.
▪ Abstractive summarization generates new sentences to improve readability but may
introduce errors or lose crucial details.
▪This project proposes a hybrid text summarization approach that combines both extractive
and abstractive techniques using deep learning.
▪This hybrid approach has the potential to improve information retrieval, making it useful for
applications like news summarization, legal document analysis, and academic research.
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

4. 2. Introduction
4
•Text summarization aims to filter key information from a text, providing a condensed and
coherent representation.
•Two methods for text summarization are extractive and abstractive methods, each with its
strengths and limitations.

•Our approach leverages state-of-the-art neural network architectures, including Recurrent
Neural Networks (RNNs) and Transformer-based models, to achieve a balanced text summary.
•The main component of our hybrid model employs neural networks to identify and extract
salient sentences or words from the source text
23/-5/2025 Hybrid approach for Text Summarization using Deep Learning

5. 2. Introduction
•Complementing this, the abstractive component utilizes Transformer-based models, which excel
in understanding context and generating human-like language.
•Even though several methods exist to solve the text summarization problem as a classification
problem using deep learning, they have some limitations. Firstly, these methods trained their
models based on only one way of representing sentences in a text. Secondly, their models were
evaluated based on classification performance or summary quality rather than both.
5
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

6. 3. Objectives
1.Combine content retention and coherence in summaries.
2.Support both extraction and abstraction for summarization.
3.Ensure context-aware and user-centric summary generation.
4.Support cross-domain applicability of hybrid models.
5.The audio application to spell summarized text.
6
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

7. 4. Literature review
“Abstractive text summarization using LSTM-CNN based deep learning”.[2]
Authors: S. Song and T. Ruan
Year of Publication:2019

Publisher:Springer
Description of work:
ATSDL Approach: ATSDL employs LSTM-CNN for Abstractive Text Summarization, extracting semantic phrases and
generating summaries in two stages. Superior Performance: Experimental results on CNN and Daily Mail datasets demonstrate
ATSDL's competitive edge, outperforming existing models in semantic and syntactic aspects for quality evaluation.
7
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

8. 4. Literature review
“A Hybrid Approach for Arabic Text Summarization Using Domain Knowledge and Genetic Algorithms” [3]
Authors: Al-Radaideh, Q. A., & Bataineh,
Year of Publication: 2018
Publisher:Springer
Description of work:
ASDKGA outperforms three existing Arabic text summarization methods, as evidenced by its average F-measure of
0.605 in summarizing Arabic political documents using the ROUGE framework on Arabic Summaries Corpus. The evaluation
results indicate that ASDKGA demonstrates promise for generating concise and informative summaries, particularly for Arabic
political content, with effectiveness showcased at a compression ratio of 40%.
8
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

9. 4. Literature review
“COSUM: Text summarization based on clustering and optimization” [4]
Authors: N. R., Abdi, & Idris . N
Year of Publication: 2019
Publisher : Article
Description of work:
The document covers various text summarization techniques, including clustering algorithms, and unsupervised
learning, emphasizing the importance of avoiding redundancy, ensuring topic coverage, and providing diversity in
summaries.The COSUM model, a two-stage sentence selection approach based on clustering and optimization, is proposed in the
document as a solution to challenges in text summarization, demonstrating superior performance compared to existingmethods.
9
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

10. 4. Literature review
“Learning-Free Unsupervised Extractive Model” [5]
Authors: Myeongjun Jang and Pilsung Kang
Year of Publication: 2021
Publishe r: IEEE
Description of work:
Document outlines challenges: Traditional methods lack semantic depth, while deep learning models demand
significant resources and labeled data. Proposed solution: Introduces a training-free summarization framework using principal
component analysis, covering both extractive and abstractive approaches.
10
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

11. 4. Literature review
“HTS-DL: Hybrid Text Summarization System using Deep Learning” [6]
Authors: Chitra Dadkhah and Nasim Tohidi.
Year of Publication: 2022
Publisher : IEEE
Description of work:
HTS-DL Model: Encoder-Decoder with attention mechanism, addressing limitations, but slower learning. ETS
Integrates Textrank algorithm for Enhanced Text Summarization to reduce input length and Textrank used for unsupervised
sentence extraction, providing an efficient NLP summarization solution. This output has been fed to Encoder-Decoder with
attention mechanism to perform abstractive text summarization using pre-defined datasets.
11
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

12. 4. Literature review
“Extractive Summarization - A Comparison of Pre-Trained Language Models and Proposing a Hybrid Approach”
[7]
Authors: Vanshika Taneja, Sanjana B ,Dinesh Singh,
Year of Publication: 2023
Publishe r: IEEE
Description of work:
The context outlines algorithms for summarization, including Word2Vec, TextRank, CNN-based methods, and a blend
of extractive and abstractive techniques. It emphasizes Sum Evaluation for comprehensive modelassessment. By using BERT
and BART for text summarization, it has achieved highest accuracy among others models.
12
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

13. 4. Literature review
1323/05/2025 Hybrid approach for Text Summarization using Deep Learning
“Deep Reinforcement Learning for Sequence-to-Sequence Models” [9]
Authors: Yaser Keneshloo, TianShi, Naren Ramakrishnan, and Chandan K. Reddy
Year of Publication: 2019
Publisher : IEEE
Description of work:
Seq2Seq Challenges: Despite widespread use in applications like translation and summarization, seq2seq models face
issues like exposure bias and training-testing inconsistency. Reinforcement Learning Solution: Researchers propose using
reinforcement learning to texposure bias and enhance consistency, enabling seq2seq models to better handle long-term
dependenciesindata.

14. 4. Literature review
14
“A Systematic Literature Review on Text Generation Using Deep Neural Network Models” [11]

Authors:, Sher Muhammad D, and Abdullah S
Year of Publication: 2022
Publisher : IEEE
Description work
This document explores the challenges of text generation using deep learning models, highlighting issues like
vanishing gradients in RNNs, evaluation methods, and the lack of comprehensive reviews on datasets, metrics, and languages.
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

15. 4. Literature review
15
“Abstractive Summarization: An Overview of the State of the Art” [15]
Authors: Som Gupta, S. K Gupta
Year of Publication:2019
Publisher : IEEE
Description work
The paper provides a comprehensive review of the field of abstractive summarization, discussing various techniques, tools,
evaluation methods, challenges, and future research directions. It categorizes the methods into structure-based, semantic-based,
and deep learningbased approaches.
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

16. 4. Literature review
16
“Automated Abstractive Text Summarization using Deep Learning” [16]
Authors: G. Karuna, M. Akshith
Year of Publication:2023
Publisher : IEEE
Description work
The paper presents a model for automated abstractive text summarization using deep learning techniques,
specifically the Sequence-to-Sequence (Seq2Seq) model with Long Short-Term Memory (LSTM) networks. The approach
preprocesses large text datasets, encodes them into vectors using an encoder, and generates summaries through a decoder.
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

17. 5. Conclusion
17
Hybrid Text Summarization using Deep Learning represents a compelling synthesis of extractive
and abstractive approaches, harnessing the strengths of both methodologies. By combining the precision
of extractive techniques with the creativity of abstractive methods, this hybrid model achieves a balanced
summarization output. The synergy enables the model to preserve factual accuracy while generating
coherent and contextually rich summaries, contributing to improved information comprehension.

23/05/2025 Hybrid approach for Text Summarization using Deep Learning

18. 6. References
18
[1] Jigisha M Narrain, Vanshika Taneja, Sanjana B Atrey, Jahnavi Sivaram, Dinesh Singh, “Extractive Summarization - A
Comparison of Pre-Trained Language Models and Proposing a
Hybrid Approach”, IEEE,2023.
[2] Mengli Zhang, Gang Zhou, wanting Yu, Ningbo Huang, Wenfen Liu, "A Comprehensive Survey of Abstractive Text
Summarization Based on Deep Learning", Computational Intelligence and Neuroscience, Volume 2022.
[3] Wojciech Kryściński, Bryan McCann, Caiming Xiong, Richard Socher, "Evaluating the Factual Consistency of Abstractive
Text Summarization”, Salesforce Research ,2019. [11] Nikolaos Giarelis, Charalampos Mastrokostas, Nikos Karacapilidis,
MDPI, Basel, Switzerland, 2019.
[4] Som Gupta, S. K Gupta, “Abstractive Summarization: An Overview of the State of the
Art”, 2019.
23/05/2025 Hybrid approach for Text Summarization using Deep Learning

19. 23/05/2025 19
Thank you
Hybrid approach for Text Summarization using Deep Learning