International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 08 | Aug 2024
p-ISSN: 2395-0072
www.irjet.net
Combining Deep Learning and Heuristic Search for Efficient Text Summarization Divya Beeram San Jose State University, California, USA. -------------------------------------------------------------------------***-----------------------------------------------------------------------Abstract— This work is a research project aimed at studying how deep Learning based on advanced DNN architectures can be integrated with heuristic search algorithms to make text summarization faster and of higher quality. Existing summarization methods are traditionally manual or rule-based, and they have drawbacks in terms of time consumption and the need for several essential information. This work introduces a novel feel that leverages deep learning models (in the form of neural networks) with heuristic search methods to create compact and informative summaries. Our approach leverages the representation learning and attention mechanism present in deep-learning models for identifying keywords and key sentences from a longer text; considering the relevance to the topic adds important heuristic information to choose how many sentences are needed. We hope the integration can improve text summarization in general, and we aim to use this method in news, research articles, and other types of content. Keywords— Deep Learning, Heuristic, Summarization, Leverages, Mechanism I. INTRODUCTION Today, the volume of information available to us is magnified exponentially. Its huge amount of data makes it difficult for people to keep up with. Consequently, there is a high demand for text summarization methods that can compress large volumes of textual corpus to their meaningful essence as rapidly as possible. That is the point where deep Learning and heuristic search are integrated [1]. Deep Learning, as a category of machine learning, leverages artificial neural networks to process data in complex forms so that machines can understand them better. It is used for some of the most common tasks in machine learning, including image classification, speech recognition and natural language processing. C) Heuristic search: This is a problem-solving methodology used to solve problems from intuition or a rule of thumb and reach optimal solutions in a considerably large state space [2]. This essay focuses on the possible benefits of integrating these two approaches for successful text summarization using deep Learning and heuristic search. First, we introduce the separate approaches and how their combination enables us to enhance summarization. Deep Learning performs best in NLP tasks, such as text summarization. The traditional text summarization methods divide the process of Figure 3 into two phases, mainly extractive and abstractive summaries. Extractive summarization involves pulling key sentences or phrases from the original text and mashing them together to create a summary [3]. The system understands the meaning and generates a summary by paraphrasing/rephrasing human-readable text while Convergent summarization. Deep learning models can learn from that (giant) data; the performance of a deep model is also much greater than the traditional method. Text summarization usually uses Convolutional neural networks (CNNs) and recurrent neural networks (RNNs). TextCNNs can extract features from text using convolutional filters. RNNs are good at dealing with sequential data and generating abstractive summaries by predicting the next word according to context [4]. The convolutional neural attention model, a combination of these two techniques, has also been implemented in text summarization. It uses CNN to extract features from the input text and attention layer to decide which portion of the text is more relevant for a summary. To this day, that mix has proved effective at producing accurate summaries. Heuristic search is a methodology that helps to solve large, complex problems by breaking them down into simpler, more manageable steps [5]. On the other hand, in the summarization of text, much information can be sought from a corpus using realistic search algorithms. A* is one of the heuristic search algorithms that uses a cost function to evaluate potential solutions and get closer towards finding an optimal solution. Heuristic search algorithms have been employed in extractive summarization approaches, and most of them fall under the umbrella field of text summarization. The algorithms examine the text and rank sentences and phrases by their importance to this job. This makes it easy to choose the important items and summarize them. While the approach to applying deep Learning with the heuristic search for text summarization is relatively new, it has already exhibited reasonably good results. This combination enables to leverage the advantages of both methods, leading to more efficient and accurate summaries [6]. The biggest issue of concern in text summarization is that tons and tonnages (i.e., large amounts) of information are available. Deep learning algorithms can work through massive data sets to determine the crucial words within text. This is unlikely to be efficient, especially in the case of extensive data sets. At the same time,
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 23