Leveraging Natural Language Processing Applications Using Machine Learning: Text Summarization Employing Deep Learning

Leveraging Natural Language Processing Applications Using Machine Learning: Text Summarization Employing Deep Learning

Janjanam Prabhudas (VIT-AP University, India) and C. H. Pradeep Reddy (VIT-AP University, India)
DOI: 10.4018/978-1-5225-9643-1.ch016


The enormous increase of information along with the computational abilities of machines created innovative applications in natural language processing by invoking machine learning models. This chapter will project the trends of natural language processing by employing machine learning and its models in the context of text summarization. This chapter is organized to make the researcher understand technical perspectives regarding feature representation and their models to consider before applying on language-oriented tasks. Further, the present chapter revises the details of primary models of deep learning, its applications, and performance in the context of language processing. The primary focus of this chapter is to illustrate the technical research findings and gaps of text summarization based on deep learning along with state-of-the-art deep learning models for TS.
Chapter Preview


The basic unit of information called data which is alarming to the heights of uncertainties over period laid many challenges to researchers. It is well known that the data generated in the past 20 years is been produced in a week. As such data leaps over the gigantic curve, efficient models have evolved to process and analyze information for effective applications. Collaborating abundant information, computational abilities, and outstanding Artificial Intelligence techniques, the internet has upgraded from web 2.0 (dynamic web page interaction, independence to exhibit opinions, social blogging and media) to web 3.0 (leveraging information extraction, semantic web, artificial intelligence) elucidating the significance of processing Natural Language (NLP) by employing Machine Learning.

NLP is termed as an automatic process of analyzing, understanding and generating human language utilizing computational models. It has transformed from the early ages where batch processors are used to processing a single sentence that could consume 7 minutes to an era of search engines that process and analyze hundreds of web pages in not more than second. Today with redefined methods in NLP, few popular real-time applications that outperformed literature results are Machine Translation, Speech Recognition, Text Summarization, Virtual assistant chatbots, Sentiment analysis.

Since the language itself is ambiguous and unstructured, traditional methods and machines can’t able to cope up with unstructured data and insufficient in dealing with challenges arises due to lexical and structural ambiguities. Lexical ambiguities include similar words that occur in a sentence tent to behave differently according to its context, structural ambiguities are common in a language where a sentence can have multiple inferences or predictions. Consider the following Sentence, “Due to environmental disasters there was a big hole in the forest; a team of professionals is looking into it”. Here, humans can easily understand as “a team is working on the cause of the hole in the forest” but a system can interpret it as “a team is staring into the hole” which is of no meaning. To catch the clear meaning of the sentence the system should understand the language thoroughly and hence Machine Learning changed the perspective of earlier methods in dealing text.

There was a sudden leap of NLP applications in its performance when employing Machine Learning (ML) models. The notable achievements are during the 1990s, core NLP tasks are transformed via learning models from large quantities of available data. Core NLP tasks like tagging words with its grammatical structure using the Hidden Markov model, Named Entity Recognition, Syntax trees, Language modeling (N-Grams) using Markov chain process have achieved great improvements over traditional approaches. With the available Linguistic Data in the form of text and acoustics made available by Linguistic Data Consortium 1992, data was used to train and learn efficient ML models. Effective improvement of parallel processing made the evolution of statistical Natural Language Processing advanced applications like Sentiment analysis (Munjal, Kumar, Kumar, & Banati, 2019), text classification, spam email detection, automatic word correction, sentence recommendations in search engines.

Employing supervised or unsupervised popular and successful ML techniques like SVM, Logistic regression, Bayesian Classifier, Clustering for NLP tasks are based on words which are a statistical method. Though these methods succeeded in building successful applications like Apple Siri, IBM Watson, internally they rely on Bag of words (BOW), language modeling which are statistical approaches that work with individual word co-occurrences and probabilities. To make this informal these systems are like a learned parrot which repeat words without known actual meaning (Cambria & White, 2014)). The limitation with traditional ML models is they only consider statistical facts leaving semantics of language which can be overcome by engaging distributional semantics. The trends of NLP and the transformation of applications of NLP over the decades are seen in Figure 1.

Figure 1.

Trends of NLP and its transformation over the decades of time


Complete Chapter List

Search this Book: