Text Summarization and Its Types: A Literature Review

Introduction

In the present era, nobody has enough time to go through the full documentation to get the full meaning. So, there is an immense need for automatic text summarization to save time and to make it easy for humans. Text summarization helps in creating bulletin, headings, summary, brief description, finding out the essential words, and so on. Text summarization helps in creating a summary of the data set without altering the actual meaning of data, which includes essential information as well. The need for text summarization can also be understood by an example suppose someone wants to read documents which are related to text mining from a vast miscellaneous database, and he starts reading all documents one by one and hence consumes much time, but if there is a list of all headings, then the person can directly read materials related to text mining. Other examples are – news headlines in the newspaper, the title of a book, and many more. Text summarization is essential because a massive growth in the information requires high maintenance.

Abstractive text summarization and extractive text summarization are the two major categories in text summarization. Abstractive summarization refers to recreate the whole document in a few words or lines, which may include new words as well. Extractive summarization refers to extract the critical words or lines from the original document. To clear the difference between abstractive and extractive to consider an example – a man is reading a document, and while reading, he is highlighting the main sentences to remember the vital part; after that, he writes the whole document using a pen to make notes. The highlighted sentences and pen-written notes both describes the original document in brief but are placed in different categorization. A highlighted summary is an example of extractive summarization, and a pen-written summary is an abstractive summarization. Machines or tools used to create a summary are known as summarizers. These tools are language-dependent.

Summarizers take text data as input to produce a summary. If the tool produces a summary in one or more languages other than the original language, then it is referred to as multilingual text summarization. The main idea of multilingual text summarization is to save time and complexity. Earlier summarizers worked for the single document, but now many documents can be feed to the machine as input and referred as multiple document summarizer. Tools are mainly designed by keeping the target reader in mind. It can be indicative, informative of query focused. Without giving much content, indicative summaries help in providing an idea about the text, while informative summaries provide a shortened version of the content. Necessary steps in creating a summary are:

• a)

  feed text input to the machine in one or more languages;

• b)

  pre-processing and feature extraction;

• c)

  sentence selection and assembly;

• d)

  summary generation in the desired language.

Figure 1 shows a generic architecture of text summarization. Text data is taken as input; important terms are extracted, and on their basis sentence ranking is done; high ranked sentences are selected, and low-rank sentences are rejected and then combined to form a summary; output can be in one (monolingual) language or different (multilingual) language.

The proposed chapter will be helpful to all the readers in understanding the basics of text summarization. To make this chapter enjoyable and to make concepts more precise, a proper flow will be maintained. Real-life examples will be given to clear the picture of text summarization. Pictorial explanations will support this chapter to make the chapter more interesting for the students, notably. Different Multilingual text summarization approaches will be discussed in detail, explaining the importance of each and difference among all. Comparison of different Multilingual text summarization tools will be made based on their accuracy (in percentage), shortcomings, and effectiveness. Queries like how to calculate accuracy and what are the parameters to calculate the accuracy will be covered in this chapter. While reviewing the tools, the dataset used will also be the significant concerned area for comparisons as efficiency and accuracy can be depicted based on the dataset used. The trend of text summarization will help the reader to know about the past, present, and future of text summarization.