A Relative Study of Neural Network and Fuzzy Logic Systems in Multi Lingual, Multi Document Textual Summarization

Since a long period of time, the internet has shown tremendous expansion which has led to increased volume of information accessible on the Web. One such type of digital information available is in the form of online news stories and articles. Summarization systems have been designed for people who do not have much time to invest in reading every news article in detail. These systems generate a brief gist of the news story so that the reader gets an overview of what is there in the complete article. Such systems can be used to generate headlines of any news article. The paper proposes two different machine learning algorithms applied to multiple documents containing the same news story and then combining all important sentences from the various documents into one short summary. The algorithms used are neural networks and fuzzy logic systems and they are being compared based on certain statistical and linguistic criteria. The focus is on two very common languages used for communication in India i.e. English and Hindi, making this system a multi linguistic, multi document text summarization system. It is observed that in comparison to neural networks, fuzzy logic systems outperformed in achieving improved results. When compared with existing summarization systems, an average improved precision of 8–9% for Hindi and 45–48% for English text documents has been achieved.