A Novel Framework for Semantic Oriented Abstractive Text Summarization

Internet continues to be the most communal of all the mass media turning information content from being scarce to superabundant that is evidenced by its increase tenfold every five years. A powerful text summarizer can aid in balancing this overload but generating quality summaries of the target content thereby reducing time and effort to mine the required information. The proposed system aims is to develop a Semantic Oriented Abstractive Summarization to generate abstractive summaries with increased readability and qualitative content. The contribution of our works are Joint Model Predicate Sense Disambiguation and Semantic Role Labelling termed as Joint (PSD+SRL) is proposed to better capture the semantic representation of text. The content selection involves semantic based content selection and feature extraction are selected by Genetic Algorithm. The proposed system can be very useful for the students who want to read a whole book in a short time. Our experimental study is carried out using DUC, a typical corpus for text summarization.