Optimized Abstractive Text Summarization using Semantic-Graph based Approach to Analyze Mental Health
Mental health is one of the primary concern for all physical human beings to perform socially. Generating abstractive text summary from health records of patients suffering from mental illness is the concern in this article. In order to provide precise and complete information, humans write long documents. While such documents are useful under many circumstances, often what is required is a shorter version of the document that conveys the important aspects of the document. Such shorter documents are easier and quicker for humans to process and enable one to get the gist of the information contained in the longer document. Text summarization is the process of obtaining the vital information from a source document and presenting the concise version to the reader. Automatic summarization can be broadly classified into two categories - extractive summarization and abstractive summarization. In extractive, based on certain heuristics or rules, certain units of the input are deemed to be important and returned as the summary. In contrast, the Abstractive method tries to generate a summary of the way a human would summarize. A human while summarizing does not necessarily chose a subset of the original sentences, he may condense the overall information in a succinct manner using words and sentences that may not be found in the original input. Thus, the abstractive approach is the more challenging form of automatic summarization. In this paper a novel approach is proposed towards abstractive summarization using a semantic-graph based model. A semantic graph will be generated by capturing nouns and verbs of a sentence available in several mental health records along with their topological relations. Then generated graph will be reduced by using set of heuristic rules. Further, by using domain ontology and WordNet summary will be generated for the reduced graph. The obtained results are the evidence to say that this is one of the optimal abstractive summarization approaches.