Extractive Text Summarization Using Supervised Learning and Natural Language Processing

The amount of textual data that we are exposed to is growing each day. It is very difficult to browse through all the available textual matter to find relevant material or to read through all the information in order to stay updated. To keep up with the pace, the need for a tool that can automatically reduce the amount of content while also retaining the key points and essence of long pieces of text arises. Automatic text summarization mechanisms form a solution well suited to this problem which is what our proposed model aims to implement. In this paper, a Natural Language Processing based extractive approach is used for summarization of a single document. An extractive summary is assembled by selection of a subset of information rich sentences from the source document. A supervised approach is used here in which Support Vector Machine, K-Nearest Neighbour and Decision Tree algorithms are used to generate models whose performances are compared using ROUGE metric. The highest scoring model is used to summarize an unseen document. The summary is displayed as text and converted to audio form. The results obtained using the proposed approach are sufficiently good as average F1 scores secured for ROUGE-1, ROUGE-2 and ROUGE-L are 0.706, 0.630 and 0.434 respectively.