KnowSum: Knowledge Inclusive Approach for Text Summarization Using Semantic Allignment
Krishnan Nallaperumal, Gerard Deepak
Abstract
Text summarization plays an important role in delivering compact, most relevant, and efficient text to the user. It is also applied on the field of community question answers. There is a large amount of data on the internet pertaining to each topic. The question needs to be analyzed properly so that optimized, most relevant, and summarized text answer is generated. This paper proposes an ontology-based text summarization technique using Semantic Alignment and information gain along with LSTM and flower pollination algorithm. Here MS Marco Data set is used. From this for classifying question and answers LSTM is used. The top half of the data is only taken. With respect to each domain term from domain ontology feature extraction is done using information scent. Community question answer data such as Yahoo answers and Quora dataset are taken and classified. Both of these are then mapped together based on semantic alignment using flower pollination algorithm. After mapping, the answers are prioritized based on semantic similarity and information gain. Top 5 answers are chosen and summarized. The architecture's performance is calculated and compared with the baseline approaches and it is clearly observed that the proposed ontology-based text summarization technique is predominant in terms of performance and attained a precision and accuracy of 99.94% and 96.54 % respectively.