Redundancy Removal Method for Multi-Document Query-Based Text Summarization

We present RedunWSD, a Word Sense Disambiguation (WSD) based redundancy removal method for multiple text documents query-based text summarization. Recognizing and identifying the redundancy from the text documents is a challenging task that has not been fully resolved yet. In the case of multiple text document summarization processes, redundancy removal helps in getting maximum content coverage by minimizing any redundant or repetitive sentences. Another novel contribution of our work is the disambiguating a word's sense, since it helps us in getting accurate semantic similarity score between two sentences. Along with the semantic similarity score, we also consider the order of words between two sentences. Our model has the additional advantage of detecting the redundant sentences based on its meaning. Our experimental results show that the proposed WSD method outperforms many existing and current WSD methods. We have evaluated the proposed RedunWSD method on Document Understanding Conference (DUC) benchmark datasets and show that it helps in getting better recall values than many top existing query-based text summarization methods.