A Deep Network Model for Paraphrase Detection in Punjabi

Abstract

Paraphrase refers to the text which tells the same meanings but with different expressions. It is important in NLP as it deals with many applications such as information retrieval, information extraction, machine translation, query expansion, question answering, summarization and plagiarism. Paraphrase detection is to find that given two texts are semantically similar or not similar. Though paraphrase detection has wide literature, there is no proper algorithm for paraphrase detection in Punjabi language. A new paraphrase detection model for Punjabi language is developed in this paper. We use two deep learning methods to map sentences as vectors, and these vectors are further used to detect paraphrases. Despite other implementations of paraphrase detection, our model is simple and efficient to detect paraphrases. Qualitative and quantitative evaluations prove the efficiency of the model and can be applied to various NLP applications. The proposed model is trained on Quora’s question pair dataset which makes new directions for paraphrasing in Indian languages.