Paraphrase detection is presented as an emerging research area under Natural Language Processing (NLP). The system assists in determining the semantic and lexical similarities between the user’s multiple sentences. If both sentences have the same meaning, then the penalties will be denoted as paraphrases to each other, otherwise, they will not be paraphrases to each other. As paraphrasing removes the communication barrier and helps people to express their feelings more readily, it takes part in vital tasks in different domains, like healthcare, teaching, plagiarism detection, research, and many other fields. But still, there are many limitations of paraphrasing. One of the most important challenges of paraphrasing is usually when the writer replaces just one or two words in the source phrases with synonyms. This type of paraphrasing does not show sufficient understanding and engagement with the text. Instead, the writer must try to take ideas and information and put them into his own words. To overcome these challenges, the paraphrasing model should be more developed, so that there is no over-fitting or under-fitting situation, and the dataset should be more enriched with various regional usages of sentences. This article deals with different paraphrase detection procedures with the main goal of analyzing different solutions using neural methods.