Abstract：COMPAREA Taxonomy and Dataset of Comparison Discussions in Peer Reviews Comparing research papers is a conventional method to demonstrate progress in experimental researchWe present COMPAREa taxonomy and a dataset of comparison discussions in peer reviews of research papers in the domain of experimental deep learningFro Retrieval and IndexingAs the research focuses on comparative studies of research papersthis can be used to improve the retrieval efficiency of materials in the libraryespecially for digitized theses and journalsUsing these technologiesit is possible to find papers related to specific topics or keywords more quickly and accuratelyInstitutional CollectionThis technology can be used to analyze and organize the institutional collections of the libraryespecially when the collection involves a large number of research papersWith such a systemit is possible to more effectively compareclassifyand annotate the literature in the collectionCollection ClassificationUsing deep learningThis article introduces COMPAREa taxonomy and dataset for comparative discussion in peer reviews of research papers in experimental deep learningFrom detailed observations of numerous review sentencesthe authors established a taxonomy of comparative discussion categories and proposed a detailed annotation scheme to analyze these discussionsThe article also experimented with various methods to identify comparative sentences in peer reviewsreporting the highest F1 score of 049Additionallythey pre-trained two language models on MLNLPand CV paper abstracts and reviews to learn meaningful representations of peer reviews