Discovery and recognition of formula concepts using machine learning

Abstract Citation-based Information Retrieval (IR) methods for scientific documents have proven effective IR applications, such as Plagiarism Detection or Literature Recommender Systems in academic disciplines that use many references. In science, technology, engineering, and mathematics, researchers often employ mathematical concepts through formula notation to refer prior knowledge. Our long-term goal is generalize citation-based apply this generalized method both classical references concepts. paper, we suggest how formulas could be cited define a Formula Concept task with two subtasks: Discovery (FCD) Recognition (FCR). While FCD aims at the definition exploration of ‘Formula Concept’ names bundled equivalent representations formula, FCR designed match given assigned unique concept identifier. We present machine learning-based approaches address tasks. then evaluate these on standardized test collection (NTCIR arXiv dataset). approach yields precision 68% retrieving frequent recall 72% extracting name from surrounding text. enable citation within facilitate semantic search question answering, well document similarity assessments plagiarism detection recommender systems.