Interpretable Document Representations for Fast and Accurate Retrieval of Mathematical Information

ABSTRACTA study conducted by the International Data Corporation predicted that by the year 2021, the total amount of digital information resources would have reached the 40 zettabyte mark [2]. According to a rule formulated by Merrill Lynch, 80 to 90% of these resources are unstructured [7]. Despite this, users expect digital libraries to provide them with fast and interpretable access to digital information resources that will satisfy their information need. Math information retrieval emerged as a subfield of information retrieval in 2008 [8], when it became clear that standard information retrieval techniques used for text documents are inadequate to accurately retrieve documents in digital mathematical libraries.