Search

This paper compares the search capabilities of a single human brain supported by the text search built into Wikipedia with state-of-the-art math search systems. To achieve this, we compare results of manual Wikipedia searches with the aggregated and assessed results of all systems participating in the NTCIR-12 MathIR Wikipedia Task. For 26 of the 30 topics, the average relevance score of our manually retrieved results exceeded the average relevance score of other participants by more than one standard deviation. However, math search engines at large achieved better recall and retrieved highly relevant results that our ‘single-brain system’ missed for 12 topics. By categorizing the topics of NTCIR-12 into six types of queries, we observe a particular strength of math search engines to answer queries of the types ‘definition lookup’ and ‘application look-up’. However, we see the low precision of current math search engines as the main challenge that prevents their wide-spread adoption in STEM research. By combining our results with highly relevant results of all other participants, we compile a new gold standard dataset and a dataset of duplicate content items. We discuss how the two datasets can be used to improve the query formulation and content augmentation capabilities of match search engines in the future.