Formula search on a million scientific documents (arxivsearch.mathweb.org)

We, at KWARC, developed a search engine capable of searching arXiv, see [1]. It can search latex-style formulae and CMML formulae. In addition, one can also use "query variables", which means that if I search for, e.g., "?a + ?a" (?a is a query variable), I will get all sums of two equal terms. ?a could be an integral, a fancy function, or just a one-letter variabl