entiti categor larg document collect venkatesh ganti arnd nig rare vernica extract entiti peopl movi document identifi categori painter writer belong enabl structur queri data analysi unstructur document collect paper focu problem categor extract entiti prior approach develop task analyz local document context entiti occur paper significantli improv accuraci entiti categor entiti context multipl document ii exploit exist larg list relat entiti list actor director book approach introduc comput challeng context entiti aggreg document list relat entiti larg develop techniqu address challeng present experiment studi real data set demonstr increas accuraci scalabl approach 