Authors: Kui Lyu, Tianhao Tang, Minghao Wang

Description: In this work, we first used the two-stage text detector based on EAST and CRNN networks designed in Task1 and Task2 to locate text on scanned documents accurately. Then we designed a novel method of non-neural network model to extract the key information from the unstructured text list detected by our text detect. The core of the method is to used keyword positioning to obtain the location information of the scanned receipts from the unstructured text list and then perform local search find the key information.