Most previous studies of document-level event extraction mainly focus on\nbuilding argument chains in an autoregressive way, which achieves a certain\nsuccess but is inefficient in both training and inference. In contrast to the\nprevious studies, we propose a fast and lightweight model named as PTPCG. In\nour model, we design a novel strategy for event argument combination together\nwith a non-autoregressive decoding algorithm via pruned complete graphs, which\nare constructed under the guidance of the automatically selected pseudo\ntriggers. Compared to the previous systems, our system achieves competitive\nresults with 19.8\\% of parameters and much lower resource consumption, taking\nonly 3.8\\% GPU hours for training and up to 8.5 times faster for inference.\nBesides, our model shows superior compatibility for the datasets with (or\nwithout) triggers and the pseudo triggers can be the supplements for annotated\ntriggers to make further improvements. Codes are available at\nhttps://github.com/Spico197/DocEE .\n