Into the crossfire: evaluating the use of a language model to crowdsource gun violence reports