Do We Need Neural Models to Explain Human Judgments of Acceptability?