Text this: Language models for data extraction and risk of bias assessment in complementary medicine