Text this: Can large language models detect drug–drug interactions leading to adverse drug reactions?