- The rules were implemented in MinorThird [29] and we use its notation here. Only the part in brackets (the string of interest) is being extracted as a mention (i.e., annotation); the rest of the rule (if any) specifies the context/anchors. The rules use explicit matching of spans (e.g., eq(‘times’)), the dictionary matches for single (e.g. a(verb) – matching verbs that indicate the administration of a medication (e.g., take, insert)) and multiword terms (e.g., @period, see Additional file 1: Table S2). Number models numerical expressions including those belonging to the dictionary “number”; a(timeUnitLy) matches the words of the dictionary “timeUnitLy” with prescription text that indicates an adverb of time e.g., “daily”, “weekly”, etc; @perTimeUnit recognises syntactical patterns of the dictionary “perTimeUnit” that contain both numeric and word phrases in prescription text e.g., “four times a day”, “2 times per week”, etc.; a(timeUnit) identifies the words from the “timeUnit” dictionary (see Additional file 1: Table S2)
