Hi Sean and fellow Fast Dictionary Lookup fans, I notice that the UmlsJdbcRareWordDictionary doesn't seem to index terms from PREFTERM, only CUI_TERMS.
For instance, my custom dictionary script has the term "Angina Pectoris" as a PREFTERM, but "angina pectoris" isn't among the CUI_TERMS inserts: INSERT INTO PREFTERM VALUES(2962,'Angina Pectoris') INSERT INTO CUI_TERMS VALUES(2962,0,3,'ischemic chest pain','ischemic') INSERT INTO CUI_TERMS VALUES(2962,0,3,'ischaemic chest pain','ischaemic') INSERT INTO CUI_TERMS VALUES(2962,0,1,'angina','angina') INSERT INTO CUI_TERMS VALUES(2962,4,5,'pain ; chest , ischemic','ischemic') INSERT INTO CUI_TERMS VALUES(2962,0,2,'anginal discomfort','anginal') INSERT INTO CUI_TERMS VALUES(2962,4,5,'chest ; pain , ischemic','ischemic') INSERT INTO CUI_TERMS VALUES(2962,0,2,'anginal syndrome','anginal') INSERT INTO CUI_TERMS VALUES(2962,2,3,'syndrome ; anginal','anginal') INSERT INTO CUI_TERMS VALUES(2962,0,1,'stenocardia','stenocardia') INSERT INTO CUI_TERMS VALUES(2962,0,3,'anginal ; syndrome','anginal') INSERT INTO CUI_TERMS VALUES(2962,0,2,'angor pectoris','angor') INSERT INTO CUI_TERMS VALUES(2962,0,1,'stenocardias','stenocardias') INSERT INTO CUI_TERMS VALUES(2962,2,3,'chest pain ischemic','ischemic') INSERT INTO CUI_TERMS VALUES(2962,0,2,'anginal pain','anginal') INSERT INTO CUI_TERMS VALUES(2962,0,1,'anginas','anginas') So, in text containing the phrase "angina pectoris", the UmlsLookup annotators identify only the CUI_TERMS term "angina" as a SignSymptomMention. First off, am I missing something? I haven't used the default ctakessnorx.script dictionary for years. Is this a peculiarity of the custom dictionaries I've been building? Is there an option to include PREFTERMs in the rare-word index? Or is there some reason *not* to include PREFTERMs -- would they mess up the rare-word indexing somehow? Certainly many PREFTERMs would never occur in the wild (e.g. "Benign essential hypertension (disorder)"), but there are quite a few common clinical terms that are in PREFTERM but not CUI_TERMs. Off the top: C0017168 gastroesophageal reflux disease, C0018802 congestive heart failure, C0022104 irritable bowel syndrome, ... We've been adding these to a supplementary BSV file as they come up, but there are many more. This HSQL query for PREFTERM-only disorders on my custom dictionary returns 175K+ rows; at first blush, 20% look legit. select cui,lcase(prefterm) as prefterm from tui t join prefterm p on p.cui=t.cui and t.tui in (19,20,33,34,37,40,41,42,43,44,45,46,47,48,49,50,56,57,184,190,191) except (select cui,text from cui_terms c where c.cui=cui); Thanks for any thoughts, and happy holidays. Kean Kaufmann RecordsOne
