I have planned to develop a prototype for academic purpose. I have list of snomed id with corresponding medical term. Some doctor write medical term in short expression so naturally those words have no standard snnomed id. my job is to predict and suggest the close standard snomed id medical term from those free text. I am now experimenting with gate software's Annie gazetteer.
I need some alternate suggestion.
I am new in nlp and machine learning
In the Snomed community there is a way described to create an index of all possible three, four, five, etc letter combinations so you can find possible fits quickly. For example: "emergency appendectomy" would result in an index having eme, mer, erg, rge, gen, enc, ncy, cy , y a, ap, ppe, etc. So when someone starts typing, from start or somewhere in the middle, after three characters you are able to present a list of terms. But the index will be huge and will cost time to create.