Disease normalizer isn't being given entire input text, only the keyword

[mukund109]

On scanning through the code I can see that you don't seem to be giving the sieve-based normalizer the full abstract as input, only the keyword (see here).

In that case how does it do abbreviation detection? Or is that being skipped?

Abbreviation detection is responsible for >5% of the disease normalizer's accuracy so would be great if you could clarify :smile:

On another note, thanks a lot for this repo, has been very useful.