Added ICD 10 classification with attention (best results), attention with context (worst results) and char level classification (running atm).

# Conflicts:
#	code/data/train/FR/training/aligned/PavillonLaurent.pdf
#	code/data/train/FR/training/aligned/README.txt
#	code/data/train/FR/training/aligned/corpus/AlignedCauses_2006-2012full.csv
#	code/data/train/FR/training/aligned/corpus/AlignedCauses_2013.csv
#	code/data/train/FR/training/aligned/corpus/AlignedCauses_2013full.csv
#	code/data/train/FR/training/aligned/corpus/AlignedCauses_2014.csv
#	code/data/train/FR/training/aligned/corpus/AlignedCauses_2014_full.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2006-2010.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2011.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2012.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2013.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2014.csv
#	code/data/train/FR/training/aligned/dictionaries/Dictionnaire2015.csv
#	code/data/train/FR/training/aligned/eval/clefehealth2017Ta...

Initial version of representation learning model for eCLEF18-Task1. The model is able to distinguish between correct line-ICD10-code pairs (originating from the training data) and randomly sampled corrupted pairs.