We help minimize false positives and false negatives with a pre-processing pipeline to handle common errors with dependency parsing, lemmatization, POS-checking etc. We also improve entity recognition, concept/synonym recognition, etc.