I need to parse a Spanish corpus with syntactic dependencies. I have tried both Freeling and Maltparser, they both give similar results, but with some divergence. I need the highest accuracy possible, mainly in the dependencies between the verb and its dependents. Which one should I use?
I suggest to use more than one tool.
When you have three different tools you can use majority voting to determine your tags. In case of ties, a human arbiter can fix the tags.
Also note that the tools can be trained. When you can afford to create some tailored training data from your corpus you can improve the accuracy of the tagging significantly.