* merge_entities sets the vector in the vocab for the merged token * add unit test * import unicode_literals * move code to _merge function * only set vector if vocab has non-zero vectors
Sofie Van Landeghem