Twitter Sentimental Analysis

Twitter Sentimental Analysis

I would like to know, if I have 3 labels (positive,negative, and neutral )
for sentimental analysis for twitter , and this dataset (Dataset1) is
unbalanced 1000 records for neutral, 570 for positive, and 450 for
negative.. I combined it with another dataset Dataset2 to get balanced,
For example, 1000 neutral from Dataset1 1000 positive from Dataset2 1000
negative from Dataset2
and make classification ,I tested it and I found that the accuracy is low,
so my question is : Does merging/combining dataset like mentioned before
affect the classifier ?, or I have to investigate in another direction..
Also if anyone knows where I can find balanced database for twitter
sentimental analysis including (pos,neg, neutral), it will be better to
use it once.